DEVELOP3D workstation special report - August 2021 by X3DMEDIA

AUGUST / SEPTEMBER 2021 | DEVELOP3D.COM

Workstation special report for CAD, design viz, reality modelling and more THE LATEST

Pro

GPUs FROM A AND NVIDIAMD REVIEWED

THE BATTLE OF THE CPUs WHAT DO INTEL AND AMD’S NEW PROCESSORS REALLY MEAN FOR CAD-CENTRIC WORKFLOWS? WS01_D3D_AUGSEPT21_Cover.indd 1

DESKTOP WORKSTATIONS HOW TO BUY A WORKSTATION + OUR TOP PICKS OF THE BEST MACHINES FOR CAD & BEYOND

BEST LIGHTWEIGHT LAPTOPS ULTRA-PORTABLE MOBILE WORKSTATIONS TO TAKE CAD AND DESIGN VIZ ON THE ROAD 12/08/2021 09:27

workstation special report

Desktop workstation buyer’s g Greg Corke goes back to basics with some general advice for those looking to buy a workstation for product development workflows If you read DEVELOP3D, the chances are that you already own a workstation. But how much do you know about it and how involved were you in its purchase? If you leave everything to your general IT department, you could end up with a machine that slows you down or, worse still, is simply not able to run your more demanding applications. We all work within budget limits, so it is really important to spend your money in the right areas. We’ve heard plenty of horror stories, including that of one large engineering firm that spent loads on dual CPU workstations with lots of cores that ran at a low frequency (GHz), even though its designers only ever used CAD. In that example, buying workstations with a single high-frequency CPU with fewer cores would not only have saved money, but significantly increased the productivity of the design team. Choosing the right spec is very important – and we cover this in depth in this special report – but it is not the be all and end all of buying a workstation. In this article, we look at other important points to consider.

Warranty Virtually all workstations come with a three-year warranty, but the level of cover can vary considerably between manufacturers. This isn’t just about protecting your investment. In the event of a

HP Z2 Tower G8 running Solidworks

WS2

August / September 2021

WS02_03_D3D_AUGSEPT21_Buyers_guide.indd 2

failure, you need to get machine. Warranties can be your workstation back up extended and negotiated, so and running as quickly as also bear that in mind. possible and with minimal The chassis hassle. A next business day (NBD) Workstations are not just on site warranty is common about their constituent parts. with workstations from The chassis can be incredibly major OEMs. Some custom important. There are three manufacturers only offer main points to consider: size, ‘return to base’ as standard noise and serviceability. But or NBD on site for the first having easy access to things Having to send like USB ports or headphone year. Having to send away away your your workstation to be sockets can also make a big repaired could mean days workstation to difference. Ask yourself the without its use, which could following questions: Will the be repaired be catastrophic when on a machine fit on your desk, or could mean tight project deadline. will it have to be kept down Many minor repairs can days without it, by your feet? If you need to be done by yourself. Some which could be move it, is it heavy and does manufacturers specifically it have built-in handles? catastrophic make their workstations Are the fans so noisy that when on a tight they are distracting? Can easy to service, so customers can be guided over the project deadline you expand the system with phone or video call and GPUs, storage or memory get themselves back up (are there free memory and running the same day (or the next slots)? Is it easy to service internal day if a part needs to be sent out). Other components? manufacturers state that any repairs or Of course, if you are the type of person alterations that are carried out by the that would never dream of getting inside a customer invalidate the warranty, so computer then serviceability is a moot point. check with your supplier first. Think about the things that matter to you. When comparing prices of workstations, For some, aesthetics are very important. make sure you are looking at details of the warranty and not just the Custom manufacturer or OEM specifications of A frequently asked question is whether t h e to buy from a custom manufacturer or a major OEM like Dell, HP, Lenovo or Fujitsu? There’s no simple answer, but there are a number of things to consider. Custom manufacturers are more flexible in how machines can be configured and are generally quicker at introducing new technologies. Many such firms overclock their CPUs, which can boost performance. OEMs, on the other hand, spend much more time on R&D to make

‘‘

’’

www.develop3d.com

12/08/2021 14:05

workstation special report

s guide sure that components work well together and do extensive studies into acoustics, electromagnetic interference and thermodynamics. OEM workstations tend to be more solid and easier to service, particularly as some custom manufacturers try to cut costs by using budget PC chassis. OEM workstations also go through extensive testing and certification to make sure they work well and will be fully supported with many different CAD applications. With custom manufacturers, you are usually just getting the certification of the GPU. Of course, many large design and engineering firms will only buy from major OEMs for reasons of support, management or global availability, so the decision is already made.

Performance Nobody ever complains that their workstation is too fast, but the additional money one must spend to increase performance needs to be weighed up against the benefits experienced. Some performance increases are easy to quantify, such as the time it takes to render a scene or process a point cloud, so it is easy to envisage the potential benefits. 3D graphics performance, however, is harder to measure. More frames per second (FPS) is always better, but if you are unable to discern a difference, does it really matter? Generally speaking, anything over 25 FPS is fine for 3D design work. You can get away with lower frame rates on the desktop, but Virtual Reality is different. An underpowered GPU could make you feel sick as it can’t keep up with your head movements – or it may mean you have to spend hours manually optimising CAD models every time you bring them into VR.

Know your bottlenecks Understanding how your software works and where your bottlenecks occur can help you make informed decisions on workstation purchases. Many of the major workstation OEMs offer free workstation tuning software that can monitor resources in real time and over a set period. Other good tools include GPU-Z for monitoring GPU usage, CPU-Z for CPU and Windows Performance monitor. If you are experiencing slow 3D performance, for example, check to see how much of your GPU is being used. If it’s 100%, then investing in a more powerful GPU would probably help. However, if usage is significantly less, you could be wasting your money and would be better off buying a higher frequency CPU. Once you understand where your bottlenecks occur, it can help you assign budget to the correct areas and tailor a workstation for your specific workflows. Read this DEVELOP3D article to learn more (develop3d.com/hardware/know-your-bottlenecks). www.develop3d.com

WS02_03_D3D_AUGSEPT21_Buyers_guide.indd 3

Processor (CPU) In a workstation, you should always aim for a CPU with a high frequency (GHz). This is good for general system and modelling performance as many operations are single threaded. i.e. they use only one CPU core. Multi-threaded operations can use multiple CPU cores (and sometimes virtual CPU cores). Rendering is the best example; as a rule of thumb, doubling the number of cores halves rendering time. Simulation software and point cloud processing software also tends to be multithreaded, but there can be diminishing returns as you use more cores. In saying that, it is possible to do multiple simulations on a single machine at the same time, providing you have sufficient memory and memory bandwidth, storage bandwidth and software licences. CPUs that have lots of cores typically run at lower frequencies, so it’s important to strike

Graphics (GPU) The professional 3D graphics card or graphics processing unit (GPU) is one of the key components that defines a professional 3D workstation. Unlike consumer GPU drivers, pro graphics drivers are specifically designed to work with professional 3D software (especially CAD software) and are given an official stamp of approval for each application through ‘certification’. This comes with the promise of full support from the software developer. Pro drivers can mean better performance, better stability and access to specific features, such as RealView in Solidworks, Order Independent Transparency (OIT) in PTC Creo, and AMD Radeon Viewport Boost (see page WS26). If there are any display or stability issues in your CAD software, they are more likely to be fixed in a subsequent driver release. Nvidia has the lion’s share of the add-in GPU market with its Nvidia RTX and Nvidia Quadro family, but AMD Radeon Pro also plays

Storage (SSD / HDD) M.2 NVMe Solid State Drives (SSDs) have quickly become the standard in workstations. Sequential read/write speeds are superior to 2.5-inch SATA SSDs, although this won’t always result in real world benefits. M.2s are also smaller, which has helped reduce the size of workstations. Newer workstations, including those with AMD Ryzen 5000 or 11th Gen Intel Core CPUs, support PCIe Gen 4 NVMe SSDs, which offer double the sequential read/write performance

Memory (RAM) You should always aim to have enough memory so you never run out, as performance can significantly slow down if your workstation has to page data to your SSD. Memory requirements will always change over time. Your datasets will get bigger and the

a balance. Another strategy is to optimise your workstation for modelling and send your multi-threaded calcs to a server or the cloud. Intel used to dominate the workstation CPU market, but this is changing. AMD now has extremely competitive processors for all different workflows. However, with the exception of Lenovo, AMD-based workstations are still only available from specialist manufacturers like Scan, BOXX, and Workstation Specialists. For CAD centric workflows, choose 11th Gen Intel Core or Intel Xeon W-1300 CPUs (up to 8 cores) or AMD Ryzen 5000 series (up to 16 cores). Intel still has the edge in single threaded performance, but AMD offers much better multi-threaded performance as it has double the number of cores (see page WS4). For very high-end multi-threaded workflows, it’s very hard to recommend anything other than AMD Threadripper or AMD Threadripper Pro. It’s important to note that CPUs from different families cannot be directly compared by GHz.

a very important role. Some Intel CPUs include integrated graphics. While the performance can be OK for entry-level 3D workflows, certain professional features might not be supported and the number of application certifications is much less. Still, professional GPUs are facing increased competition from their consumer counterparts — Nvidia GeForce and AMD Radeon. Some of the major workstation OEMs now offer consumer GPUs in their workstations — both desktop and mobile. High-end consumer GPUs tend to feature less memory than high-end professional GPUs but can easily compete on raw processing power. For CAD workflows we still recommend sub-£500 entry-level to mid-range pro GPUs, but when you get into the high-end, an area where you need more 3D performance for real-time visualisation, GPU rendering or virtual reality (VR), the difference in price is greater and pro graphics cards are often a harder sell for those on a budget. Nvidia even has a GeForce ‘Studio’ driver for applications including Unreal Engine, V-Ray and Enscape.

of PCIe Gen 3 models. Those working with very large datasets, such as point cloud, simulation or video editing, should benefit most. The cost of NVMe SSDs has come down a lot but Hard Disk Drives (HDDs) continue to offer the best price per GB. HDDs should absolutely not be used as a primary drive for operating system and applications, but they are good for secondary storage and do not impact load / save times of CAD models that much. However, if you need to process large point cloud or simulation datasets, especially if you don’t have enough RAM to hold the data entirely in system memory, then NVMe SSDs are always best.

memory footprint of applications will increase with new releases. With this in mind, it’s good to buy a workstation with spare RAM slots for easier upgrades. However, for best performance, memory should be installed in pairs, quads or eights, matched to the number of CPU memory channels. ECC memory can protect against crashes, which becomes important for lengthy calculations as you can lose hours of work.

August / September 2021

WS3

12/08/2021 14:05

SOME BUILDING PROJECTS WON’T MOVE AHEAD WITHOUT YOU Get the ultimate software experience using the AMD Radeon™ PRO W6000 graphics series, offering high-performing hardware raytracing, lightning-fast framebuffers, optimizations for up to 6 Ultra-HD displays, and superior multitasking capabilities. All wrapped around an award winning graphics architecture, called AMD RDNA™ 2 which is the established graphics foundation for the leading, visually rich games consoles. Allowing you to focus on those projects that require even more from you. For everything else let AMD Radeon PRO graphics help.

amd.com/RadeonPRO

R E A D Y C R E ATO R

workstation special report

WS4

August / September 2021

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 4

www.develop3d.com

12/08/2021 14:38

workstation special report

Intel Core vs AMD Ryzen for CAD and beyond With 11th Gen Intel Core and AMD Ryzen 5000, competition in workstation CPUs has never been so strong. Greg Corke explores the best CPUs for designcentric workflows from CAD to reality modelling and rendering

ne of the questions we get asked most often at DEVELOP3D is ‘which processor is best for CAD?’ Normally, we wouldn’t hesitate in recommending Intel Core or low core count Intel Xeon processors (after all, Intel has demonstrated a clear lead in single threaded performance for the last fifteen years) but with AMD’s recent resurgence in the CPU market, things are no longer so clear cut. In October 2020 AMD launched its AMD Ryzen 5000 Series, based on its 7nm ‘Zen 3’ architecture, and finally took the performance crown from Intel. With more cores (16 vs 10), high clock speeds and superior Instructions Per Clock (IPC), AMD Ryzen 5000 outperformed 10th Gen Intel Core in both single threaded and multi-threaded workflows. It quickly became the processor of choice for users of CAD applications like, Inventor and Solidworks, as well as CPU ray trace renderers like V-Ray and KeyShot, or those built into the CAD applications themselves. Well — it did for those who could

AMD Ryzen 5000 There are four AMD Ryzen 5000 processor models, differentiated largely by the number of cores – 6, 8, 12, or 16. This is the same number of cores offered on the previous Ryzen 3000 series, but the IPC improvement is significantly higher – also 19% generation over generation. The top-end AMD Ryzen 9 5950X has 16 cores and 32 threads. It has the lowest base frequency (3.4 GHz) but the highest boost frequency (4.9 GHz), making it very well suited to both single threaded and multithreaded workflows. The other three models in the range are the Ryzen 5 5600X (6 cores, 3.7 GHz, 4.6 GHz boost), the Ryzen 7 5800X (8 cores, 3.8 GHz, 4.7 GHz boost), and the Ryzen 9 5900X (12 cores, 3.7 GHz, 4.8 GHz boost). See table below for the full specs.

11th Gen Intel Core Intel has significantly more models in its 11th Gen Intel Core family. These are differentiated mostly by power draw and base frequency, and less so by the number of cores. With 11th Gen Intel Core you only

AMD Ryzen 5 5600X

AMD Ryzen 7 5800X

AMD Ryzen 9 5900X

AMD Ryzen 9 5950X

Intel Core i5-11600

Intel Core i5-11600K

Intel Core i7-11700

Intel Core i7-11700K

Intel Core i9-11900

Intel Core i9-11900K

# of CPU Cores

# of CPU Threads

3.70 GHz

3.80 GHz

3.70 GHz

3.40 GHz

2.80 GHz

3.90 GHz

2.50 GHz

3.60 GHz

2.50 GHz

3.50 GHz

Max Boost Frequency

4.60 GHz

4.70 GHz

4.80 GHz

4.90 GHz

4.80 GHz

4.90 GHz

5.00 GHz

5.20 GHz

5.30 GHz

Cache

L2 Cache 3 MB L3 Cache 32 MB

L2 Cache 4 MB L3 Cache 32 MB

L2 Cache 6 MB L3 Cache 64 MB

L2 Cache 8 MB L3 Cache 64 MB

12 MB Intel Smart Cache

16 MB Intel Smart Cache

Default TDP / TDP

65 W

105 W

65 W

125 W

65 W

125 W

65 W

125 W

Price (Ex VAT)*

£229

£317

£424

£574

£176

£192

£262

£292

£344

£441

Base Frequency

*PRICE TAKEN FROM SCAN.CO.UK ON 5/7/21

actually get hold of one. The global chip shortage meant demand massively outweighed supply and even if AMD Ryzen 5000 CPUs were in stock, prices were sometimes hugely inflated. This bought Intel some time, and in March 2021 the chip giant hit back with its 11th Gen Intel Core desktop processor family (code-named Rocket Lake-S). Based on its ageing 14nm manufacturing process, expectations were low. However, the new chip family came with the promise of a massive 19% IPC performance improvement, generation on generation. In other words, if a single core on a 10th Gen and 11th Gen Intel Core CPU ran at the same frequency, the 11th Gen would be 19% faster. This was obviously big news for users of CAD software for whom single threaded performance really matters, and it immediately put pressure back on AMD. However, the single threaded boost came at the expense of multi-threaded performance. 11th Gen Intel Core had two fewer cores than 10th Gen Intel Core, so rendering times actually went up.

www.develop3d.com

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 5

August / September 2021

WS5

12/08/2021 14:38

workstation special report

get a choice of 6 or 8 cores. identical to 11th Gen Intel Core processors, All Core i5 models have 6 cores and 12 but include support for Error Correcting threads, while the Core i7 and Core i9 Code (ECC) memory. models have 8 cores and 16 threads. As Finally, it’s important to note that you mentioned previously, this is a step down can’t directly compare AMD and Intel from the 10th Gen Intel Core processors CPUs in terms of their frequency. It’s not which offer up to 10 cores and 20 threads. just the GHz of the CPU that matters, but With fewer cores than the AMD Ryzen the Instructions Per Clock (IPC) that it 9 5900X and Ryzen 9 5950X, it’s very can execute. You should only compare hard for Intel’s 11th Gen Core processors frequency when trying to decide between to compete in highlytwo CPUs from the same threaded workflows like series, although you should rendering. It means the get a fairly good idea When it comes still main battleground for Intel across families. to rendering is in single threaded or more On test lightly threaded workflows, AMD Ryzen centring on CAD. In an ideal world we would 5000 wins There are three main have tested all different hands down. processors — the topmodels of Intel and AMD With the 11th end Intel Core i9-11900K CPUs, but the time it takes (3.5 GHz base, 5.3 GHz to test with real-world Gen Core Turbo), the Intel Core applications (and processor processors i7-11700K (3.6 GHz, 5.0 GHz availability) meant this was maxing out at not viable. Turbo), and the Intel Core i5-11600K (3.9 GHz, 4.9 GHz eight cores Intel Instead, we focused on Turbo), all of which can be two different workstations simply can’t overclocked (see later). — the Scan 3XS GWP-ME compete Intel also offers non “K” A132R with the top-end versions, which are locked AMD Ryzen 9 5950X (16 so can’t be overclocked, and cores) and the Workstation have a lower base frequency, a slightly Specialists WS-184 with the Intel Core lower turbo frequency and draw less i9-11900 (8 cores). Unfortunately, we power. There are also slightly cheaper were unable to get hold of an ‘unlocked’ “F” variants which have the integrated Intel Core i9-11900K which has a higher graphics disabled. With most CAD and base clock and a slightly higher boost design viz workstations featuring discrete frequency (see table on previous page). pro GPUs from AMD or Nvidia this could We’re aware that this is a bit of an save a bit of cash. ‘apples and pears’ comparison, so the Major workstation manufacturers like results should be interpreted accordingly. Dell and HP tend to offer all three variants The AMD Ryzen 9 5950X has double in their machines, as well as Intel Xeon the number of cores of the Intel Core i9W-1300 series CPUs, which are virtually 11900, so is obviously better suited to

‘‘

’’

highly-threaded workflows like rendering. It also costs around 65% more (£574 vs £344 Ex VAT). However, in terms of single threaded workflows, the two workstations should give a good indication of relative performance between the two processor families. It should also be noted that different results might be seen with different motherboards and memory configurations. For consistency, both machines featured 64 GB (2 x 32 GB) of 3,200 MHz memory and a 2 TB Samsung 980 Pro NVMe PCIe 4.0 SSD. To streamline our extensive testing, different GPUs were used for different tests, but the same GPU was always used in both machines. The full specs can be seen on page WS9.

CAD and beyond CAD tools like Solidworks and Autodesk Inventor and Building Information Modelling (BIM) software like Autodesk Revit are bread and butter tools for designers, engineers and architects. In the main they are single threaded and while some processes can use more than one CPU core, it’s only usually ray trace rendering that can take full advantage of all the processor cores, all the time. In Autodesk Revit 2021, the RFO v3 benchmark (fig. 3) showed a clear lead for the Intel workstation in model creation and export, although the AMD workstation predictably won out in rendering. We saw similar results in Solidworks with the SPECapc 2021 benchmark (fig. 1), although Intel’s lead was extended considerably in ‘model rebuild’. We also did some manual tests in Solidworks 2021 (fig. 2) and the AMD workstation closed the gap, coming in

Solidworks appears to run fastest on 11th Gen Intel Core

WS6

August / September 2021

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 6

www.develop3d.com

12/08/2021 14:38

workstation special report

Reality modelling Reality modelling is becoming much more prevalent in the AEC sector. Agisoft Metashape is a photogrammetry tool that generates a mesh from multiple hi-res photos. It is multi-threaded and uses a combination of CPU and GPU processing. We tested using a benchmark from specialist US workstation manufacturer Puget Systems (fig. 8). In most of our tests Intel had a clear lead, but AMD did claim top spot in one test. Interestingly, Puget Systems reports that the AMD Ryzen 7 5800X, which has half as many cores as the AMD Ryzen 9 5950X but a higher base frequency (3.8 GHz), is faster in three out of the four tests. In fact, it even beat Intel. In point cloud processing software Leica Cyclone Register 360 (fig. 9), which can run on up to 6 CPU threads (as long as the workstation has sufficient memory), the AMD machine had the lead when registering both of our point cloud datasets. www.develop3d.com

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 7

Processor frequency in relation to number of CPU threads used (using Cinebench custom run)

6.0 5.8

Processor frequency (GHz)

a touch behind the Intel workstation in model opening and IGES export. Storage does play a part in these processes, but the CPU still plays the major role by far. We also tested with the new InvMark for Inventor 2022 benchmark by Cadac Group and TFI (fig. 7). Intel had the lead in most sub tests, apart from the solid sweep modelling operation and (of course) rendering. Interestingly, while Task Manager indicated that some processes are multithreaded (for example, data translation, drawings, and FEA) multiple cores are either used sparingly or only for short bursts in combination with single threaded processes, so overall there’s no real benefit to having more than eight cores. Here, it’s worth pointing out the chart to the right where you can see how frequency drops as more cores (threads) are enabled. Between 12 and 16 threads, Intel maintains a higher frequency than AMD, so this could help give Intel a lead in workloads that use a similar number of threads. Interestingly, even though the Intel Core i9-11900 has a moderate base frequency of 2.5 GHz, the workstation’s cooling lets it maintain a 4.6 GHz boost on all eight cores. Of course, the central code in many CAD tools is quite old and new generation tools like nTopology, which focus on generative design, are built from the ground up for multi-core processors (and, more recently, GPU computation). For our nToplogy geometry optimisation test (fig. 4) we solely focused on the CPU and, with all cores in use, the AMD workstation had a clear lead.

5.6

Workstation Specialists WS-184 (Intel Core i9-11900)

5.4

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

5.2 5.0 4.8 4.6 4.4 4.2 4.0 3.8 3.6 3.4

Number of threads

Rendering There were no surprises when ray trace rendering with V-Ray, Cinema4D and KeyShot, or any renderer built-in to our CAD and BIM tools (fig 5/6). With twice as many cores as Intel, the AMD workstation was always going to come out on top, even if its cores run at a lower frequency. We also did stress tests to see if CPU frequency dropped over time. When rendering in Cinebench for over an hour, the Intel maintained 4.59 GHz and the AMD maintained 3.74 GHz, testament to the excellent cooling in both workstations.

Scan 3XS GWP-ME A132R » AMD Ryzen 9 5950X CPU » 64 GB (2 x 32 GB) Corsair Vengeance DDR4 3,200 MHz memory » 2 TB Samsung 980 Pro NVMe PCIe 4.0 SSD » Asus Pro WS X570-ACE motherboard » Noctua NH-D15 air cooler » Full review on page WS14

Graphics CPU frequency has some influence over graphics performance, but the extent to which it does depends on the application. In Revit, Solidworks and Inventor — applications that tend to be more CPU limited — Intel showed around a 10% Frames Per Second (FPS) performance lead over AMD when using the same GPU (fig. 10). This went to down to 5% in Solidworks when RealView, Shadows and Ambient Occlusion (AO) were enabled — graphics effects that place much bigger demands on the GPU. In Unreal Engine 4 — an application renowned for being GPU limited, rather than CPU limited — the difference between Intel and AMD was negligible with our viz test scene.

Workstation Specialists WS-184 » Intel Core i9-11900 CPU » 64 GB (2 x 32 GB) 3,200 MHz dual channel Corsair Vengance RGB Pro DDR4 memory » 2 TB Samsung 980 Pro NVMe PCIe 4.0 SSD » Asus Prime Z590-P motherboard » Be Quiet liquid cooler » Full review on page WS15

Multi-tasking These days, very few architects, engineers or product designers use single applications, and with compute intensive workflows on the rise, such as reality modelling, rendering and simulation, it’s very important to consider multi-tasking when choosing a CPU. Even with 8 cores, it’s possible to leave August / September 2021

WS7

12/08/2021 14:38

workstation special report

one or more multi-threaded tasks running in the background, and still leave resources free for bread and butter 3D modelling tasks. To explore multi-tasking potential, we pushed both machines to their limits, with a demanding AEC workflow, consisting of point cloud processing and photogrammetry (fig. 11). We registered a 24 GB point cloud dataset in Leica Cyclone Register 360 while processing a series of high-res photographs in Agisoft Metashape at the same time. If done sequentially it would have taken 1,218 secs on the AMD machine and 1,207 secs on the Intel, but running both jobs in parallel the AMD finished in 895 secs and the Intel in 954 secs. With more threads trying to run concurrently, the Intel machine starts to slow down. To push the machines even harder we added ray trace rendering into the mix, rendering an 8K scene in KeyShot 10 using 4 cores and 8 threads. Here, the AMD’s 16 cores showed a real benefit, completing all three tasks in 1,024 secs compared to 1,459 secs on Intel. The AMD machine also has the headroom to push the rendering load up to 8 cores and 16 threads which reduced the time to 863 secs. Doing the same on Intel, and dedicating all of its 8 cores to rendering really took its toll as it caused Agisoft Metashape to crash.

Solidworks 2021 SPECapc benchmark

2.63

CPU simulate 0

1,579

235.3

42.3

14,854

Multi-core test

26.6 0

100

150

200

250

24,361

300

5000

10000

15000

20000

Workstation Specialists WS-184 (Intel Core i9-11900)

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

Solidworks 2021 application testing

1.23 Time secs (smaller is better)

nTopology 3.24

Geometry optimisation

10.48

Assembly open

98.99

IGES export 0

101.41

73.40

74.57 20

115.87

120

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 8

11,064

V-Ray 5.0 CPU Benchmark

19,839

2.7 0

Workstation Specialists WS-184 (Intel Core i9-11900)

August / September 2021

V-Ray 5.0 / Keyshot 10

KeyShot 10 Benchmark

70.77

Workstation Specialists WS-184 (Intel Core i9-11900)

WS8

Ray trace rendering

Bike seat (Voronoi pattern)

100

25000

1.23 Benchmark score (bigger is better)

1.23 Time secs (smaller is better)

9.44

Solidworks open

1,649

282.0

Rendering

1.86 1.64

Cinebench R23 (ray trace rendering)

Single core test

Export 2.20 2.05

CPU convert

Having read multiple reviews from respected tech websites like Anandtech, we were a little surprised that the Intel

1.23 Benchmark score (bigger is better)

89.5

2.39 1.53

The verdict

79.1

Model creation

4.35

1.23 Time secs (smaller is better)

CPU rebuild

SMT/HT: Simultaneous multi-threading (SMT) on AMD and HyperThreading (HT) on Intel essentially splits each of the CPU’s physical cores into two virtual

CPU tests

1.23 Benchmark score (bigger is better)

CPU raytrace

cores. It’s enabled by default and gives Both workstations were tested straight out a proven performance benefit when of the box. However, AMD and Intel CPUs rendering, but it has been known to slow can also be tuned in different ways. down other processes. Disabling SMT / HT in the BIOS made Overclocking: Both machines were virtually no difference to performance configured with stock CPU settings, where in most applications. Puget Systems has the frequency on one or more cores gets an reported a benefit of disabling SMT on automatic boost or turbo, as long as it stays AMD workstations when using Agisoft within set thermal and power limits. Metashape, but that was not what we To push performance higher, CPUs can observed in our tests. While processing also be overclocked, a process that typically time went down in one of the datasets, for involves tweaking frequencies and voltages. the other three, it actually went up. On Intel Core processors (those with the ‘K’ suffix) overclocking permanently Availability increases the frequency of all cores. Workstation brand also has a huge Different manufacturers take different influence over your choice of CPU. Pretty approaches to overclocking and some much every vendor offers a workstation push CPUs harder than others. A few with 11th Gen Intel Core CPUs. However, manufacturers hand-select CPUs that they workstations with AMD Ryzen 5000 know will overclock best. Overclocking is CPUs are only available from specialist not supported by the major manufacturers. manufacturers like Scan and Workstation With AMD Ryzen 5000, overclocking Specialists (see page WS16 for a round-up). can be controlled through Precision Boost This could change, of course. With AMD Overdrive 2, a feature which essentially offering superior performance in some makes AMD’s standard boost technology multi-threaded workflows, we are certain more aggressive. It can’t be overclocked on that Dell, HP, Fujitsu and Lenovo are all cores in the same way that Intel can, but paying close attention. Lenovo already has some manufacturers are finding new ways a workstation with an AMD Threadripper to get more out of the processor. Pro CPU, with up to 64-cores.

Autodesk Revit 2021 RFO v3 benchmark

CPU tests

Workstation tuning

100

120

5000

10000

5.07 15000

20000

25000

30000

Workstation Specialists WS-184 (Intel Core i9-11900) Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

www.develop3d.com

12/08/2021 14:38

workstation special report

workstation performed so well. In single threaded workflows it demonstrated a clear lead over the AMD workstation. It was typically around 10% faster, but in some tests it was as little as 1% or as much as 56% (when rebuilding a model in Solidworks). For CAD and BIM software, single threaded performance is where it counts. Yes, there are some processes within these applications that are multi-threaded, but very few can use all of the AMD Ryzen 9 5950X’s 16 cores. And, even if they can, it’s often only for short bursts, mixed in with single threaded operations. There are caveats to this, such as executing a solid sweep in Inventor, where AMD has a clear lead. Also, in new generation modelling tools like nTopology, which have been built from ground up for multi-core CPUs. Some of the more advanced CAM and simulation tools can also take advantage of more than eight cores, although this is not the case for Solidworks Simulate or dynamic simulation and FEA in Inventor. Even photogrammetry software Agisoft Metascan doesn’t show any major benefit from the Ryzen 9 5950X although AMD does have a small but significant lead in point cloud processing software Leica Cyclone Register 360. All of these observations are only relevant if you’re using your workstation to do one thing at a time. The Ryzen 9

InvMark for Inventor 2022 (by Cadac Group and TFI)

5950X has significantly more potential than Intel when it comes to multi-tasking, using its 16 cores to full effect to run several CPU intensive tasks concurrently. This all leads us onto rendering where the AMD Ryzen 5000 wins hands down. With the 11th Gen Core processors maxing out at 8 cores Intel simply can’t compete. And if rendering is part of your day-to-day workflow then the AMD Ryzen 9 5950X looks to be the obvious choice. Of course, the Ryzen 5000 Series and 11th Gen Intel Core isn’t just about the top-end models. For CAD users on a tight budget the 6-core Intel Core i5-11600 looks like a great value CPU. And, as

Agisoft Metashape Professional 1.73

1,272 1,855

Assemblies

Ray Tracing

1,477

1,455

2,439 2,095

1000

5,888 5,994

2000

3000

24.90

4000

5000

27.89

Sketchy view

6000

30.90 20

Workstation Specialists WS-184 (Intel Core i9-11900 with Quadro RTX 4000)

Workstation Specialists WS-184 (Intel Core i9-11900)

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X with Quadro RTX 4000)

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

1,237

Leica Cyclone Register 360

1,088

Point cloud registration

1,499

1,390

1.23 Time secs (smaller is better)

Multi-tasking workflow benchmarks (Cyclone Register 360, Metashape and Keyshot)

1.23 Time secs (smaller is better)

1,159

Saving to disk

Dynamic Sim

3,671

22.82

1,894

Data Translation

Opening files

1,522

Graphics testing (with AMD Radeon Pro W6800)

Park Map

1,686

Autodesk Revit 2021 RFO benchmark

Standard view

391 426

School model

mentioned previously, in some workflows, the 8-core AMD Ryzen 7 5800X can actually outperform the AMD Ryzen 9 5950X in multi-threaded workflows, even though it has fewer cores. In summary, choosing a workstation processor boils down to how you intend to use your machine. 11th Gen Intel Core looks to be the CPU of choice for vanilla CAD and some lightly threaded workflows. However, AMD is not that far behind. AMD is also a strong contender for those who multi-task and if ray trace rendering is also part of your day-to-day workflow then the AMD Ryzen 9 5950X wins hands down.

1.23 Time secs (smaller is better)

111 120

Rock model

School Map Drawing

Puget Systems benchmark 1.23 Time secs (smaller is better)

1.23 Benchmark score (bigger is better)

Modelling

Your choice of CPU has very little bearing on 3D performance in GPU limited applications like Unreal Engine

963

1,992

1,313

FEA

665

1,777

1,263

Large dataset (99 GB)

Cyclone and Metashape

699

Small dataset (24 GB)

1,888

1,785

Cyclone, Metashape and Keyshot on 4 cores Cyclone, Metashape and Keyshot on 8 cores

500

1000

1500

2000

895

1,439

1,024

Test did not finish 0

954

863 300

600

900

1200

1500

Workstation Specialists WS-184 (Intel Core i9-11900)

Workstation Specialists WS-184 (Intel Core i9-11900 with Quadro RTX 4000)

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X)

Scan 3XS GWP-ME A132R (AMD Ryzen 9 5950X with Quadro RTX 4000)

0 500 1000 1500 2000 2500 3000 3500 4000

www.develop3d.com

WS04_05_06_09_10_11_D3D_AUGSEPT21_Intel_AMD.indd 9

August / September 2021

WS9

12/08/2021 14:38

workstation special report

Multicore for CAD talking heads

Kevin Schneider, Senior director, Fusion 360 Group technology strategist / Autodesk With eight CPU cores now standard in workstations, rising in some cases to as many as 64, designers and engineers typically have a huge amount of processing power at their fingertips. So is there scope for making CAD software more multithreaded, enabling it to use more cores, more efficiently – or does the sequential nature of many operations mean the hands of CAD software developers are tied? We asked the experts to find out more Autodesk Inventor (below) nTopology (bottom)

AD applications have limits on how aggressively they can use multicore computing. For example, history-based modelling has operations that must be performed serially, and therefore cannot be parallelised across cores. Fortunately, as CAD applications mature and integrate other valuable capabilities, many of those added capabilities leverage more cores. Autodesk’s design and manufacturing applications leverage multicore compute in specific cases. These include importing solid geometries; performing mass property and sheet metal calculations; toolpath and nesting calculations; and simulations such as e-cooling, injection moulding, and meshing and model preparation. Graphics operations, like converting BREP (boundary representation) surface data to facet information for real-time rendering, are also multicore capable.

In any discussion of multicore capabilities, the importance of GPU and cloud compute cannot be overlooked. Autodesk provides multiple ray trace rendering tools, such as Autodesk VRED, which use GPU compute. In fact, rendering and ray tracing leverage as many cores as are available, and scale almost linearly with the number of cores. Autodesk also leverages ondemand cloud compute, scaling CPU/GPU resources in the cloud to deliver compute-intensive output such as simulation, rendering and generative design. Generative design can scale to using many cores and GPUs in parallel to solve its challenges and produce multiple outcomes. If you only look at history-based part modelling operations, the value of multiple CPUs is low. But CAD applications have evolved way beyond just part modelling, and customers receive significant benefits from the CAD capabilities that do use multicore computing. If you are designing assemblies, working with imported data, rendering, creating drawings and frequently calculating manufacturing toolpaths, those cores will increase your productivity. Add the ability to use cloud compute on demand and customers have even more compute available to help them push their design and manufacturing processes further and faster. ■ autodesk.com

‘‘

If you are designing assemblies, working with imported data, rendering, creating drawings and frequently calculating manufacturing toolpaths, those cores will increase your productivity

’’

WS10

August / September 2021

WS10_11_12_D3D_AUGSEPT21_Talking_Heads_CAD.indd 10

www.develop3d.com

18/08/2021 14:09

workstation special report

Danny Sicking Director of Software Engineering, NX Platform / Siemens

n developing new capabilities and expanding our products, we are always looking for ways to leverage compute power to deliver the best performance. Today’s frontier of parallel compute exists on the cloud. As microservices and horizontal scaling emerge for CAD and engineering workloads on the cloud, we can learn from recent

models in real time, based on insession dynamic content. Additionally, model and assembly load times can be reduced by spreading the load of reading data across cores. Analysing the structure from the PLM system and prepping for the load can also be parallelised. As a general guideline, compute-intensive operations with minimal UI feedback or end-user interaction are usually the best places to multithread and improve performance. UI operations must be processed on the main thread, so these cannot be executed on multiple cores. Another barrier to multithreading is when results are order-dependent, like parametric feature updates. Performance improvements can be achieved by other means, such as segmentation and reducing the

‘‘

Display-related functions such as bounding box calculations, hidden line removal, view section creation, edge extraction and rendering are easily sped up by leveraging multiple cores

’’

advancements on desktop that leverage multicore CPUs. Display-related functions such as bounding box calculations, hidden line removal, view section creation, edge extraction and rendering are easily sped up by leveraging multiple cores, mainly because the results are not orderdependent. High-end rendering also can be greatly speeded up by using GPU acceleration. Our recent advancements in AI/ML leverage multithreading to build

number of features participating in the update. At Siemens, we look to leverage advances in hardware and computing technologies in products such as NX, as well as our kernel modelling tools such as Parasolid. We do this with a focus on the value it can offer to our customers and our commitment to preserving their data and maintaining a high level of quality. ■ sw.siemens.com

www.develop3d.com

WS10_11_12_D3D_AUGSEPT21_Talking_Heads_CAD.indd 11

Chris Hall Director of software performance engineering / AMD

omputer hardware has become so much faster since I started in CAD in the early 1990s, that for many designers working on smaller parts, the single-threaded nature of CAD applications is not an obstacle to productivity in terms of designing that widget. But I think the world has moved on from simply designing an object by hand and then turning around and having it made. We’ve moved into a world now where the compute power that’s available from high core count processors allows for a more robust and rigorous design process, in a couple of ways in particular. First, with additive manufacturing, we’ve entered an era where generative design becomes important, and this is an area where having more cores is definitely a performance advantage. Second, we’re now capable of doing a lot more simulation on everything that we make. It’s no longer the realm of a

supercomputer or a cluster. With a 64-core Threadripper Pro, you can easily do simulation on your desktop. In fact, because you have so many cores available, you can set that simulation running and while it might take several hours to complete, you still have enough compute available to continue doing design work with your CAD application of choice. So, while you don’t necessarily need the additional cores for traditional design tasks, they provide a boost in those areas where CAD has evolved. Often an afterthought, the fact is that once you design a part, you still have to manufacture it. With injection moulding, you need to run a simulation of plastic flow through the part, and that simulation is multithreaded. Another element that people don’t often consider is generating CNC toolpaths – software like CAMworks, for example, is highly multithreaded. In practice, that means with Threadripper Pro, you have the opportunity to iterate more often to refine that toolpath, which can a) increase production-line throughput, and b) decrease wear on cutters and save material, both of which increase company profitability. So, it’s not the old-school way to think about CAD, but in terms of contributing to the bottom line of the company you’re working for, that’s definitely a place where high core count helps. ■ amd.com

‘‘

It’s not the old-school way to think about CAD, but in terms of contributing to the bottom line of the company you’re working for, that’s definitely a place where high core count helps

’’

August / September 2021

WS11

18/08/2021 14:09

workstation special report

Blake Courter Chief Technical Officer nTopology

Topology’s unique computational approach allows it to take full advantage of the available CPU and GPU resources of a desktop or virtual machine. We have pioneered several technologies that work together to enable this achievement. Our field-driven design and implicit modelling calculations are highly parallelisable, unlike B-rep and mesh-based

will simultaneously use all the CPU and GPU cores provided. If your hardware has any issues with cooling or the power supply, it will likely expose them. To avoid over-consuming resources, we match our worker threads to the available cores. An added benefit is that, unlike most engineering software, nTopology lets you keep working in your document while evaluation runs in the background. Our generative design tools also leverage precise B-rep, mesh, voxel and FE modelling components, solvers and algorithms. Some of these blocks inherit compute limitations from those components. Although such bottlenecks do not yet appear on the critical path for our customers, our evaluation system is stateless and purely functional, so we have the opportunity to parallelise those when possible. nTopology is still a young

‘‘

We can use double precision on the CPU for engineering computations, while simultaneously using single precision on the GPU for visualisation and real-time interaction

’’

modelling systems, which rely on sequential algorithms. The straightforward math expressions that represent our models are theoretically infinitely precise and are computed at the appropriate fidelity on the fly. For example, we can use double precision on the CPU for engineering computations, while simultaneously using single precision on the GPU for visualisation and real-time interaction. The result is that nTopology WS12

product with a natural affinity for more compute resources. Our recent 3.0 release, which demonstrated a 10-times to 100-times improvement using the GPU, exemplifies such compute flexibility. We are now experimenting with porting our implicit expressions to arbitrary platforms, from Raspberry Pi to HPC to CAD add-ins. Although we’ve been focused on individual PCs so far, distributed approaches are likely in our future as well. ■ ntopology.com

August / September 2021

WS10_11_12_D3D_AUGSEPT21_Talking_Heads_CAD.indd 12

Brett Chouinard Chief Technical Officer Altair

ecent developments in hardware technologies have indeed put multicores and threads via CPUs and GPUs at the fingertips of developers in all industries. Some of these industries were architecturally designed to benefit from multicores naturally, like the gaming industry, where applications like system AI [Artificial Intelligence], sound handling, network layers and graphic rendering could be handled through multiple cores and threads. Other industries, such as CAD or CAE, had to follow, and needed to consider the benefits of this architectural redesign of the applications over their original infrastructure, or at least identify areas within the applications’ workflows that would greatly benefit from these types of changes and updates. Altair uses multicores and threading in many aspects of its simulation suite; for example,

within the meshing and morphing capabilities of Altair HyperWorks, when reading or importing large data files and results and for graphic renderings. A good example is Altair’s graphical capabilities, which have been added to our Inspire Studio Renderer solution and now include CPU and GPU multicore support with the addition of hardware ray tracing and denoising support from the GPU. The results must be seen to be believed; they are faster than ever before and look great in real time. The more obvious areas in Altair products that show tremendous multithreading benefits are our solver solutions. One such product is Altair NanoFluidX, a solution from our CFD solver package, utilising multicores over multinodes (including cloud solutions) and Nvidia Cuda GPU hardware. Some of our test results show speed enhancements of up to 20 times, a boost that allows designers, analysts and engineers to use proper tools to turn their ideas into the best designs and products possible. Altair is a company that brings innovation to its products while utilising cutting-edge technologies, be it multithreading CPUs, compute and AI GPU power or HPC cloud-based solutions to allow people to turn their ideas and design into reality. ■ altair.com

‘‘

The more obvious areas in Altair products that show tremendous multithreading benefits are our solver solutions

’’

www.develop3d.com

18/08/2021 14:09

DISCOVER YOUR SOFTWARE’S FULL POTENTIAL Welcome to Our Greatest PRO GPU. Ever.

Welcome to Exceptional Performance. Introducing the Incredible Radeon™ PRO W6800 GPU. Delivering a Gigantic 32 GB of High-Performance Memory with ECC Support, Hardware Raytracing, Optimizations for 6 Ultra-HD HDR Displays, Accelerated Software Multi-Tasking, PCIe® 4.0 for Advanced Data Transfer Speeds, and Certifications for Many Popular Applications. ALL AS STANDARD.

Enhanced Compute Units To Decrease Latency

Ray Accelerators For Realtime Raytracing

26.8 Billion Transistors Within 520mm2

Importance to Professional Software Engineered from the ground up with superior performance and power efficiency, AMD RDNA™ 2 architecture introduces significant advancements in the form of an all new AMD Infinity Cache™ Memory, an enhanced Compute Unit, Hardware Raytracing support, combined with a new visual pipeline for added efficiencies. These advancements help enable high resolution performance and increased vivid visuals within your professional software of choice. The latest AMD Radeon PRO W6000 graphics series is compatible with modern workstations offering additional optimizations for select AMD Ryzen™ based processors, harnessing the GPU’s and system’s full performance potential.

PCIe® 4.0 x16 Support Removing Bottlenecks

AMD Infinity Cache Bandwidth Amplifier

AMD RDNA 2 graphics processor within the high performing AMD Radeon PRO W6800.

Professional Graphics for Exceptional Performance with Reliability, Stability and Software Certifications at its Core. ©2021 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo, AMD Infinity Cache, Radeon, AMD RDNA, Ryzen and combinations thereof are trademarks of Advanced Micro Devices, Inc. PCI Express and PCIe are registered trademarks of PCI-SIG Corporation. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies. PID#: 21822206

workstation special report

Scan 3XS GWP-ME A132R 11th Gen Intel Core workstations might have have the edge in CAD and BIMcentric workflows, but this Ryzen 9 5000 beast from Scan wins hands down when it comes to rendering and extreme multi-tasking, writes Greg Corke

ike many workstation manufacturers, it’s been a frustrating six months for Bolton-based Scan. AMD Ryzen 5000 CPUs have been in huge demand, but the global chip shortage has meant limited supply. This now appears to have eased and Scan workstations with ‘Zen 3’ AMD CPUs are now rolling off the company’s production line. The delay has brought some benefits. Scan can now pair AMD’s impressive desktop CPU with a choice of three Nvidia ‘Ampere’ pro GPUs. The Nvidia RTX A4000 (16 GB), Nvidia RTX A5000 (24 GB) and Nvidia RTX A6000 (48 GB) are all available as options inside the new Scan 3XS GWP-ME A132R. For our review machine, Scan has chosen the top-end AMD Ryzen 9 5950X CPU. With 16 cores, 32 threads, a base frequency of 3.4 GHz and a boost of 4.9 GHz, this processor is adept at handling all different types of workflows. You get excellent performance in single threaded CAD software, as well as multithreaded power for simulation, point cloud processing, photogrammetry, generative design, CAM and, of course rendering.

WS14

August / September 2021

WS14_D3D_AUGSEPT21_Scan.indd 14

As you’ll learn from our indemanding graphics-centric Product spec depth article on page WS4, the workflows (or a tighter budget) 11th Gen Intel Core CPU still a downgrade to the Nvidia ■ AMD Ryzen 9 5950X processor has the edge in many workflows. RTX A4000 will shave close (3.4 GHz, 4.9 GHz However, when it comes to to £1,000 off the £4,167 (Ex boost) (16 cores, 32 threads) rendering and extreme multiVAT) price tag. See our in-depth ■ Nvidia RTX A5000 tasking, the 8-core Intel Core review of both Nvidia GPUs on GPU (24 GB) i9-11900 or Intel Core i9-11900K page WS28. ■ 64 GB (2 x 32 GB) simply can’t compete with the The Asus Pro WS X570Corsair Vengeance DDR4 3,200 MHz 16-core AMD Ryzen 9 5950X. ACE motherboard is ■ 2 TB Samsung 980 To explore the machine’s workstation-grade and has Pro NVMe PCIe 4.0 SSD + 4 TB Samsung multi-tasking potential, we what Scan describes as an 870 Evo SATA SSD pushed it to its limits, with ‘over-engineered cooling and ■ 3XS workstation a demanding engineering power delivery system’. It has case with tempered glass window workflow, consisting of point four memory slots, populated ■ Microsoft Windows cloud processing in Leica in our test machine with 64 GB 10 Professional Cycone Register 360 and (2 x 32 GB) of Corsair Vengeance 64-bit photogrammetry processing DDR4 3,200 MHz memory. For ■ 3 Years warranty – 1st Year Onsite, 2nd in Agisoft Metashape. Not only memory-hungry workflows and 3rd Year RTB did it reduce processing time like point cloud processing, it’s (Parts and Labour) to 889 secs from 1,306 secs (the easy to upgrade to 128 GB, as ■ £4,167 (Ex VAT) time it took if both jobs were run we did when testing with Leica scan.co.uk/3xs sequentially), but the machine Cyclone Register 360, though was still responsive enough to you’ll first need to clip off one of model comfortably in Revit. See our in- the CPU fans to get to the slots. depth AMD vs Intel article on page WS4 to With a second PCIe Gen 4 slot, you can find out more. add a second Nvidia RTX A5000 GPU at a The 2 TB Samsung 980 Pro NVMe PCIe later date although you may need a bigger 4.0 Solid State Drive (SSD) certainly helps PSU than the 750W Corsair RMX, which here as it’s able to read and write data very is 80PLUS Gold rated. Scan technicians quickly, so storage doesn’t become too can advise here. much of a bottleneck. Everything is housed in a new custom Rather than supplementing the M.2 SSD Scan 3XS workstation case, which is with a 3.5-inch SATA Hard Disk Drive available with both solid and tempered (HDD) for secondary storage, Scan has glass side panels. It’s a nice sturdy chassis included a 4 TB Samsung 870 Evo SATA with a 3XS branded vented front that SSD instead. Costing less than 10p per GB, helps ensure cool air runs freely from SSDs are quickly becoming the future of front to back. all workstation storage, and not just for A substantial dual radiator, dual fan OS, apps and select datasets. Noctua NH-D15 air cooler helps keep the Coupled with the Nvidia RTX processor running cool. We observed A5000 GPU, the workstation a peak frequency of 4.91 GHz in single has an obvious bias towards threaded workflows, going down to 3.74 graphics intensive workflows, GHz with all 16 cores running flat out for suggested by the ‘ME’ (Media extended periods. Scan says the Noctua and Entertainment) abbreviation is generally quieter than an equivalently in the product name. rated hydrocooler, and it’s hard to The Nvidia RTX A5000 is a argue with that. The entire system was hugely powerful professional incredibly quiet, even under heavy loads. GPU for real-time viz and GPU rendering. At 4K resolution, The verdict we got in excess of 20 frames This is another excellent machine from per second in most of our test specialist workstation manufacturer applications. The exceptions Scan. With the AMD Ryzen 9 5950X were Unreal Engine 4 when real- CPU, the 3XS GWP-ME A132R is equally time ray tracing was enabled and adept at handling single threaded or in Autodesk VRED Professional multi-threaded workflows. Whether when anti-aliasing was set to that’s for rendering stills and animations, ultra-high. However, these or for demanding multi-tasking reality are extreme workloads. Most modelling workflows, it doesn’t really mainstream viz users should matter. And with an Nvidia RTX A5000 expect a smooth, reactive GPU backing it all up with substantial viewport when navigating graphics horsepower, it can handle pretty models, and superfast results much anything you throw at it. And it when GPU rendering. does all this while giving out little more For those with less than a gentle hum. www.develop3d.com

12/08/2021 14:43

workstation special report

Workstation Specialists WS-184 The 11th Gen Intel Core processor might be a step back in some departments, but in single threaded workflows like CAD it remains king. Furthermore, coupled wth the right components — as it is in this new tower from Derby-based Workstation Specialists — it can also provide a solid foundation for more demanding workflows. By Greg Corke

hen Intel launched its ‘Rocket Lake’ processor family earlier this year it came under a lot of flak for having fewer CPU cores than the previous generation. 10th Gen Intel Core maxed out at ten, but the new 11th Gen Core CPUs were limited to eight. The reality is, this is only an issue if you use software that can take full advantage of that many cores. And for most architects, engineers and product designers, that’s typically only ray trace rendering. For those that rely solely on CAD software, it continues to be all about frequency and Instructions Per Clock (IPC). And here, Intel delivers in spades. The Intel Core i9-11900 CPU at the heart of this Workstation Specialists desktop tower might have a base clock of 2.50 GHz, but in many of our single threaded CAD tests it hit 5.15 GHz on a single core, just a touch slower than the stated maximum of 5.20 GHz. And this isn’t even Intel’s fastest 11th Intel Core CPU. If you can get hold of one, the Intel Core i9-11900K can go all the way up to 5.3 GHz. Despite ‘only’ having eight cores, the Core i9-11900 performed very well in some multi-threaded workflows like point cloud processing and reality modelling. Applications like Leica Cyclone Register 360 and Agisoft MetaShape might be multi-threaded but there appears to be little additional benefit (or none at all) to having more than eight cores. It’s only when you get into highly-threaded applications like rendering and geometry optimisation that CPUs with more cores, such as the AMD Ryzen 9 5950X, have an advantage. You can learn more about this in our AMD vs Intel article on page WS4. Our test machine came with 64 GB (2 x 32 GB) of 3,200 MHz dual channel Corsair Vengance RGB Pro DDR4 memory. This is a good amount for an allround CAD-centric workstation. Those who simply work with average sized models in Solidworks or other CAD tools should get away with 32 GB. www.develop3d.com

WS15_D3D_AUGSEPT21_WS.indd 15

Conversely, those who deal with new Nvidia RTX A4000 which Product spec huge point cloud datasets, would comes with a substantial 16 GB likely be better served with 128 of memory. As you see from our ■ Intel Core i9-11900 processor (2.5 GHz, GB — the maximum the Asus review on page WS28, this is an 5.3 GHz Turbo) PRIME Z590-P motherboard impressive all-round GPU for (8 cores, 16 threads) can take in its four memory slots. workflows that include GPU ■ 64 GB (2 x 32 GB) 3,200 MHz dual The motherboard fits snuggly rendering, real-time 3D, realchannel Corsair inside the Fractal Design Define time ray tracing and VR. Vengance RGB Pro DDR4 memory 7 Compact chassis. At 427 x For an additional £767 you ■ 2 TB Samsung 210 x 474 mm this is a shrunkcan upgrade to the Nvidia RTX 980 Pro NVMe PCIe 4.0 SSD down version of the popular A5000 (24 GB). But if you’re ■Asus Prime Z590-P workstation case used by many very much focused on CAD motherboard of the UK’s custom workstation and other less demanding 3D ■ Fractal Design manufacturers. applications, you could easily Define 7 Compact desktop chassis For easy access, there are two drop down to the Nvidia T1000 (427 x 210 x 474 mm) USB 3.2 Gen 1 ports up front, but (4 GB), or the Nvidia RTX ■ Microsoft Windows no USB 3.2 Gen 2x2 Type-C. If A2000 (6 GB), which will be 10 Professional 64-bit you have devices that need the available in October. ■ 36 Months more modern USB standard, premium RTB The verdict you’ll probably want to buy a hardware warranty with remote hub to plug into the single USB The Workstation Specialists engineer diagnostics by next business day Type C port at the rear. WS-184 is an excellent choice ■ £2,414 (Ex VAT) The machine has three fans for CAD-centric workflows. In that move air from front to applications like Solidworks ■ Upgrades back and, together with the ‘Be it delivers where it counts, 128 GB memory Quiet’ CPU air cooler, keep the but also has enough cores - add £315 machine running pretty quietly. to support more processorNvidia RTX A5000 - add £767 For a machine of this type, it intensive workflows like 8 TB Samsung 870 offers the perfect combination point cloud processing and QVO - add £696 of size and expandability. In photogrammetry. the bottom section of the case, For those with a focus on workstation specialists.com nestled alongside the 750W rendering, the fact that the ATX 80-Plus Platinum Certified Workstation Specialists WSPower Supply Unit (PSU), there are two 184 would struggle against an AMD 3.5-inch drive bays. Ryzen 9 5950X-based workstation could For most CAD-centric workstations, we’d be a moot point if your software is tuned usually expect one of these bays to be fitted for GPU. The Nvidia RTX A4000 GPU with a Hard Disk Drive (HDD) to give a cost- (or RTX A5000) is more than capable of effective mass storage partner to the on- doing the heavy lifting here. board NVMe SSD. For example, in a budget Conversely, strip out the high-end GPU, workstation, a combination of 512 GB Solid drop down to 32 GB of RAM, and pare State Drive (SSD) and 2 TB HDD is typical. back on the storage and you should have However, for this machine, Workstation a very affordable, but fast workstation for Specialists has gone for SSDs throughout. bread and butter CAD work. For the main system drive there’s the high-performance 2 TB Samsung 980 PRO M.2 PCIe 4.0 NVMe SSD, which offers up to 7,000 MBps read and 5,100MBps write. But this has been partnered with an 8 TB Samsung 870 QVO 2.5-inch SATA III SSD. With 560MBps read, and 530MBps write, the 870 QVO offers significantly lower sequential read/write performance than its PCIe 4.0 counterpart, but with a whopping 8 TB to play with, it’s a great alternative to an HDD in I/O intensive workflows like point cloud processing, where collosal datasets are frequently read from / written to disk. For graphics, there’s the August / September 2021

WS15 15

12/08/2021 14:44

workstation special report

CAD workstation round up The latest workstations for CAD-centric workflows: 11th Gen Intel Core (up to 8 cores) and AMD Ryzen 5000 (up to 16 cores) to go beyond 3D design and into the realms of rendering, reality modelling and simulation 1

BOXX Apexx S3

Dell Precision 3450 SFF

Fujitsu Celsius W5011

BOXX has built a major part of its business around overclocking and with the BOXX Apexx S3 it permanently boosts 11th Intel Core clock speeds across all eight cores. The Intel Core i7-11700K runs at 5.0 GHz, while the Intel Core i7-11900K runs at 5.3 GHz. As with all BOXX workstations, there’s also a huge focus on build quality, with the custom chassis made from ‘aircraft-grade’ aluminium, offering a strength and rigidity way beyond that of most off-the-shelf cases.

The Dell Precision 3450 SFF is a very compact desktop workstation. Measuring a mere 290 x 93 x 293mm it can even be mounted behind a display (pictured left). The ‘small form factor’ chassis does mean a more limited set of processor options, maxing out at the 65W Intel Core i9-11900 or 80W Intel Xeon W-1390. It’s also restricted to entry-level pro GPUs, but the AMD Radeon Pro WX 3200 and Nvidia Quadro P1000 are perfectly suited to 3D CAD workflows.

This 21 litre ‘micro tower’ is not as deep as comparable workstations, as the motherboard and GPU span its entire depth. It offers a choice of 11th Gen Intel Core or Intel Xeon W-1300 processors and can be configured with a massive range of GPUs, from the entry-level CAD-centric Nvidia T400 up to the Nvidia Quadro RTX 5000 (it doesn’t yet offer the new Nvidia RTX A4000 / A5000). Other features include up to 128 GB of memory, multiple drives and tool-less access.

■ boxx.com ■ boxx-tech.co.uk

■ dell.com/precision

■ fujitsu.com

Lenovo ThinkStation P350 Tiny

Boston VENOM R41-10NP

Broadberry CyberStation SFF

Tiny by name, tiny by nature — this is the smallest workstation on the planet, measuring a mere 37 x 183 x 179mm. However, it still has everything you need for mainstream CAD workflows, including an 11th Gen Intel Core i9 CPU (8 cores, 5.2 GHz), up to 64 GB of DDR4 3200 MHz memory and a choice of Nvidia P1000 or T600 GPUs. There’s no room for a HDD but with an M.2 PCIe Gen4 NVMe SSD up to 2 TB there’s still plenty of storage. Plus, built in WiFi.

Boston offers a huge variety of desktop workstations in its Venom range, from Intel Core and Intel Xeon, to AMD Threadripper, Threadripper Pro and AMD Epyc. This AMD Ryzen 5000 machine can be fitted with optional liquid cooling for ‘maximum performance and whisper quiet operation’ and matched with up to 128 GB of DDR4 3,200 MHz stock or 4,733 MHz overclocked memory. All of Boston’s workstations are available to lease.

With its 250 x 203 x 367 mm Fractal Design Core 500 chassis, the Broadberry CyberStation SFF is one of the smallest AMD Ryzen 5000 workstations. This, together with the built-in WiFi, make it well suited to home workers. Despite its size it can still take a whole host of pro GPUs, up to the Nvidia RTX A6000. Two RAM slots on the Gigabyte AMD Ryzen X570 I AORUS PRO motherboard mean it’s limited to 64 GB, but that’s still plenty for most CAD-centric workflows.

■ lenovo.com/workstations

■ boston.co.uk

■ broadberry.co.uk/amd-ryzen-workstations

Interpro IPW-R9 Dell Precision 5760

Novatech ProStation WR7-WX41

Overclockers RENDA

As the name suggests, the InterPro IPW-R9 features a choice of 3rd Gen AMD Ryzen 9 CPUs, including the 12-core Ryzen 9 5900X and 16-core Ryzen 9 5950X. To keep clock speeds running as high as possible for longer periods, the UK firm uses a range of Corsair all-inone liquid CPU coolers. Different BIOS profiles can be created for customers, matched to their workflows. For example, to allow higher clock speeds by temporarily sacrificing cores.

The Novatech ProStation WR7-WX41 is built around the AMD Ryzen 5000 Series with a choice of three CPUs — the 8-core 5800X, 12-core 5900X and 16-core 5950X. It’s fully customisable, but not just core components like memory, graphics and storage. Customers can also choose from 15 different CPU coolers (air or liquid) and nine different chassis, from full towers like the Phanteks Enthoo Pro (pictured) to 4U rack mounts like the Chenbro RM41300G.

It’s not hard to guess what this UK firm specialises in. The Overclockers RENDA workstation is all about pushing the limits of performance, while maintaining stability. Professional overclocker Ian Parry (aka 8Pack) heads up the R&D, delivering hand-built machines based on each customer’s workflow requirements. With a custom water cooling solution he says he can push the AMD Ryzen 9 5950X to 5.1 GHz on one core and 4.6 GHz on all cores.

■ ipworkstations.com

■ novatech.co.uk

■ overclockers.co.uk

WS16

August / September 2021

WS16_17_D3D_AUGSEPT21_RoundUp.indd 16

www.develop3d.com

12/08/2021 14:46

workstation special report

13 12

BIMBOX Stryker III

HP Z2 G8 SFF

BOXX Apexx Denali

Armari Magnetar

As its name suggests, BIMBOX is laser focused on the AEC (Architecture, Emgineering and Construction) sector, but can apply its extensive knowledge of Building Information Modelling and viz software to mechanical CAD workflows. The firm takes overclocking extremely seriously. For its Stryker III workstation it ‘delids’ the Intel Core i9-11900K CPU, taking off the standard heat spreader and mounting its own liquid cooler directly onto the silicon. This brings down the CPU temperature considerably so it can safely run at 5.3 GHz on all cores. BIMBOX is based in the US, but its machines will soon be built, sold and supported in the UK and other countries.

With its 338 x 308 x 100mm chassis, the HP Z2 G8 SFF is slightly bigger than the Dell Precision 3450 SFF (top left) but has the option of more powerful processors. These include the 125W Intel Core i9-11900K and the Nvidia Quadro RTX 3000, which extends the reach of the workstation beyond 3D CAD and into the realms of entry-level viz. According to HP, the Z2 G8 SFF offers ‘Unthrottled performance’ thanks to Z’s ‘industry-leading’ thermals that keep the processor and graphics card cool, so they can run at max performance for extended periods of time. ‘Unthrottled performance’ also extends to the new PCIe Gen 4 Samsung PM9A1 SSD.

With its Apexx Denali, BOXX was one of the first workstation manufacturers to offer a Ryzen 5000 Series workstation. It uses the same compact custom ‘aircraft-grade’ aluminium chassis as the Intel-based Apexx S3 and at 174 x 388 x 452mm it’s smaller than most AMD Ryzen 5000 tower workstations. But that doesn’t come at the expense of expandability. The Apexx Denali can house up to two high-end Nvidia RTX, Nvidia GeForce or AMD Radeon Pro GPUs and two 3.5-inch Hard Disk Drives (HDDs). To keep the processor running at peak frequencies it uses a liquid-cooled closed loop system with a sizeable radiator.

The Armari Magnetar V16R-RA850G2-2S is one of the smallest AMD Ryzen 5000 Series workstation out there, measuring a mere 360 x 87 x 400mm. The custom chassis features a high quality Japanese steel frame which was designed in-house by the specialist UK manufacturer. Unlike many other small form factor workstations, there is no compromise on graphics and the workstation can support one dual slot GPU like the AMD Radeon Pro W6800 or two single slot GPUs like the Nvidia RTX A4000. Custom fans and a 14cm all-in-one liquid CPU cooler help maintain peak performance, while still preserving quiet operation.

■ bimboxusa.com

■ hp.com/z

■ boxx.com ■ boxx-tech.co.uk

■ armari.com

www.develop3d.com

WS16_17_D3D_AUGSEPT21_RoundUp.indd 17

August / September 2021

WS17

12/08/2021 14:46

workstation special report

Best lightweight workstation laptops 2021 Our top picks for ultra-portable mobile workstations to take CAD and design visualisation on the road — all under 20mm and most below 2kg

HP ZBook Studio G8

The 15.6-inch HP ZBook Studio G8 is HP’s first mobile workstation to offer both pro and consumer graphics options in the same machine, up to the Nvidia RTX A5000 (16 GB) or GeForce 3080 (16 GB). Both GPUs are ideal for design viz, VR and GPU rendering but as the laptop is very slim (17.5mm) we expect the same GPU might run faster in the thicker (22.8mm) HP ZBook Fury G8 15, which should offer better cooling and increased power draw. The HP ZBook Studio G8 offers a choice of 11th Generation Intel Core H-Series processors up to the Intel Core i9-11950H but only up to 32 GB RAM, which might be a little light for some workflows. It also features an optional HP DreamColor display with a 120Hz refresh rate, a billion on-screen colours, 100% DCI-P3, and ‘end-to-end’ colour accuracy with Pantone validation. It starts at 1.79kg. ■ hp.com/z

Dell Precision 5760 The Dell Precision 5760 is somewhat unique as it remains the only thin and light 17-inch mobile workstation from a major vendor. It is a replacement for the Dell Precision 5750 but features an enhanced thermal design including dual output fans, vapour chamber and a hidden exhaust venting through the hinge. Like the 15-inch Dell Precision 5560 (see top right) it features a combination of aluminium and carbon fibre for the chassis and a 94% display to body ratio thanks to the 4-sided InfinityEdge, 16:10 aspect ratio display. The thin (8.67 mm - 13.15 mm) and light (2.15kg) design means some compromise on graphics with the Nvidia RTX A2000 (4 GB) and Nvidia RTX A3000 (6 GB) being the only options, although the latter is ‘VR Ready’. However, it offers the same broad choice of 45W 11th Gen Intel Core and Xeon CPUs and supports up to 4 TB of PCI Gen4 SSDs and 64 GB of DDR4, 3,200 MHz memory. ■ dell.com/precision

WS18

August / September 2021

WS18_19_D3D_AUGSEPT21_Laptops.indd 18

www.develop3d.com

12/08/2021 14:47

workstation special report

MSI Creator Z16 Dell Precision 5560

MSI’s new pro-focused laptop marks a change in aesthetics for the Taiwanese company. The slimline 16mm chassis is made from CNC-milled aluminium with a ‘Lunar Gray’ finish. It starts at 2.2kg. With a 16-inch 16:10 aspect ratio display you get a bit more viewing space than the traditional 16:9. ‘True Pixel technology’ means extremely accurate colours and the display is hardware calibrated in the factory to give 100% coverage of the DCI-P3 colour gamut. QHD+ (2,560 x 1,600) resolution means pixel density is lower than a typical 4K (3,840 x 2,160) laptop display. The Z16 features a choice of 11th Gen Intel Core H series processors, an Nvidia GeForce RTX 3060 laptop GPU with Nvidia Studio drivers, up to 64 GB memory and up to 4 TB of storage spread across two M.2 NVMe PCIe Gen4 SSDs.

The Dell Precision 5560 wins hands down when it comes to portability. It’s the thinnest and lightest out of all the 15.6-inch mobile workstations — a mere 7.7mm at the front, 11.64mm at the rear and starting at 1.84kg. And with ultra thin bezels, it’s also notably smaller than comparable machines. In order to achieve this sleek aesthetic, it only includes entrylevel graphics options including the Nvidia T1200 (4 GB) and Nvidia RTX A2000 (4 GB), which are best suited to 3D CAD and entry-level viz workflows. However, there’s no compromise on the CPU with options going up to the Intel Xeon W-11955M (8 Core, 2.60 GHz up to 5.00 GHz). The laptop supports up to 64 GB of DDR4 3,200 MHz memory and 4 TB of NVMe PCI 4.0 storage. The IPS 4K ‘Gorilla Glass’ display is also top notch — 500 nits, 100% AdobeRGB and 99% DCI-P3. ■ dell.com/precision

■ msi.com

Lenovo ThinkPad P1 Gen 4 The first three generations of this thin and light mobile workstation featured a 15.6-inch display and CAD-focused pro graphics. The G4 edition is a ‘clean sheet’ design with a 16-inch display and higher-powered GPUs, including the ‘professional’ Nvidia RTX A5000 (16 GB) and ‘consumer’ Nvidia GeForce 3080 (16 GB). Memory and storage capacity remains the same with up to 64 GB DDR4 3,200 MHz and up to two 2 TB M.2 NVMe PCIe Gen4 SSDs. To accommodate the higher-end GPUs, which draw significantly more power than those in previous generation ThinkPad P1s, Lenovo has developed a new thermal design. It is also using AI to dynamically manage the ‘cooling budget’. For example, if a workflow is dependent on both the CPU and the GPU, it might set the Total Graphics Power (TGP) to 80W, whereas if a workflow is totally reliant on the GPU, such as GPU rendering, it could go as high as 90W or 100W. Despite more powerful GPUs and a larger display, the ThinkPad P1 Gen 4’s carbon fibre and magnesium alloy chassis has only increased slightly in size and weight – 361.8 (w) x 245.7 (d) x 18.4mm (h) and starting at 1.81kg. ■ lenovo.com/workstations

Microsoft Surface Book 3 (15-inch) The Microsoft Surface Book 3 launched in 2020 so is the oldest machine in this round up, but it warrants inclusion because it offers something different. At the push of a button, you can remove the touchscreen display and turn it into a tablet. And with the optional pressure sensitive Surface Pen, use it for precision sketching. The 15-inch display has a resolution of 3,240 x 2,160 and an aspect ratio of 3:2, which is deeper than all the other machines. As it’s last year’s model, the Surface Book 3 features a 10th Gen Intel Core processor - the Core i7-1065G7. With a boost of 3.9 GHz, performance in CAD will be OK, but with four cores and a base clock of 1.5 GHz it will be significantly slower than others in multi-threaded workflows like rendering. For graphics, you have the option of the Nvidia Quadro RTX 3000 (6 GB), which is designed for entry-level viz, but its ‘Max-Q Design’ means it will run slower than other machines with the same GPU. The Microsoft Surface Book 3 has a thickness of between 15mm and 23mm and weighs 1.9kg with the keyboard and 0.81kg without. ■ surface.com

www.develop3d.com

WS18_19_D3D_AUGSEPT21_Laptops.indd 19

August / September 2021

WS19

12/08/2021 14:47

workstation special report

AMD Radeon Pro W6800 This beast of a card is the first pro GPU from AMD with hardware-based ray tracing built in. With a whopping 32 GB of on board memory it’s designed for the most demanding arch viz workflows, writes Greg Corke Price $2,250 amd.com/radeonpro

memory to protect against crashes. And instead of three axial fans that recirculate air inside the machine, it has a single ‘blower’, which draws in cool air from the top of the card and pushes it out the rear of the machine. This design can be particularly beneficial in multi-GPU workstations. Perhaps most importantly, however, is that the Radeon Pro W6800 has a colossal 32 GB of on-board GDRR6 memory, double that of its consumer counterparts, and more than Nvidia’s pro GPUs at the same price point.

Monster memory

t’s been a long time coming but AMD has finally delivered its first professional GPU with hardware ray tracing built in. And with 32 GB of VRAM, the AMD Radeon Pro W6800 is a beast of a graphics card. Priced at $2,249, the Radeon Pro W6800 goes head-to-head with the 16 GB Nvidia RTX A4000 ($1,000) and 24 GB Nvidia RTX A5000 ($2,250), both of which we review on page WS28. In terms of raw performance, the Radeon Pro W6800 sits somewhere between AMD’s ‘consumer’ Radeon RX 6800 and Radeon RX 6800 XT. But as a workstation-class card there are several key differences. First, it will be certified for a wide range of pro applications, including all the major CAD and Building Information Modelling (BIM) tools. This can be especially significant for enterprise customers. It can also support up to six displays, which can be important for powerwalls, and features Error Correcting Code (ECC)

32 GB is a huge amount of memory for a GPU, surpassed only by the 48 GB Nvidia RTX A6000 which costs twice as much. It means the Radeon Pro W6800 can handle some seriously demanding visualisation datasets. This could be a huge multi-disciplinary city-scale model with immense detail or one with less geometry but hyper realistic assets such as 8K textures or detailed vegetation. It’s a huge step up from its predecessor, the AMD Radeon Pro W5700, which only had 8 GB and highlights AMD’s ambitions for high-end design viz and real-time ray tracing. The new GPU features enhanced Compute Units (CU) with dedicated ‘Ray Accelerators’. As this is AMD’s first pro GPU with hardware ray tracing, there aren’t currently a huge number of applications that can take advantage of its ‘Ray Accelerators’, but this is changing. The list currently includes applications that support DirectX Raytracing (DXR), such as Unreal Engine. Also, any that

Solidworks 2021 SP3 (OpenGL) SPECapc benchmark (FSAA) - RealView, shadows & AO 1.23 Benchmark score (bigger is better)

1.72

AMD Radeon Pro W6800

Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

N/A 0.0

0.5

AMD Radeon Pro W5700 2.08 2.29 2.52

1.0

1.5

2.0

2.5

2.26

AMD Radeon Pro W6800

Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

N/A

3.0

0.0

0.5

1.0

3.02 2.90 3.35

1.5

2.0

2.5

3.0

Solidworks Visualize 2021 SP3 (ProRender)

Computer model (denoising disabled)

Computer model (denoising enabled)

1,000 passes, accurate quality (1,500 x 1,500 resolution)

1.23

AMD Radeon Pro W5700

Render time (secs) (smaller is better)

AMD Radeon Pro W6800

Nvidia Quadro RTX 4000

Nvidia RTX A4000

100 passes, accurate quality (1,500 x 1,500 resolution)

358 225

100

150

200

Nvidia Quadro RTX 4000

Nvidia RTX A4000

250

300

350

400

67 45 56 46

3.5

Render time (secs) (smaller is better)

AMD Radeon Pro W6800

Nvidia RTX A5000

174 0

1.23

AMD Radeon Pro W5700

351

211

Nvidia RTX A5000

WS20_21_22_23_24_D3D_AUGSEPT21_RadeonPro.indd 20

The Radeon Pro W6800 supports a new pro driver feature called Radeon Pro Viewport Boost, which is designed to reduce latency and boost viewport navigation performance. It detects when a 3D model is moving quickly in the viewport then

SPECapc benchmark (FSAA) - shaded with edges

AMD Radeon Pro W5700

August / September 2021

Viewport boost

Solidworks 2021 SP3 (OpenGL) 1.23 Benchmark score (bigger is better)

WS20

feature Radeon ProRender 2.0, the latest version of AMD’s physicallybased rendering engine. This includes Solidworks Visualize, Acca Software, Autodesk Inventor, Rhino, Autodesk Maya, and Blender. Looking to the future, it will also extend to any application that supports Vulkan Ray tracing, including those in development at Solidworks (Project Romulan — tinyurl.com/SW-graphics), Autodesk (One Graphics System - see page WS32) and Enscape. The Radeon Pro W6800 will not accelerate ray tracing in Nvidia RTXenabled applications such as Luxion KeyShot, Chaos V-ray, Chaos Vantage, Enscape 3.0 and others. Of course, the Radeon Pro W6800 can also be used for many other applications that don’t rely on hardware ray tracing. This includes those that use the OpenGL or DirectX graphics APIs, including real-time design viz tools like Lumion or Twinmotion, Virtual Reality (VR) or photogrammetry software. The GPU is very much focused on viz and is not optimised for FP64 (Double Precision) code, so applications like engineering simulation will likely to continue to be best served by the AMD Radeon Pro VII ($1,899).

40 20

www.develop3d.com

12/08/2021 14:50

workstation special report

automatically drops the resolution in specific areas to reduce the number of pixels the GPU needs to process. Then, as soon as that movement stops, it restores the full pixel count. According to AMD, this can increase Frames Per Second (FPS) dramatically without impacting the visual experience. AMD Radeon Pro Viewport Boost currently works with Revit, 3ds Max, Twinmotion and Unreal Engine (for packaged projects only, not currently Unreal Engine Editor). Support for other applications is coming soon. We explore this in more detail on page WS26

best with PCIe 4.0 workstations. With double the bandwidth of PCIe 3.0, data can theoretically be fed into the GPU much quicker, although it won’t make a difference in all workflows. PCIe 4.0 compatible CPUs include 11th Generation Intel Core, Intel Xeon W-1300, AMD Ryzen 5000, AMD Ryzen Threadripper 3900X and Threadripper Pro 3900WX series. The AMD Radeon Pro W6800 is also designed to work better with AMD CPUs with AMD Smart Access Memory. This essentially gives the CPU better access

Unreal Engine 4.26

Specifications The AMD Radeon Pro W6800 is the first workstation GPU to be based on AMD’s 7nm RDNA 2 architecture. AMD states peak FP32 Throughput (Single Precision) as 17.83 Teraflops of Compute Performance. It is not optimised for FP64 (Double Precision). With six Mini DisplayPort outputs it can drive up to six displays at 5K resolution or up to two displays at 8K resolution. The board itself is full height, double slot, with a peak power of 250W. It requires a 6-pin and an 8-pin power connector and should fit most mid-sized tower chassis. The AMD Radeon Pro W6800 is a PCIe 4.0 graphics card. While it is fully compatible with older PCIe 3.0 workstations, it’s designed to work

to the GPU’s onboard memory. AMD says it unlocks higher performance for ‘key professional workloads’ but did not elaborate further. The Radeon Pro W6800 also includes 128 MB of AMD Infinity Cache, a ‘lastlevel’ data cache integrated on the GPU die designed to reduce latency and power consumption.

The Radeon Pro W6800 on test We put the AMD Radeon Pro W6800 through a series of real-world application

Lumion 11.5 (DirectX 12 - rendering)

Architectural house

AMD Radeon Pro W5700

Frames Per Second (FPS) (bigger is better)

1.23

AMD Radeon Pro W6800

9.90

Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

18.40 11.60 16.53

22.30

8K (7,680 x 3,840 resolution)

AMD Radeon Pro W5700

AMD Radeon Pro W6800

Nvidia Quadro RTX 4000

Nvidia RTX A4000

510 294 390 280

Nvidia RTX A5000

212 0

Render time (secs) (smaller is better)

1.23

100

200

300

Lumion 11.5 (DirectX 12 - real time)

Lumion 11.5 (DirectX 12 - rendering)

Colossal building (28 GB)

4K (3,840 x 2,160 resolution)

Frames Per Second (FPS) (bigger is better)

1.23

AMD Radeon Pro W5700

2.30

AMD Radeon Pro W6800

6.80

2.20

Nvidia RTX A4000

AMD Radeon Pro W6800

Render time (secs) (smaller is better)

1,283

1,243 673

Nvidia RTX A5000 5

600

218

Nvidia RTX A4000

3.40 1

1.23

500

Nvidia Quadro RTX 4000 3.10

Nvidia RTX A5000

8K (7,680 x 3,840 resolution)

400

AMD Radeon Pro W5700

Nvidia Quadro RTX 4000

WS20_21_22_23_24_D3D_AUGSEPT21_RadeonPro.indd 21

Over the past few years Unreal Engine has established itself as a very prominent tool for design viz, especially in architecture and automotive. It was one of the first applications to use GPU-accelerated real-time ray tracing, which it does through Microsoft DirectX Ray tracing (DXR). It means the AMD Radeon Pro W6800 is fully compatible. For testing, we used two datasets, both freely available from Epic Games: an arch viz interior of a small apartment and the Automotive Configurator, which features an Audi A5 convertible. Both scenes were tested with ray tracing enabled (DirectX Ray tracing (DXR)) and without (DirectX 12 rasterisation). The Radeon Pro W6800 did well with DirectX 12 rasterisation, showing a vast improvement over the Radeon Pro

Lumion 11.5 (DirectX 12 - real time) 4K (3,840 x 2,160 resolution)

www.develop3d.com

benchmarks, for GPU rendering, real-time visualisation and 3D CAD. All tests were carried out using the AMD Ryzen 5000-based Scan 3XS GWPME A132R workstation (see page WS14 for a full review). Resolution was set to 4K (3,840 x 2,160) and we used AMD’s enterprise 21.Q1 graphics driver. For comparison, we used AMD’s previous generation ‘RDNA’ workstation GPU, the AMD Radeon Pro W5700 (8 GB), plus Nvidia’s brand new ‘Ampere’ workstation GPUs, the Nvidia RTX A4000 (16 GB) and Nvidia RTX A5000 (24 GB), which we review on page WS28.

520 0

300

August / September 2021

600

900

1200

1500

WS21

12/08/2021 14:50

workstation special report

W5700, and sitting somewhere between the RTX A4000 and RTX A5000. With real-time ray tracing enabled, however, it fell notably behind both Nvidia GPUs. Without hardware ray tracing built-in, the Radeon Pro W5700 pretty much ground to a halt.

especially in real-time 3D where it was very hard to navigate the scene. With the smaller scene, however, the Nvidia RTX A5000 demonstrated a clear lead over the Radeon Pro W6800 and the RTX A4000 also stood up well, edging out the Radeon Pro W6800 when rendering.

Autodesk VRED Professional 2022

Autodesk VRED Professional is an automotive-focused 3D visualisation, virtual prototyping and VR tool. It uses OpenGL and delivers very high-quality visuals in the viewport. It offers several levels of real-time anti-aliasing (AA), which is important for automotive styling, Lumion 11.5 Enscape 3.0 as it smooths the edges of body panels. Lumion is a real-time rendering tool Enscape is a real-time viz and VR tool for However, AA calculations use a lot of GPU popular with architects. The 11.5 release architects that delivers very high-quality resources, both in terms of processing and uses DirectX 12 rasterisation. It does not graphics in the viewport. The software memory. We tested our automotive model currently support hardware-based ray has used elements of ray tracing for some with AA set to ‘off’ and ‘ultra-high’. tracing. time and version 3.0 is RTX-enabled, so As we have seen previously with AMD The software can work with 8K textures hardware ray tracing is supported on GPUs, the AMD Radeon Pro W6800 did and has a vast object library including Nvidia RTX GPUs. Later versions will OK with anti-aliasing set to off, but was trees with leaves that move still significantly behind in the wind, all of which the RTX A5000. With antican place huge demands aliasing enabled, however, The AMD Radeon Pro W6800 stands out on GPU processing and performance dropped memory. considerably, with even from the competition due to its substantial We tested the GPUs in the RTX A4000 taking a 32 GB of memory, surpassed only by the two ways: one measuring Nvidia RTX A6000 which costs twice as much substantial lead. real-time 3D performance Solidworks Visualize 2021 in terms of Frame Per The name of this GPUSecond (FPS) and two, recording the time it takes to render an use the more modern Vulkan API and accelerated physically-based renderer is a support ray tracing on both Nvidia and bit misleading as it works with many more 8K scene. applications than the CAD application of Lumion supplied us with two datasets: AMD GPUs. For our tests, we used a large scene of the same name. It can import models from a standard architectural house with surrounding vegetation, which will fit a building complex and its surrounding PTC Creo, Solid Edge, Catia and Inventor, into 8 GB of GPU memory; and a colossal area in Enscape 3.0 (non RTX). At 9.5 GB, as well as several neutral formats. The software was initially programmed building model which needs 28 GB, more the GPU memory requirements of this than the capacity of the Nvidia RTX model are relatively high, but Enscape to work with Nvidia Iray and, more recently, Nvidia RTX. However, in the A4000 (16 GB) and RTX A5000 (24 GB). models can be much larger. In terms of performance, the Radeon 2020 release, AMD Radeon ProRender It came as no surprise that the AMD Radeon Pro W6800 came out top when Pro W6800 delivered a very smooth was added, so users now have a choice testing the 28 GB model as it was the only experience at 29 FPS, more than double of two rendering engines. Both support GPU able to load the entire dataset into that of the Radeon Pro W5700. It edged denoising, a post-processing technique memory. The Nvidia RTX A4000 (16 GB) out the Nvidia RTX A4000 but was a bit that filters out noise from unfinished / noisy images and means you can get better and RTX A5000 (24 GB) really struggled, behind the Nvidia RTX A5000.

‘‘

’’

Unreal Engine 4.26 (DirectX 12 - rasterisation)

Unreal Engine 4.26 (DirectX 12 - DXR)

Audi car configurator model (ray tracing disabled)

Audi car configurator model (ray tracing enabled)

4K (3,840 x 2,160 resolution)

AMD Radeon Pro W5700

N/A

AMD Radeon Pro W6800 Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

1.23

Frames Per Second (FPS) (bigger is better)

8.38

Nvidia RTX A4000 33.00

9.40

Nvidia Quadro RTX 4000 24.61

AMD Radeon Pro W6800

13.44

Nvidia RTX A5000 0

19.02 10

Unreal Engine 4.26 (DirectX 12 - rasterisation)

Unreal Engine 4.26 (DirectX 12 - DXR)

Arch Viz interior model (ray tracing disabled)

Arch Viz interior model (ray tracing enabled)

AMD Radeon Pro W5700

1.23

AMD Radeon Pro W6800

Frames Per Second (FPS) (bigger is better)

24.30

Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

46.30 26.89 39.40

1.23

Frames Per Second (FPS) (bigger is better)

AMD Radeon Pro W6800

5.30

Nvidia Quadro RTX 4000

5.21

Nvidia RTX A4000 51.15

4K (3,840 x 2,160 resolution)

AMD Radeon Pro W5700 N/A

WS20_21_22_23_24_D3D_AUGSEPT21_RadeonPro.indd 22

28.83 16.98

4K (3,840 x 2,160 resolution)

August / September 2021

4K (3,840 x 2,160 resolution)

AMD Radeon Pro W5700 N/A

WS22

Frames Per Second (FPS) (bigger is better)

1.23

8.04

Nvidia RTX A5000 0

11.30 6

www.develop3d.com

12/08/2021 14:50

workstation special report

looking renders with significantly fewer rendering passes. We tested both AMD and Nvidia GPUs with Radeon ProRender using the PC model from the SPECapc for Solidworks 2021 benchmark. We rendered at 1,500 x 1,500 resolution with 1,000 passes (denoising disabled) and 100 passes (denoising enabled) with accurate quality. Both settings produced excellent visual results. With denoising enabled, there was little between the Radeon Pro W6800, and Nvidia’s Ampere GPUs, but the RTX A5000 had a bigger lead with denoising disabled.

With six Mini DisplayPort outputs the W6800 can drive up to six displays at 5K resolution or up to two displays at 8K

Solidworks 2021 While most CAD applications won’t benefit from any GPU more powerful that the Nvidia Quadro P2200 or AMD Radeon Pro W5500, Solidworks 2021 is an exception. By using OpenGL 4.5, a more modern version of the popular graphics API, more algorithms can be pushed onto the GPU so there is a benefit to higher performance cards. Even so, the application is still CPU limited to some extent, so the performance benefit of more powerful GPUs isn’t as big as you’d expect from a dedicated real-time viz tool. Like most CAD tools, the most popular way to view models in Solidworks is in shaded with edges mode. Using the SPECapc for SolidWorks 2021 benchmark we saw a small improvement over the Radeon Pro W5700, although the Radeon Pro W6800 was behind both Nvidia GPUs. Solidworks also features more realistic display styles for viewing models in real

With a peak power of 250W, the W6800 requires a 6-pin and an 8-pin power connector

Enscape 3.0 (OpenGL)

Autodesk VRED Professional 2022 (OpenGL)

Large building complex

Automotive model (No Anti Aliasing)

Automotive model (Anti Aliasing - Ultra-high)

4K (3,840 x 2,160 resolution)

1.23

AMD Radeon Pro W5700

Frames Per Second (FPS) (bigger is better)

AMD Radeon Pro W6800

Nvidia Quadro RTX 4000 Inconsistent results Nvidia RTX A4000

AMD Radeon Pro W5700

Nvidia Quadro RTX 4000

Nvidia RTX A4000

Nvidia RTX A5000

36 20

30.65

4K (3,840 x 2,160 resolution)

AMD Radeon Pro W5700

30.88

5.45

Nvidia Quadro RTX 4000 48.80

3.30

AMD Radeon Pro W6800

51.85

7.30

Nvidia RTX A4000 63.30

12.25

Nvidia RTX A5000

VRMark - Cyan Room DirectX 12

Frames Per Second (FPS) (bigger is better)

AMD Radeon Pro W5700

1.23

49.65

Nvidia RTX A4000

Nvidia RTX A5000

99.77 57.7

182.43

AMD Radeon Pro W6800

351.17 173.69

Nvidia RTX A4000 98.25 100

Frames Per Second (FPS) (bigger is better)

Nvidia Quadro RTX 4000 75.1

15.98 10

AMD Radeon Pro W5700

Nvidia Quadro RTX 4000

Frames Per Second (FPS) (bigger is better)

1.23

DirectX 11

AMD Radeon Pro W6800

WS20_21_22_23_24_D3D_AUGSEPT21_RadeonPro.indd 23

Frames Per Second (FPS) (bigger is better)

VRMark - Blue Room 1.23

www.develop3d.com

1.23

AMD Radeon Pro W6800

Nvidia RTX A5000 0

4K (3,840 x 2,160 resolution)

238.4

Nvidia RTX A5000 0

100

August / September 2021

150

323.43 200

250

300

350

400

WS23

12/08/2021 14:50

workstation special report

Arch viz studio Beehive pushed the W6800 to its limits in Lumion on ‘Aedas City’, a visualisation project that features six of the international architecture firm’s building designs. It uses 28 GB (yes, 28 GB) of GPU memory

time. Solidworks RealView, which is only Nvidia RTX A6000 which costs twice as competition from Nvidia. The 16 GB supported by pro GPUs, adds realistic much. But you have to take design viz very Nvidia RTX A4000, for example, materials and supports environment seriously to need such a huge amount. generally offers a little less performance reflections and floor shadows. Meanwhile, Architectural visualisation studio than the Radeon Pro W6800 but costs ambient occlusion adds more realistic Beehive certainly does. It pushed the half as much. Meanwhile, the 24 GB shadows and helps bring out details. W6800 to its limits in Lumion, while Nvidia RTX A5000 offers parity on Both viewing styles are more GPU- working on ‘Aedas City’, a visualisation price, but has a clear performance lead intensive, so performance is less limited project that features six of the international in some workflows and better software by the frequency of the CPU. In our architecture firm’s building designs. compatibility. One can’t help but wonder tests, we saw a bigger benefit to the more And because the project could be held if AMD has missed a trick by not pricing powerful GPUs when RealView, shadows entirely within GPU memory, it managed the Radeon Pro W6800 more aggressively and ambient occlusion were enabled. to massively reduce render times for a 891 to make it more competitive in workflows We were unable to test where large memory capacity the Nvidia RTX A5000 as is less important. Or perhaps Solidworks 2021 Service there’s room for a Radeon Pro There is a strong workflow argument for Pack 3 did not recognise the W6700? card. We expect this to be Nvidia also appears to have having so much memory on a GPU, by not fixed in SP4, out soon. a clear lead in DXR hardware having to worry so much about optimising ray tracing, although this geometry or textures VRMark is perhaps to be expected. We also tested with VRMark, AMD’s ray accelerators are a dedicated Virtual Reality ‘first generation’ and there is benchmark that uses both DirectX 11 and frame video – from 36 hours and 11 mins also scope for driver improvements. DirectX 12. It’s biased towards 3D games, (with the 24 GB Nvidia Quadro RTX 6000 AMD is innovating in other areas, so not perfect for our needs, but should GPU) to 9 hours and 27 mins (with the however. The Radeon Pro Viewport give a good indication of the performance Radeon Pro W6800). Boost, for example, is an exciting one might expect in ‘game engine’ viz There are big benefits for projects like feature that takes a smarter approach to tools, although all datasets are different. this but this is an extreme example. Most how precious GPU resources are The Radeon Pro W6800 came out top design-centric visualisation workflows allocated. And this is certainly one to in the ‘Cyan room’ test which measures require significantly less GPU memory, watch for the future. DirectX 12 performance. AMD itself has although with ever increasing demands We also wait with interest to see how highlighted how its ‘RDNA 2’ architecture for realism and resolution, this will the new 8 GB AMD Radeon Pro W6600 performs well in DirectX 12 applications. likely change in the future. There is also a shapes up, which launched just as we strong workflow argument for more GPU went to press. At $649 it should hit the The verdict memory, by not having to worry so much sweet spot for CAD users who also want a The Radeon Pro W6800 stands out from about optimising geometry or textures. real-time 3D, ray tracing or VR capability. the competition due to its substantial For now, in more mainstream viz We will have a full review in the next 32 GB of memory, surpassed only by the workflows, AMD faces very stiff edition of DEVELOP3D Magazine.

‘‘

’’

WS24

August / September 2021

WS20_21_22_23_24_D3D_AUGSEPT21_RadeonPro.indd 24

www.develop3d.com

12/08/2021 14:50

Now is the time to invest in VDI or Cloud based Workstations. Support your remote workers with the best performance.

Implementing a VDI (Virtual Desktop Infrastructure) solution, will enable your firm to centralise resources, reduce operating and management costs and empower and enable employees to work and collaborate more productively from anywhere, while improving the security and control that your business demands.

On Premise VDI

Private Cloud Desktops

Microsoft Azure

The most cost-eﬀective solution — All the benefits of VDI, utilising Citrix or Vmware, you buy the servers and host them, we deploy it and support you.

IMSCAD provision physical servers with your Applications and Workflows in mind, built to service your specific needs, hosted in our Cloud. Fully Managed service.

Azure Virtual Desktops, Provisioning of Desktop Disaster Recovery – If the worst happened you can fail over to Azure to run all your users in The Cloud.

Benefit from IMSCAD’s experience of deploying 100’s of successful virtual desktop solutions globally. Come and talk to the experts…

Try a VDI solution for free contact@imscadglobal.com | www.imscadglobal.com

IMSCAD Print Ad July 2021 2.indd 1

10/07/2021 11:45

workstation special report

AMD Radeon Pro Viewport Boost In recent years AMD has allocated significant resources to the development of its Radeon Pro graphics drivers. The new 21.Q2 release promises to increase 3D performance by dynamically reducing viewport resolution, and without impacting the visual experience. Radeon Pro Viewport Boost works with any AMD Radeon Pro GPU but Greg Corke tests it out with the new Radeon Pro W6800 tinyurl.com/viewport-boost

t the beginning of June AMD launched the AMD Radeon Pro W6800, a monster 32 GB professional GPU, which we review in-depth on page WS20. Such a huge amount of on-board memory certainly makes the W6800 stand out from other GPUs in its class. However, the ‘RDNA 2’ workstation card also features a new pro graphics driver feature called Radeon Pro Viewport Boost, which is designed to reduce latency and boost viewport navigation performance. The idea behind the technology is simple but smart. It detects when a 3D model is moving quickly in the viewport, then dynamically drops the resolution to reduce the number of pixels the GPU needs to process. Then, as soon as that movement stops, it restores the full pixel count. According to AMD, this can increase Frames Per Second (FPS) dramatically without impacting the visual experience. AMD Radeon Pro Viewport Boost currently works with Autodesk Revit 2021, Autodesk 3ds Max 2021, Twinmotion and Unreal Engine 4 (for packaged DirectX 11 projects only – not currently DirectX 12 or Unreal Engine Editor). Support for other applications is coming soon. It works best in GPU limited workflows. i.e. those where the GPU is being pushed to its limits and is the bottleneck in the workstation. And, with this in mind, it should deliver the biggest benefits at higher resolutions (4K and above), with larger models and when visual quality settings are maxed out.

Testing Radeon Pro Viewport Boost Radeon Pro Viewport Boost is enabled in the AMD Radeon Pro 21.Q2 driver under graphics settings. Users have control over the minimum dynamic resolution that WS26

August / September 2021

WS26_27_D3D_AUGSEPT21_ViewportBoost.indd 26

the application viewport will drop down With this in mind, the benefits of to, expressed as a percentage of its native Radeon Pro Viewport Boost in CAD or resolution. It can be set between 50% and BIM applications like Autodesk Revit are 83.3%. The lower the value, the bigger the less clear. potential performance boost. In Revit, the most popular way to view To show the extent to which it is BIM models is in ‘shaded’ mode. However, working at any given moment, one to with this display style enabled we saw no four small green dots appear in the top benefit to Radeon Pro Viewport Boost. As left corner of the viewport – one being the with many CAD and BIM tools, the GPU least, four being the most. simply isn’t stressed enough, so the CPU The smart thing about Radeon Pro becomes the bottleneck instead. Viewport Boost is that it only works when It’s only when you start ramping up the model is in fast motion, when the eye is the quality settings that more demands less sensitive to a loss of visual detail. are placed on the GPU and Radeon Pro In both of our Unreal Engine arch viz Viewport Boost can come into effect. interior scenes, for example, it only kicked And while the performance increases in when ‘running’ (shift, up arrow) and not can be large, we only found a few select when ‘walking’ (up arrow only). At 50%, scenarios where a substantial benefit the drop in resolution is clearly visible could be seen. but only really when you actively look out We tested with four relatively small for it. At 83.3% it was very hard to see any Revit models and found that if the difference. following criteria were met — realistic When modelling in Revit or 3ds max, display style, smooth lines with antilines become more pixelated. But with aliasing, transparency enabled, and the speed with which one tends to pan, viewport set to perspective mode — then rotate, or zoom-in, to quickly shift focus to there was a huge performance gain; a different part of the model, it’s really not almost double the Frames Per Second. detrimental to the overall experience. But without all of those enabled — We tested on a fairly standard 4K especially with the viewport set to (3,840 x 2,160 resolution) 60Hz IPS orthographic — the performance gains panel. There may be a bigger discernible were minimal, or there were none at all. difference on higher spec We would be interested displays. to learn how the system The performanceworks with significantly It’s good to see benefits can be huge. larger models. AMD innovating In Unreal Engine 4.26, It’s also worth pointing testing with a Paris out here that most CAD by taking a interior scene from and BIM applications smarter approach already have a built-in arch viz artist Benoit to how GPU Derau (benoitdereau.com) feature to help improve resources are we saw frame rates viewport performance more than double when working with large allocated. Why (116%) when minimum models. In Revit, for bother rendering resolution was set to example, the ‘simplify pixels that most 50%. In Unreal Engine’s display during view freely available Arch viz feature people won’t even navigation’ interior scene packaged (which is switched on notice when as a DirectX11 project it by default) suspends models are moving certain graphics effects increased by 59%. at speed? In Twinmotion the and temporarily removes boost was around 30% some objects when the with the ‘materials room’ model is in motion. demo scene when visual settings were set Radeon Pro Viewport Boost gives the to ‘ultra’. In 3ds max we saw around a 20% best performance boost when this feature improvement with AMD’s ‘snow bike’ is enabled, so you are getting a lower res model with high anti-aliasing and “High representation of a model that has already Quality” shading. been simplified. Of course, Revit like many CAD and What about CAD/BIM? BIM applications, is renowned for being For design viz applications like Unreal CPU limited so it’s hardly surprising we Engine and Twinmotion having the found reduced benefits for Radeon Pro highest quality graphics is always the Viewport Boost. This is especially true for ultimate goal. However, in 3D CAD and a high-end graphics card like the Radeon BIM modelling workflows it’s usually less Pro W6800, which is complete overkill important, with the focus instead on the for Revit. clear representation of geometry. In CAD applications that make better

‘‘

’’

www.develop3d.com

12/08/2021 14:53

workstation special report

use of the GPU, such as Solidworks 2021, AMD Radeon Pro Viewport Boost could have a bigger impact in a broader set of viewing styles. It’s also important to note that CAD applications are changing, with new graphics engines that use modern APIs like Vulkan to push more processing onto the GPU and reduce the CPU bottleneck. This includes future versions of Solidworks (Project Romulan tinyurl.com/SW-graphics) and Autodesk Revit (and other Autodesk applications) which will use the new One Graphics System (see page WS32).

With Radeon Pro Viewport Boost enabled and min resolution set to 50% some pixelation is noticeable on this wall painting when ‘running’ in this Unreal Engine scene. The four green dots (top left) show that Viewport Boost is in full effect

What we think It’s good to see AMD innovating by taking a smarter approach to how GPU resources are allocated. Why bother rendering pixels that most people won’t even notice when models are moving at speed? From our tests we see a clear benefit for design visualisation, where applications almost always push the GPU to its limits and visual quality is of paramount importance. We’re less convinced with the broader advantages for CAD and BIM software. In Revit, for example, it appears you have to use a fairly specific combination of visual settings in order to benefit. And, in a workflow where the clear representation of geometry is usually the priority, one would also question how many people actually view models that way. At the moment, application support is quite limited, but this will grow. We imagine AMD is working on support for Unreal Engine Editor as well as DirectX 12, which should be a big attraction for viz artists, especially those working with huge datasets that approach the substantial 32 GB memory limit of the Radeon Pro W6800. When AMD first announced Viewport Boost it was exclusive to the Radeon Pro W6800 and W6600 GPUs. AMD has now confirmed that it will be expanding support to prior generation Radeon Pro GPUs as well. So, for those that already own an AMD Radeon Pro GPU, this could be a great way to get more out of your investment. And it’s perhaps with less powerful GPUs like these, that users will get the biggest benefits. In all of our tests we experienced pretty good viewport performance (most well above 20 FPS) even with Viewport Boost disabled. But it’s when frame rates drop lower, and viewports become choppy, that any performance increase can make a huge difference to practical workflows and become far more important than numbers on charts. www.develop3d.com

WS26_27_D3D_AUGSEPT21_ViewportBoost.indd 27

As soon as you stop ‘running’ the four green dots disappear and the full resolution image is instantly restored

In 3ds Max, with Radeon Pro Viewport Boost enabled and min resolution set to 50%, there is little difference between the moving image (left) and the static image (below) although the lettering on the bike is not as sharp

August / September 2021

WS27

12/08/2021 14:53

workstation special report

Nvidia RTX A4000 / A5000 Nvidia’s new Ampere-based pro GPUs, the Nvidia RTX A4000 and RTX A5000, offer a big step up from the Turing-based Quadro RTX family. With more memory and significantly enhanced processing, they promise to make light work of demanding realtime ray tracing, GPU rendering and VR workflows, writes Greg Corke Price $1,000 (A4000) / $2,250 (A5000) nvidia.com

n February 2021 we reviewed the Nvidia RTX A6000, the first pro desktop GPU to be based on Nvidia’s ‘Ampere’ architecture. With 48 GB of memory and buckets of processing power, the dual slot 300W graphics card is designed for the most demanding visualisation workflows – think city-scale digital twins or complex product visualisations using very hi-fidelity textures, such as those captured from real-life scans. Of course, the Nvidia RTX A6000 is complete overkill for most architects or product designers who simply want a capable GPU for real-time visualisation, GPU rendering or VR. And it’s here that the new Nvidia RTX A4000 and Nvidia RTX A5000 come into play. Announced at Nvidia’s GTC event this year, the PCIe Gen 4 ‘Ampere’ Nvidia RTX A4000 and Nvidia RTX A5000 are the replacements for the PCIe Gen 3 ‘Turing’ Nvidia Quadro RTX 4000 and Quadro RTX 5000, which launched in 2019. The RTX A4000 and A5000 are midrange ‘Quadro’ GPUs in everything but name. Nvidia might be retiring its longserving Quadro workstation brand, but the features remain the same. Both GPUs offer more memory than their consumer GeForce counterparts, are standard issue in workstations from Dell, HP and Lenovo, and come with pro drivers with ISV certification for a wide range of CAD/BIM applications. And with an estimated street price of $1,000 for the Nvidia RTX A4000 and $2,250 for the Nvidia RTX A5000, they have much more palatable price tags than the Nvidia RTX A6000 which costs $4,650. WS28

August / September 2021

WS28_29_30_31_AEC_D3D_AUGSEPT21_Nvidia.indd 28

Nvidia RTX A4000 (16 GB) With 16 GB of GDDR6 ECC memory, the Nvidia RTX A4000 offers a big step up from the 8 GB Quadro RTX 4000. 8 GB is fine for mainstream viz workflows but for more complex projects it can be limiting, so delivering 16 GB in a sub $1,000 pro GPU is a big step forward. Previously, 16 GB was only available on the ‘Turing’based Quadro RTX 5000. As you’d expect from Nvidia’s new ‘Ampere’ architecture, the Nvidia RTX A4000 also offers a significant improvement in processing. This can be seen in all areas of the GPU with more CUDA cores for general processing, thirdgeneration Tensor Cores for AI operations and second-generation RT Cores for hardware-based ray tracing. It leads to a substantial performance increase in many different applications (see later on). Furthermore, as the Nvidia RTX A4000 is a single slot GPU with a max power consumption of 140W delivered through a single 6-pin PCIe connector, it’s available in a wide range of desktop workstation form factors. This includes compact towers like the HP Z2 Tower G8 and Dell Precision 3650. The board features four DisplayPort 1.4a ports and can drive up to four displays at 5K resolution. It is cooled by a single ‘blower’ type fan, which draws in cool air from the top and bottom of the card, pushes it through a radiator and then directly out of the rear of the workstation chassis. This is in contrast to most consumer GeForce GPUs which use axial fans that recirculate air inside the machine. There are pros and cons to each design, but with a blower fan you can stack cards within the chassis without having to leave space between them. This means you can get a very good density of GPUs inside a mid-sized chassis. With the AMD Threadripper Pro-based Lenovo ThinkStation P620, for example, you could get four Nvidia

RTX A4000s back-to-back, which could be a very interesting proposition for GPU rendering. Even though the RTX A4000 doesn’t support NVlink (so there’s no pooling of GPU memory) 16 GB is still a good amount and two, three or four RTX A4000s could work out well in terms of price/performance compared to the more powerful RTX A5000 or A6000. Another potential use case for highdensity multi-GPU is workstation virtualisation using GPU passthrough, where each user gets a dedicated GPU. Again, this workflow looks well suited to the Lenovo ThinkStation P620, which can be configured with up to 64 CPU cores and 2 TB of memory. Other more niche pro viz features include support for 3D Stereo, Nvidia Mosaic for professional multi-display solutions, and Quadro Sync II, an addin card that can synchronise the display and image output from multiple GPUs within a single system, or across a cluster of systems.

Nvidia RTX A5000 (24 GB) With 24 GB of GDDR6 ECC memory, the Nvidia RTX A5000 offers only a 50% memory uplift compared to the Quadro RTX 5000 it replaces. Like the Nvidia RTX A4000 it offers a significant upgrade in all areas of processing — CUDA, Tensor and RT cores. It’s a double height board, with a max power consumption of 230W which it draws from the PSU via an 8-pin PCIe connector, but it’s still available in compact towers. The board also features four DisplayPort 1.4a ports and is cooled by a single ‘blower’ type fan, but only draws in cool air from one side of the card. The Nvidia RTX A5000 supports all the same features as the Nvidia RTX A4000 but differs in two main areas. One, it supports Nvidia NVLink, so GPU memory can be expanded to 48 GB by connecting two 24 GB GPUs together. Two, it supports Nvidia RTX vWS (virtual workstation software) so it can deliver multiple high-performance virtual workstation instances that enable remote users to share resources. In the AMD Threadripper Pro Lenovo ThinkStation P620, for example, which we reviewed earlier this year, you could get a very high density of CAD users who only need high-end RTX performance from time to time. www.develop3d.com

12/08/2021 15:03

workstation special report

Nvidia RTX A5000 inside the AMD Ryzen 5000-based Scan 3XS GWP-ME A132R workstation

Testing the new cards We put the Nvidia RTX A4000 and Nvidia RTX A5000 through a series of real-world application benchmarks, for GPU rendering, real-time visualisation and 3D CAD. All tests were carried out using the AMD Ryzen-based Scan 3XS GWP-ME A132R workstation at 4K (3,840 x 2,160) resolution using the latest 462.59 Nvidia driver (see page WS12 for a full review). For comparison, we used the last two generations of ‘4000’ class Nvidia pro GPUs – the 8 GB ‘Turing’ Nvidia Quadro RTX 4000 (from 2019) and the 8 GB ‘Pascal’ Nvidia Quadro P4000 (from 2017). Three to four years is quite a typical upgrade cycle in workstations, so the intention here is to give a good idea of the performance increase one might expect from an older machine (N.B. to see all of the benchmark scores for the Nvidia Quadro P4000 visit tinyurl.com/ RTX4000). We also threw some Nvidia RTX A6000 scores in there. These were done on two different workstations with a 32-core Threadripper Pro 3970X and a quad core Intel Xeon W-2125 CPU. While both CPUs have lower frequencies and instructions per clock (IPC) the results should still give a pretty good idea of comparative performance, especially in GPU rendering software. The results of these tests can be seen in www.develop3d.com

WS28_29_30_31_AEC_D3D_AUGSEPT21_Nvidia.indd 29

the charts on page WS20 in our review of ‘real-time’. Vantage, for example, is built the AMD Radeon Pro W6800. from the ground up for real-time ray tracing so can maximise the usage of RT Hardware-based ray tracing cores within the new GPUs. It’s been just over two years since Nvidia introduced ‘Turing’ Nvidia Quadro RTX, Chaos Group V-Ray its first pro GPUs with RTX hardware ray V-Ray is one of the most popular tracing. physically based rendering tools, In a classic chicken and egg launch, there especially in architectural visualisation. were very few RTX-enabled applications We put the new cards through their paces back then, but this has now changed. For using the freely downloadable V-Ray 5 design viz, there’s Chaos V-Ray, Chaos benchmark, which has dedicated tests for Vantage, Solidworks Visualize, Autodesk Nvidia CUDA GPUs, Nvidia RTX GPUs, VRED, KeyShot Unreal Engine, Unity, as well as CPUs. Enscape, D5 render, Nvidia Omniverse, The results were impressive. In the Siemens NX Ray Traced Studio, Catia Live CUDA test, the Nvidia RTX A4000 rendering and others. was 1.62 times faster than the previous Nvidia RTX gave GPU rendering a generation Nvidia Quadro RTX 4000 massive kick start and while there is and in the RTX test 1.70 times faster. The increased competition from hugely lead over the Pascal-based Quadro P4000 powerful CPUs like the 64-core AMD was nothing short of colossal – 3.53 times Threadripper [Pro], we are seeing deeper faster in the CUDA test. As the P4000 penetration of GPU rendering tools, does not have dedicated RT cores, it could especially in architect / engineer / product not run the RTX test. designer friendly workflows. Stepping up to the Nvidia RTX A5000 Nvidia RTX is being used to massively will give you an additional boost. accelerate classic viz focused ray trace Compared to the Nvidia RTX A4000 it renderers like V-Ray, KeyShot and was between 1.27 and 1.37 times faster. Solidworks Visualize, which we test later Interestingly, the RTX A5000 was not on in this article. However, some of the that far behind the RTX A6000, which more exciting developments are coming costs more than twice as much. from the Architecture, Engineering and Construction (AEC) sector in tools like Luxion Keyshot Enscape, Chaos Vantage and Unreal KeyShot, a CPU rendering stalwart, is a Engine, which really make ray tracing relative newcomer to the world of GPU August / September 2021

WS29

12/08/2021 15:03

workstation special report

rendering. But it’s one of the slickest implementations we’ve seen, allowing users to switch between CPU and GPU rendering at the click of a button. In the Keyshot 10 benchmark, part of the free KeyShot Viewer, the performance leap was even more substantial than in V-Ray. The Nvidia RTX A4000 and Nvidia RTX A5000 outperformed the Quadro RTX 4000 by a factor of 1.89 and 2.51 respectively. And the RTX A5000 was only 20% slower than the RTX A6000.

Solidworks Visualize

being able to render at such speeds is quite incredible and can have a profound impact on workflows. In comparison, it took the Quadro P4000 GPU 105 seconds, so you can see just how far things have progressed in four years.

Real time 3D While GPU rendering is a major play for the Nvidia RTX A4000 and Nvidia RTX A5000, real-time 3D using OpenGL, DirectX and (in the future) Vulkan continues to be a very important part of visualisation, with applications including Unreal Engine, Unity, TwinMotion, Lumion, Enscape, LumenRT and others. Of course, the boundaries between realtime 3D and ray tracing continue to blur. In fact, out of the list above only Lumion and Twinmotion are yet to support RTX, although it should be coming to Twinmotion soon. To test frame rates, we used a combination of monitoring software including FRAPS, Xbox Game Bar and

will still likely need the top-end Nvidia RTX A6000 especially for high-res VR workflows. In Unreal Engine we used two datasets: an arch viz interior of a small apartment and the Automotive Configurator, which features an Audi A5 convertible. Both scenes were tested with ray tracing enabled (DirectX Ray tracing (DXR)) and without (DirectX 12 rasterisation). The results were pretty much as expected with good scaling between all the GPUs with DirectX 12 rasterization. With real-time ray tracing enabled, performance naturally takes a hit in general, but the Quadro P4000 really suffers without any RT cores.

The name of this GPU-accelerated physically based renderer is a bit misleading as it works with many more applications than the CAD application of the same name. It can import models from Creo, Solid Edge, Catia and Inventor, as We also tested with VRMark, a dedicated well as several neutral formats. Virtual Reality benchmark that uses Since the 2020 release the software has DirectX 11 and DirectX 12. In the DX12 supported Nvidia RT cores and Tensor test both GPUs came in second to AMD’s cores to improve rendering performance Radeon Pro W6800 (see page WS20). with Nvidia RTX GPUs. Users can choose Computer Aided Design to render scenes with or without denoising enabled. The Nvidia RTX A4000 Denoising is a postand RTX A5000 are The performance leap from the Quadro RTX processing technique overkill for most 3D based on machine learning CAD applications and 4000 to RTX A4000 is hugely impressive, that filters out noise from are unlikely to give you and the step up from the four-year old unfinished / noisy images significantly better 3D ‘Quadro P4000 is simply phenomenal and is the foundation for performance than more many RTX-accelerated mainstream GPUs like the applications. It means Nvidia Quadro P1000 / you can get better looking renders with MSI Afterburner. We only tested at 4K P2200, or the new Nvidia T1000. significantly fewer rendering passes. (3,840 x 2,160) resolution. At FHD (1,920 x However, CAD applications are changing DS Solidworks reckons that if a scene 1,080) resolution this class of GPU simply and, in the future, should be able to make routinely needs 500 passes without the isn’t stressed enough. much better use of the plentiful power of denoiser, then you may be able to achieve higher-end GPUs like the RTX A4000 the same rendering quality with 50 passes In Autodesk VRED Professional we and A5000. with the denoiser enabled. tested our automotive model with AA set In addition, it is important to note that We tested the stock 1969 Camaro car to ‘off’ and ‘ultra-high’. both GPUs will be certified for a wide model at 4K resolution with 1,000 passes Considering that this pro viz range of pro CAD applications, which (denoising disabled) and 100 passes application used to only really run is important for some firms. This is (denoising enabled) set to accurate quality. effectively on Nvidia’s ultra-high-end especially true for enterprises that buy Both settings produced excellent visual professional GPUs, it’s quite astounding 100s or 1,000s of workstations from large results. that the Nvidia RTX A4000 – a sub OEMs like HP, Dell and Lenovo and want The RTX A4000 and RTX A5000 $1,000 card – delivered over 30 FPS at assurance that the GPUs will be stable and delivered the 100-pass render in 22 4K resolution with medium anti-aliasing. that they will be properly supported by seconds and 14 seconds respectively. In saying that, those really pushing the the software developer. This isn’t the most complex scene but boundaries of automotive visualisation Certification is a major reason why

‘‘

’’

Solidworks Visualize 2021 SP3 (Iray)

1969 Camaro car model (denoising disabled)

1969 Camaro car model (denoising enabled)

1,000 passes, accurate quality 4K (3,840 x 1,080 resolution)

1.23

Render time (secs) (smaller is better)

Nvidia Quadro P4000

100 passes, accurate quality 4K (3,840 x 1,080 resolution)

1,029

Nvidia RTX 4000

349

Nvidia RTX A4000

200

600

800

1000

1200

Nvidia Quadro P4000 1

105

14 0

100

August / September 2021

WS28_29_30_31_AEC_D3D_AUGSEPT21_Nvidia.indd 30

32.94

Nvidia RTX A4000 1

Nvidia RTX A5000 1

Nvidia RTX A6000 2

120 1

WS30

5.60

Nvidia Quadro RTX 4000 1

Nvidia RTX A5000 400

1.23 Relative performance to reference system (bigger is better)

Nvidia RTX A4000

140

Render time (secs) (smaller is better)

Nvidia RTX 4000

211

Nvidia RTX A5000

Nvidia Quadro P4000

1.23

Luxion KeyShot 10 benchmark (GPU)

62.34 82.60

AMD AMD Ryzen 5950X CPU (462.59 driver)

99.62 60

100

AMD Threadripper 3970X (461.09 driver)

www.develop3d.com

12/08/2021 15:04

workstation special report

V-Ray 5 image courtesy of Toni Bratincevic

Solidworks 2021 can make better use of powerful GPUs than most CAD applications but it is still CPU limited to some extent, so the performance benefit the new cards give you isn’t as big as you’d get from a dedicated real-time viz tool. In the SPECapc for SolidWorks 2021 benchmark we saw a small improvement, generation on generation with the shaded with edges. The Nvidia RTX A4000 was 1.10 times faster than the Quadro RTX 4000 and 1.44 times faster than the ‘Pascal’ Quadro P4000. With RealView, Shadows and Ambient Occlusion enabled we saw a bigger benefit over older GPUs. The Nvidia RTX A4000 was 1.16 times faster than the Quadro RTX 4000 and 1.57 times faster than the Quadro P4000. We were unable to test the Nvidia RTX A5000 as Solidworks 2021 SP3 did not recognise the card. We expect this to be fixed in SP4, out soon.

Conclusion With the new Nvidia RTX A4000 and A5000, Nvidia has made its ‘Ampere’ GPU architecture much more accessible to a wider audience. In particular, we see the sub $1,000 Nvidia RTX A4000 hitting the sweet spot for designers, engineers or architects that want a pro viz capability in their workflow. The performance leap from ‘Turing’ to ‘Ampere’ (Quadro RTX 4000 to RTX A4000) is hugely impressive. In realtime 3D, a 45% to 60% boost, generation on generation, seems typical, with even bigger gains from real-time ray tracing when the enhanced RT and Tensor cores come into play. The step up from the four-year old ‘Pascal’ Quadro P4000 is phenomenal, especially for rendering. Equipping the RTX A4000 with 16 GB of memory is very significant. While we often see models/scenes that surpass 8 GB (the capacity of the previous generation Quadro RTX 4000) scenes that are 16 GB and above are certainly less common, and more the preserve of viz specialists than most architects or product designers who use standard materials and assets. For viz workflows that need lots of memory, Nvidia has strong competition from the 32 GB AMD Radeon Pro W6800,

Chaos Group V-Ray 5.0 benchmark

V-Ray GPU CUDA

V-Ray GPU RTX

290

Nvidia Quadro P4000 1 634

Nvidia RTX A4000 1

Nvidia RTX A6000 2

AMD AMD Ryzen 5950X CPU (462.59 driver)

1500

AMD Threadripper 3970X (461.09 driver)

1,559

Nvidia RTX A5000 1 1,531

1000

919

2,128

Nvidia RTX A6000 2 2000 1

500

AMD AMD Ryzen 5950X CPU (462.59 driver)

1000

2,490 1500

2000

2500

AMD Threadripper 3970X (461.09 driver)

August / September 2021

WS31

www.develop3d.com

200

400

500

Nvidia RTX A4000 1 1,299

N/A

Nvidia Quadro RTX 4000 1 1,025

Nvidia RTX A5000 1

For more performance data see charts on page 20 as part of the in-depth review of AMD Radeon Pro W6800 with 32 GB of on-board memory

1.23 vrays (calculations per minute) (bigger is better)

1000

Nvidia Quadro P4000 1 Nvidia Quadro RTX 4000 1

In the original edit of this article we concluded by discussing the need for Nvidia to flesh out its Ampere workstation family and bring RTX capabilities to the lower-end of the market. Then, just as DEVELOP3D was going to press, Nvidia announced the Nvidia RTX A2000 (6 GB), a $450 pro RTX GPU to replace the Pascal-based Quadro P2200. The interesting thing about the RTX A2000 is that it can be configured for both Small Form Factor (SFF) workstations and standard towers. We’re excited by what this will bring to CAD-centric workflows but wonder if it might be a little light on memory. See page 8 in the main magazine to find out more.

800

1.23 vpaths (calculations per minute) (bigger is better)

but in less demanding worflows Nvidia’s biggest competitor in pro graphics is currently itself. The new 12 GB ‘consumer’ GeForce RTX 3080 Ti, for example, might have half the memory of the Nvidia RTX A5000, but offers more performance on paper for half the price. Nvidia even has a GeForce Studio driver for applications including Enscape, Unreal Engine and V-Ray. Despite the obvious attraction of Nvidia’s consumer GPUs, Nvidia’s ‘A’ class models should continue to find favour in large firms and enterprises that buy in volume, want more memory, consistent supply, pro viz features or the assurance of certification.

600

some firms choose Nvidia’s pro-focused RTX GPUs over their ‘consumer GeForce’ counterparts so they can confidently use applications like Revit, Solidworks, PTC Creo, and Siemens NX alongside more vizfocused tools like Chaos V-Ray, Enscape, Luxion KeyShot and Solidworks Visualize.

WS28_29_30_31_AEC_D3D_AUGSEPT21_Nvidia.indd 31

12/08/2021 15:04

workstation special report

Example of ambient occlusion in Autodesk’s path traced reference application Image courtesy of Autodesk / Nvidia GTC

The Autodesk Inventor graphics boost Autodesk’s new One Graphics System (OGS) for Inventor, 3ds max and other Autodesk applications promises to improve viewport performance and visual quality through GPU ray tracing, writes Greg Corke

utodesk is working on a new graphics engine that promises to ‘radically improve’ 3D performance in its core applications, including Inventor, 3ds max, Revit and others. The ‘next-generation’ system will also improve visual quality by delivering ray tracing directly inside the viewport. The aim is to enable users to switch between a simple shaded view and a photorealistic representation very quickly – though the speed at which this can happen will depend on the graphics hardware and the complexity of the model. WS32

August / September 2021

WS32_33_34_D3D_AUGSEPT21_inventor.indd 32

The new viewport system is a complete re-architecting of the current One Graphics System (OGS), which relies solely on rasterisation, a rendering method for 3D software that takes vector data and turns it into pixels (a raster image). OGS is shared by nearly all Autodesk products, including Inventor, Revit, 3ds max, AutoCAD, Maya, Fusion, Infraworks, Navisworks and Recap. It dates back to 2007 and uses older graphics APIs including OpenGL and DirectX 9/10/11. Inventor, like many other CAD applications that use older graphics APIs, isn’t the best at harnessing the plentiful power

of modern Graphics Processing Units (GPUs). As a result, 3D performance in the software is largely bottlenecked by the frequency of the CPU. Autodesk’s ‘next generation’ viewport system should fix this, and users should get a much smoother viewport experience without having to rely on the software to temporarily simplify or discard parts of the view during model navigation by setting a minimum frame rate. By fully utilising modern graphics APIs including DirectX 12 and Vulkan, Autodesk’s applications will be able to talk straight to the GPU hardware (rather www.develop3d.com

12/08/2021 10:17

workstation special report

than having to route everything through the graphics driver) and take better advantage of multi-core CPUs. The modernised OGS will also be able to use hardware-based ray tracing, which is built into modern graphics cards such as the new ‘Ampere’ Nvidia RTX A2000, A4000, A5000 or A6000 (see page WS28), or ‘RDNA 2’ AMD Radeon Pro W6600 or W6800 (see page WS20). It will utilise open standards, such as MaterialX and Pixar’s Hydra, part of USD.

Nvidia GTC showtime Autodesk’s Henrik Edstrom, senior software architect, graphics technology, and Mauricio Vives, senior principal engineer, gave a sneak peek of the ‘next-generation’ viewport system in a presentation at Nvidia’s GTC conference earlier this year. Rather than looking at performance in simple shaded views, the focus was on the eye-catching stuff — real-time ray tracing. The idea is that when working on a model in the viewport, a product designer or engineer can quickly switch to full ray traced quality at the click of a button. Inventor currently relies on CPU rendering which takes a long time and the results can be mixed. For the new engine, Vives explains that the emphasis is to provide a good real time experience, like you get with a game engine. “Photorealism is important, but we prioritise performance over quality,” he says. According to Edstrom, it will be very easy to use for non-rendering experts. “They just need a very simple solution with literally no settings,” he says. In Inventor, for example, we imagine it will be just another visual style to add to the current drop-down for wireframe, shaded, realistic, monochrome, watercolour and sketch / technical illustration.

Lights, camera, action In a demonstration of a path traced reference application, Vives showed a welding robot model that originated in Inventor. Render quality was photorealistic and looked impressive, much better than the rasterisation engine currently used in Inventor 2022. The scene was lit by a single directional light and a low frequency environment light. Vives showed off the render quality by highlighting several aspects within the scene, including contact hardening soft shadows, real ambient occlusion and glossy reflections, focusing in on the subtle reflection of a black socket on a red paint finish. www.develop3d.com

WS32_33_34_D3D_AUGSEPT21_inventor.indd 33

In order to achieve a smooth viewport experience, the frame budget was set to 30 frames per second. This resulted in some ‘noise’ when moving the camera, but the image resolved very quickly when idle by casting more rays and using denoising. The presentation then went on to show the path tracing mode of the new engine working inside in Autodesk 3ds Max, Maya and Inventor. You can watch the full presentation at tinyurl.com/ OGS-Autodesk (registration required).

But, at Autodesk University 2019, when we asked what Autodesk was doing to improve 3D performance we got a rather prickly response from CEO Andrew Anagnost. He implied this wasn’t an issue. While Autodesk would not publicly admit the current engine needed improving, we expect work on the new OGS had already started back then. Swapping out a graphics engine is a huge undertaking, especially one that covers so many diverse applications as OGS does. The emphasis is The fact is, there are Flexible architecture to provide a good many Autodesk customers To provide more flexibility that struggle with large real time for future software and performance, from experience, like model hardware developments, load times to model disyou get with a the new viewport system is play and manipulation, based on a decoupled relying on their software game engine. architecture, where the “Photorealism is to automatically simplify applications and the or turn off important, but we geometry graphic system are not as effects during view navigaprioritise intertwined as they are in tion in order to maintain the current OGS. This is performance over frame rates. being done through Hydra, Everyone wants more quality,” says part of Pixar’s USD. 3D performance, and we Mauricio Vives, expect many will be excit“That will allow us to senior principal ed by this news. Even make changes and improve OGS and add more so, if they can get this engineer, support for new features without having to upgrade Autodesk and new APIs and new GPU hardware. According hardware and so on, withto Vives, basically any out really involving the GPU, up to five years old application teams too much,” explains from any vendor, will support the modEdstrom. “It also gives us the flexibility ernised OGS. to plug in other renderers, like Arnold, The addition of real-time ray tracing in for instance, that already supports the the viewport is also a very exciting develHydra render delegate API, or some- opment. Rather than relying on slow thing like the Nvidia Omniverse RTX CPU ray tracing on typical 4, 6 or 8 core renderer which also supports this API… workstations, it will help bring visualisand there are numerous others as well.” ation more upfront in the design process The system currently uses DirectX – not just for presentations but to better Raytracing (DXR), but Autodesk is also planning to use the recently released Vulkan Ray tracing extensions.

‘‘

’’

Conclusion In DEVELOP3D’s workstation tests over the years we’ve experienced first-hand how applications like Inventor and Revit can be very CPU limited when it comes to 3D graphics and unable to take advantage of the huge processing potential of modern GPUs.

Autodesk’s new One Graphics System (OGS) will be able to take advantage of new generation GPUs like the AMD Radeon Pro W6600 which include hardware-based ray tracing

August / September 2021

WS33

12/08/2021 10:17

workstation special report

1 Example of glossy reflections in Autodesk’s path traced reference application 2 Example of soft shadows in Autodesk’s path traced reference application Images courtesy of Autodesk / Nvidia GTC

understand complex forms through more realistic shading or to explore materials and the impact of light. Those looking to take advantage of realtime ray tracing, however, will need to invest in a GPU with built in hardware ray tracing. Currently, this would be an Nvidia RTX GPU (see page WS28) or the latest AMD Radeon Pro GPUs (see page WS20) or Intel GPUs in the future. So when can we expect to see this new engine inside shipping Autodesk products? In short, it’s too early to tell. In the presentation it was described as a ‘very early work in progress’ and there were no promises as to what will or will not be supported in Autodesk’s products. But this is a standard line for a publicly traded company. The fact that Autodesk is showing this in public now perhaps shows promise for its 2023 or 2024 series of products. While it’s good to see Autodesk actively looking to improve the 3D graphics expe-

Enhanced graphics for Solidworks

Autodesk is not the only company embracing modern graphics APIs. In February, Dassault Systèmes demonstrated Project Romulan, a new Vulkan-based graphics engine for Solidworks, designed to bring enhanced visual fidelity, GPU-accelerated ray tracing, and better performance for large assemblies to the 3D CAD tool’s viewport.

rience across the board, it’s long overdue and, with rise of powerful real-time engines like Unreal Engine, designers now expect a fast and visually rich viewport in CAD tools like Inventor and Revit.

We’re excited to see how this develops, although that same enthusiasm might not be shared by firms that already have optimised visualisation workflows. ■ tinyurl.com/OGS-Autodesk

Ray tracing and denoising are used to calculate ‘accurate ambient occlusions’ with the resulting soft shadows (see below) overlaid on top of the model

■ tinyurl.com/

SW-graphics

WS34

August / September 2021

WS32_33_34_D3D_AUGSEPT21_inventor.indd 34

www.develop3d.com

12/08/2021 10:17

ROM F

F L HE OS HT C T KE S I D E A T I ON

A R T U O G I R F S N CO

T I O A N R O L P X &E

P R O D U CT

HOT S Y HKE T WI

O N EC L I C TT O H OV S Y E R KK

L I DF 3

CO MMU T N O N AI I C

EI MP O T R

KEY

L UA S I V

SH OT NT RI T P O3 DP

SeehowKeyShoti sus edt ovi s ual i z eyourdes i gn t hr oughoutt hepr oductdev el opmentl i f ecycl e. Ke y S h o t . c o m/E n v o y VI EWI N YOURHOME!

Workstations for Media & Entertainment Workflows 3XS GWP-ME A124C Mini

3XS GWP-ME A132R

• AMD Ryzen 9 5900X • 32GB DDR4 • 16GB NVIDIA A4000 • 1TB M.2 • Microsoft Windows 10 Pro

• AMD Ryzen 9 5950X • 64GB DDR4 • 16GB NVIDIA RTX A4000 • 2TB M.2 SSD • Microsoft Windows 10 Pro

• AMD Ryzen 9 5950X • 64GB DDR4 • 24GB NVIDIA RTX A5000 • 2TB M.2 SSD & 2TB HDD • Microsoft Windows 10 Pro

From

£2,399.99

INC VAT

£3,399.99

INC VAT

£4,299.99 INC VAT

3XS GWP-ME A132R

3XS GWP-ME A1-48T

3XS GWP-ME A264T

• AMD Ryzen 9 5950X • 64GB DDR4 • 48GB NVIDIA RTX A6000 • 2TB M.2 SSD & 2TB HDD • Microsoft Windows 10 Pro

• AMD Ryzen Threadripper 3960X • 128GB DDR4 • 48GB NVIDIA A6000 • 2TB M.2 & 4TB SSD • Microsoft Windows 10 Pro

• AMD Threadripper 3970X • 128GB DDR4 • 2x 48GB NVIDIA RTX A6000 • 2TB M.2 & 2TB HDD • Microsoft Windows 10 Pro

From

£6,799.99 INC VAT

£7,999.99 INC VAT

Discover your perfect workstation at

scan.co.uk/m&e • 01204 47 47 47

From

£13,499.99 INC VAT

workstation special report

Hybrid working what does it mean for design firms? With many firms re-evaluating office space and working from home policies, we asked Adam Jull of IMSCAD about the role that virtual workstations can play – and the differences between VDI, public and private cloud

he pandemic has shifted the way we all work. As the dust settles and we try to get back to some sort of normality, a hybrid work model seems to be the favoured approach of employers and employees alike. But what does this really mean for design firms who have historically invested a lot in their offices / studios to provide great spaces for client facing and for collaboration amongst staff? The nature of hybrid work means that staff will be in the office less and split their time between the office and home – or a remote location that is not the traditional office. As a result, less office space is required, but will this translate into firms reducing their office footprint, which could save considerable cost, or will it

Learning by example VDI and private cloud

Parkhill, a US architect and engineering firm with nine offices spread across Texas, Oklahoma and New Mexico, has deployed a successful VDI solution that gives 400+ creative professionals the freedom to work anywhere, on any device using all of their applications including the Autodesk suite. Parkhill invested in an on-premise solution, with 16 servers running 430 desktops using

www.develop3d.com

WS37_38_D3D_AUGSEPT21_IMSCAD.indd 37

‘‘

As we start to come out the other side of the pandemic, and attitudes to the traditional office continue to change, firms are looking for more robust IT solutions to support flexible working

’’

lead to firms changing the way they use their existing office space? The answer is both. Those firms that can downsize now, surely will. However, those that cannot (perhaps they have a longer lease they cannot get out of) will still likely change the way they use their existing space. In both cases, it is likely that fewer desks will be needed, staff will be hot desking, and more collaboration and meeting spaces will be required. With staff working from home for two or three days a week, firms need to ensure they invest in the right remote graphics technology solutions so everyone can work from anywhere productively.

specific applications that require a lot of resource so they can be used to their full potential. Investing in the right technology to facilitate this move to hybrid working is paramount. VDI (Virtual Desktop Infrastructure) is a proven solution for giving design users the performance they require when working remotely. VDI technology enables you to deliver desktops via virtual machines (VMs) and the desktops are managed from a central server. There are several ways for firms to deliver these virtual desktops to their users. On-premise or on-prem VDI is deployed on your own infrastructure. This could be one server in your office or a private data centre, where you are responsible for the infrastructure and everything associated with it. A real positive of this sort of solution is that the servers can be provisioned to suit your use case, your specific application-mix and the individual user workloads. Private cloud or hosted VDI is when virtual desktops are outsourced. They The tech options are deployed and hosted in a hosting In the manufacturing company’s datacentre, on bare metal space, the majority of servers, and the hosting company is designers use industry- responsible for the infrastructure. Like the

Citrix. In light of the pandemic, Parkhill will tell you it has gained an advantage over its peers from a mobility standpoint and an IT management perspective. The solution has come into its own during the pandemic, with nearly all staff working from home and now working two or three days a week away from the office. On a smaller scale is a UK-based manufacturing firm that

was looking to minimise future disruption on productivity, in terms of hours and days lost with users not being able to work to their full potential when they could not access the office during the pandemic. The firm needed a solution to enable remote work for 50+ design users and chose a private cloud solution. The solution included three servers, two of which include two GPUs,

providing desktops to 20 users each and the third, one GPU providing desktops for the further 12 users. Another example of a successful on-premise VDI solution is that of HUNT EAS, a New York-based engineering, architectural and surveying firm. Hunt had an existing Virtual desktop solution in place, but the solution had never performed, and the company was looking to upgrade from

August / September 2021

the existing four server environment providing desktops for 65 users to one providing desktops to 90+ users. The new environment consists of five servers and provides all users great performance running AutoCAD, Autodesk Revit and other applications. It has proved invaluable over the last year through the pandemic as well as when major snowfall affected New York last winter.

WS37

12/08/2021 15:00

workstation special report

on-premise solution, firms get the same benefits of provisioning infrastructure to match their use cases, applications and workflows with the extra advantage of not having to buy or manage the hardware. Hyperscale or Public Cloud Desktopse.g. Microsoft Azure, AWS, Google Cloud. This is where a firm pays for ‘instances’ of virtual desktops. While spinning up a new instance is easy, a VDI solution on the public cloud still requires extra underlying infrastructure and integration components to work. From the architecture and networking to the security, the customer is responsible for everything they build in the public cloud.

(Capital Expenditure) required as the infrastructure needs to be purchased, racked and set up before you can start provisioning the desktops. Once fully functioning, the cost are minimal - yearly licence renewals and support. With a private or hosted cloud solution, there will be an upfront fee to set up and then a monthly cost, keeping CapEx to a minimum and enabling a more manageable OpEx (Operational Expenditure) model. When looking at a Public Cloud VDI solution there are no upfront costs - you just pay for the instances. However, you will need to add additional products and components to your virtual What are the benefits? infrastructure, from domain controllers The benefits of these types of solutions are to VMs and each comes with an associated many - most importantly furnishing your cost on top of the cost of each instance. users with the performance they require when working both in the office and at One step at a time home - or anywhere else for that matter. Not all firms move to VDI in one go. Historically, users have been reluctant Some firms continue to utilise their to give up their physical workstations physical workstations and often use the as they have worried about not getting VDI solution to facilitate remote work. the same performance when using other Often they will sweat the workstation technologies. until end of life, by which time the user This was highlighted for many as they is accustomed to working on a virtual tried to work from home on hastily put desktop and the change to doing so full together remote work solutions that did not time is not so dramatic. give the performance they required when Another facet to this is those firms they suddenly had to vacate the office. that are starting small, providing a VDI Your IP is obviously very important and solution for a single team or for a specific security is another great benefit of this type project - in essence using this as a pilot. of solution. Some firms will actually invest in a single With all data being stored centrally, server either on-premise or a private no data leaves the infrastructure. With cloud, specifically as a pilot project and applications also kept centrally on the get users from all over the firm to test. Both servers, all patches, upgrades and fixes are sensible approaches. can be managed centrally as a single There are also firms employing a hybrid instance and the management of users, approach, as it is a big ask to migrate to the their access to applications and data is cloud in one single leap. You also need to also centrally controlled by IT which factor in that public cloud providers might will result in a reduction in IT not be the most cost effective for the types management costs. of applications and workloads that design With users not relying on their physical and engineering firms are using. workstations and not having to be at their desks, firms are far more resilient to any Closing thoughts interruptions. In addition to obvious When the Covid-19 pandemic first hit, pandemics, this could be due to bad many manufacturing firms had to quickly weather stopping travel or a fire at the adapt existing IT to allow designers and office. With a virtual desktop solution engineers to work effectively from home. there is a built-in business continuity Data centralisation was key, so remoting capability, meaning users can remain into office workstations quickly became a working and therefore productive no temporary fix for some. matter what happens. Now as we start to come out the other side, and attitudes to the traditional What are the real costs? office continue to change, firms are Cost is obviously a major factor in looking for more robust IT solutions to determining the solution that you will go support flexible working. And proven with, and the costs associated with each of technologies like VDI present a huge these deployment methods will vary. opportunity for those looking to fully When looking at an on-premise embrace a truly hybrid future. solution, there is a large CapEx ■ imscadglobal.com WS38

August / September 2021

WS37_38_D3D_AUGSEPT21_IMSCAD.indd 38

DEVELOP3D’s cloud workstation roundup Amazon Web Services (AWS) offers two cloud workstations; G4dn uses Intel CPUs and Nvidia GPUs while G4ad features AMD CPUs/GPUs. AWS uses the latest AMD Radeon Pro GPUs but doesn’t offer the same flexibility as Microsoft in terms of how GPU resources can be allocated. BOXX has a different take to other cloud workstation providers insofar as it takes its high-performance desktop workstations and makes them available to users of CAD and demanding viz applications over a 1:1 connection. The service is currently focused on the US, but is coming to the UK soon. Cloudalize has lots of experience of 3D CAD and BIM applications. It offers a ‘pay-per-use’ or an ‘unlimited-usage’ subscription model and a big choice of cloud workstations out of its private datacentres Google offers a big choice of GPUs for its virtual workstations including the Nvidia P4, T4, and P100 GPUs. Several come preconfigured in the Google Cloud Marketplace. Microsoft Azure NVv4 virtual workstations stand out because they offer huge flexibility in how GPU resources can be allocated. The AMD Radeon Instinct MI25 GPUs can be virtualised at a hardware level, so you only pay for the GPU resources you need, as opposed to many other cloud workstations which use a whole GPU. Nutanix Frame is a Desktop-as-a-Service (DaaS) solution designed specifically to deliver 3D applications or desktops to any device. The flexibility of the Frame service allows firms to select their preferred infrastructure (Azure, AWS, GCP or hybrid environments powered by Nutanix AHV) to allow them to reach users in more than 200 countries. The company places a big emphasis on ease of deployment. Scan, best known for its desktop workstations, teamed up with Ebb3 for its Scan 3XS Cloud Workstation. Read this DEVELOP3D article to learn more tinyurl.com/ebb3-scan Tehama offers an enterprise desktop-as-aservice with a big emphasis on security. It allows firms to create ‘secure virtual rooms and desktops’ on the cloud. Workspot aims to simplify the deployment and management of cloud workstations by delivering virtual desktops as a service. This includes Windows 10 desktops and Windows 10 cloud workstations, including the AMD-based Microsoft Azure NvV4 Virtual Machines (VMs). The company has extensive knowledge of the AEC sector and is now working with Siemens to bring NX to the cloud.

www.develop3d.com

12/08/2021 15:00