Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1681
Article Thumbnail

SOLIDWORKS 2020 SP1 CPU Performance

Written on February 28, 2020 by William George
Share:

Introduction

With the recent release of SOLIDWORKS 2020's first service pack, as well as multiple new CPU launches from Intel and AMD in the past few months, we thought it would be a good time to do a roundup of processor performance in modeling, rendering, and simulations within SOLIDWORKS. In the past we have found this application to vary greatly in how it uses the CPU, with some functions being single-threaded while others are able to use all the cores effectively - and, of course, a spread of behavior in-between those extremes.

Looking for an Engineering Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Test Hardware

Here are the detailed specs of the test platforms we used:

AMD Ryzen Test Platform
CPU AMD Ryzen 9 3950X
AMD Ryzen 9 3900X
AMD Ryzen 7 3800X
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte X570 AORUS ULTRA
RAM 4x DDR4-2933 16GB (64GB total)
Video Card NVIDIA Quadro P6000 24GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit
SOLIDWORKS 2020 SP1
Intel Core Test Platform
CPU Intel Core i9 9900K
Intel Core i7 9700K
CPU Cooler Noctua NH-U12S
Motherboard Gigabyte Z390 Designare
RAM 4x DDR4-2666 16GB (64GB total)
Video Card NVIDIA Quadro P6000 24GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit
SOLIDWORKS 2020 SP1
AMD Threadripper Test Platform
CPU AMD TR 3990X
AMD TR 3970X
AMD TR 3960X
CPU Cooler Noctua NH-U14S TR4-SP3
Motherboard Gigabyte TRX40 AORUS Pro WiFi
RAM 4x DDR4-2933 16GB (64GB total)
Video Card NVIDIA Quadro P6000 24GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit
SOLIDWORKS 2020 SP1
Intel Core X Test Platform
CPU Intel Core i9 10980XE
Intel Core i9 10900X
CPU Cooler Noctua NH-U12DX i4
Motherboard Gigabyte X299 Designare EX
RAM 4x DDR4-2933 16GB (64GB total)
Video Card NVIDIA Quadro P6000 24GB
Hard Drive Samsung 960 Pro 1TB
Software Windows 10 Pro 64-bit
SOLIDWORKS 2020 SP1

Benchmark Details

Our SOLIDWORKS benchmark suite was originally developed by one of my colleagues here at Puget Systems: Matt Bach. He put together a series of AutoIt scripts that run through testing a variety of the capabilities in SOLIDWORKS, which I have updated and added to over the years. I have been aided in that process by the help of many readers who have suggested changes, provided additional files for testing, and more. I have done my best to cite their contributions when changes they help with are made to our testing.

Most recently, before this last round of benchmarks, I added a rebuild test with the help of Issac Roberts, an Aerospace Engineer. I was having trouble because our former rebuild file was taking only fractions of a second for a full rebuild in SW 2020, far too short to get any actionable data. Issac provided an artificially complex file that was built specifically to take a lot longer to rebuild - more so that most real world assemblies, but it worked well to give us more measurable times so that we can look at the behavior of different CPUs in this area.

The whole battery of tests was run multiple times on each CPU, with the fastest result (lowest time) used for this article. We didn't have any significant outlier results and saw very little variance between runs, so we opted for this method over an average of scores. The results are broken up into individual graphs below and followed by our analysis.

Results & Analysis

Here are galleries of the results from each part of our SOLIDWORKS testing. AMD chips are shown in red - darker for Threadripper, lighter for Ryzen. Intel processors are in blue, similarly with Core X models in a darker shade and the mainstream Core in a lighter color.

SOLIDWORKS start up, file open / close, rebuild, and motion study performance

In this first set of data, we can see that the tested Intel processors are somewhat faster for starting up SOLIDWORKS itself as well as opening files, though saving files is a more mixed bag - and has much less overall variance. Motion studies also show a pretty small variance, but slightly favor Intel again.

That brings us to rebuild testing, and this deserves some explanation. Our previous test file for rebuild was only taking fractions of a second in SW 2020, so a reader (Issac Roberts) reached out to me to help provide a more complex assembly that was specifically tailored to increase rebuild time... and boy did it work! However, it uncovered something odd: the times that SW itself reported for rebuilding the assembly were drastically different from the actual time that the system was unresponsive while working on it. As such, I started recording both numbers - and have provided them here on separate charts. In both cases, AMD's processors were faster for rebuilding than the Intel chips we tested - but the difference in actual, real-world time was far smaller than the numbers which SW reports. Remember that getting to these results required extra work, though, and most files won't take anywhere near this long to rebuild.

Various SOLIDWORKS simulation tests

Simulations in SOLIDWORKS come in many varieties, and show equaly varied performance across the CPUs we tested. The Stress simulation seems to be the biggest outlier, which had low core count processors doing the best... but the overall variance wasn't huge. Close behind are the Thermal and Airflow simulations, which both also had little variance between most of the processors - with the exception of AMD's Threadripper 3990X, which did particularly poorly. I don't have an explanation for that, but it may have to do with how many cores & threads it supports (64 / 128, respectively) interacting badly with SOLIDWORKS. That is just a guess, though.

Moving on to our newer "Benchmark" simulations, Intel's top Core i9 and AMD's 24- and 32-core Threadripper processors all did very well - but Intel has a price advantage at that level, making the Core i9 10980XE the best value / performance choice for this workload.

SOLIDWORKS PhotoView 360 rendering

PhotoView 360 is where high core count processors really shine, since CPU-based rendering like this scales extremely well. Because of that, we see all three of AMD's 3rd Gen Threadripper chips doing extremely well here. If you do a lot of rendering, the speed those processors offer can easily outweigh their slower performance in other areas simply because of how long rendering itself often takes. The Ryzen 9 3950X also puts in a very respectable showing here, as does Intel's Core i9 10980XE if you do lots of simulations as well.

The last gallery, below, has a different color scheme. These charts are looking at frame-rates while rotating an assembly, with varying quality settings. Instead of using colors to separate processors, this time the colors correspond to the settings used in the viewport.

Assembly rotation performance

And lastly, looking at framerates while rotating assemblies - all of these CPUs are more than fast enough for general modeling work! There is some variation, mostly limited to low quality modes, but the selection of video card ends up having a much bigger impact on this aspect of SOLIDWORKS. For more information on that, check out our companion SOLIDWORKS 2020 SP1 GPU performance article.

Conclusion

It is tricky to make a single recommendation when separate aspects of an application behave differently, so this conclusion is broken down based on usage:

  • For general engineering work, along with motion studies and simple simulations, Intel's Core i9 9900K is a great choice
  • For more in-depth simulations, Intel's Core i9 10980XE offers some of the best performance at a great price
  • For rendering, AMD's 3rd Gen Threadripper chips are clear winners - but other high core count processors do well too

We offer SOLIDWORKS-specific workstations focusing on the first two of those categories, and have other rendering-optimized systems for those whose workload is heaviest in that area.

Looking for a SOLIDWORKS Workstation?

Puget Systems offers a range of workstations that are tailor-made for your unique workflow. Our goal is to provide the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Tags: Dassault, Systemes, CPU, Processor, Performance, Intel, Core, i7, i9, Solidworks, AMD, AMD Ryzen 3rd Gen, AMD Threadripper 3rd Gen, Threadripper, Core X, Ryzen, Rendering
Misha Engel

https://www.youtube.com/watch?v=Vhh_GeBPOhs

Posted on 2020-02-29 12:40:10
Phillip LaPlante

Hello,

Wow did SW 2020 service pack 1 really increase the Flow Simulation performance that much? In the 500k flow Simulation, your last article put the 3900x at 199s to 219s (depending on 3200MHz memory or 2666MHz), with this article's test at 147s. In our own testing whit this model on SW 2019 SP3 we get 201s on a 3900x, 189s on a 3950x and 164s with a 3970x (with 3600MHz, 3600MHz and 3200MHz memory respectively). So that means that with 2020 SP1, the 12 core processor beats any processor out there running SW 2019. Did anything else change with this benchmark? With those results I could bump up to the 1000k mesh and get nearly the same solve time.

Thank you,
Phillip LaPlante
Motordyne Engineering

Posted on 2020-03-02 17:26:00

I think it must have, because I see that sort of performance increase in other processors as well. For example, look at the Intel Core i9 9900K - which has been around long enough to have been included in two rounds of SW 2019 testing here, as well as this SW 2020 test:

https://www.pugetsystems.co... - 432 seconds at 1000K mesh
https://www.pugetsystems.co... - 431 seconds at 1000K mesh
https://www.pugetsystems.co... (this article) - 281 seconds at 1000K mesh

Nothing changed in the Benchmark Simulation test files or scripts between 2019 and 2020, so I assume any performance change like what we are seeing above is due to improvements in the SW Flow Sim code.

Posted on 2020-03-02 17:57:08
Phillip LaPlante

Yup I just updated and this improvement is huge. I got about a 30% decrease in solve time on my machines in the 500k mesh. Additionally running parallel simulations, the performance increased more: I can run two simulations at once on the 3970x faster than I could one simulation on SW 2019 (148s vs 164s). And the 3970x running 16 solvers at once with 4 cores went from taking 1050s to 558s, a 46% decrease in solve time.

Posted on 2020-03-03 17:10:25

Awesome, that's excellent :-)

Posted on 2020-03-03 17:15:27
FlowJoe

Yes, William, based on information that I received from the developers, they confirm speed-up improvement in the solver with their own testing of the Flow Simulation 2020 release.

Posted on 2020-03-27 15:56:13

Awesome, thank you for checking on that :)

Posted on 2020-03-27 21:10:53
delcolux

Thank you for this testing. Like Phillip, I noticed in your numbers there are some metrics that improved over your previous testing. Rebuild times are marginally improved vs the 2019SP1 testing, but I noticed the file open time was better in 2019SP1. For File Open, the 9900K went from 14sec in 2019SP1, to 19.7sec in 2019SP3, then 15.4sec in 2020SP1. I wonder if it is typical for solidworks to slow down as the service packs stack up, then get faster again upon a new yearly release.

Posted on 2020-03-10 16:23:41

Huh, that is an interesting idea! I really should start paying closer attention to the differences in performance across multiple articles / testing sessions / versions :)

Posted on 2020-03-13 18:11:08
Evan

Is there an article discussing what the various benchmarks are/mean? For instance, there are benchmark simulations with various K numbers, which I assume are DOF, but there is no mention on if they are Static or other simulations, are they only for the meshing step?

Posted on 2020-06-02 07:20:25

A very fair question! We used to include the description of the tests in each article, but it felt repetitive so I'd dropped it in the last couple. I had switched to talking about any *new* test sections in each article, but not rehashing the old ones. Here is a link to the section containing info on those in a former publication:

https://www.pugetsystems.co...

There are two set of tests that will not describe, though: the "Benchmark" simulation, which is a conjugate heat transfer airflow simulation done at three different mesh sizes (500,000 | 1,000,000 | 2,000,000 - more about it here: https://www.pugetsystems.co... and the Rebuild test, which is described in this article. Hopefully that info helps :)

Posted on 2020-06-02 17:49:52
Evan

Thanks for the detailed response William.

What would be the procedure to use some of these results in a webinar discussing Solidworks Simulation Performance troubleshooting?

Posted on 2020-06-16 14:28:40

I think it should suffice to cite us & our testing / articles the source of the info. You can refer to the test itself as the "Puget Systems SOLIDWORKS benchmark", and to us as a company by name ("Puget Systems"). If it is published beyond the live webinar, a link to the article where you got the results would be good. Bonus points for mentioning we are a workstation manufacturer ;)

Posted on 2020-06-16 16:33:05
Michael Grimm

Have you seen or do you expect any performance improvements from Intel's latest 10-series processors?

Posted on 2020-06-15 15:33:51

That is actually something I plan to test in the coming days, so you can probably expect an article about it in the next few weeks :)

Posted on 2020-06-15 22:51:37