Table of Contents
Currently, Solidworks officially supports NVIDIA Quadro and AMD FirePro discrete video cards so if you want to use a supported card (which we highly recommend doing) you will need to use a workstation-class GPU. However, if you are a student learning Solidworks or otherwise unable to use a Quadro card, it is entirely possible to use a GeForce or Radeon card instead. You may not be able to get support from Solidworks if you have a problem with their software (which makes it a very bad idea for a professional environment) but if that does not matter to you then it can be a way to get started with Solidworks on a computer you already own.
In this article we will be looking at a range of NVIDIA GeForce cards and comparing them to the three most popular NVIDIA Quadro cards for Solidworks. To fully test the cards we will be using a work around to enabling features like RealView which are normally disabled unless you have a workstation card. This is absolutely something we don’t recommend doing if you are using Solidworks professionally, but it will help us determine the full performance differences between Quadro and GeForce cards. If you want to skip over our individual benchmark results and simply view our conclusions, feel free to jump ahead to the conclusion section.
For our test system, we used the following hardware:
|CPU:||Intel Core i7 6700K 4.0GHz Quad Core 8MB 95W|
|RAM:||4x Crucial DDR4-2133 4GB (16GB total)|
|GPU:||GeForce GTX Titan X 12GB (~$970 – 6144 GFLOPS)
GeForce GTX 980 Ti 6GB (~$650 – 5632 GFLOPS)
GeForce GTX 980 4GB (~$520 – 4616 GFLOPS)
GeForce GTX 970 4GB (~$350 – 3494 GFLOPS)
GeForce GTX 960 4GB (~$240 – 2308 GFLOPS)
GeForce GTX 950 2GB (~$160 – 1573 GFLOPS)
|Hard Drive:||Samsung 850 Pro 512GB SATA 6Gb/s SSD|
|OS:||Windows 10 Pro 64-bit|
|PSU:||EVGA SuperNOVA 850W P2|
|Software:||Solidworks 2016 SP 0.1|
Our test platform is built around an Intel Core i7 6700K as that is the CPU that should give the best possible performance in Solidworks when rotating models. For the video cards we will be testing the full range of current GeForce GTX cards as well as the three most popular Quadro card for Solidworks. In the chart above we listed both an estimate price for the card as well as the peak theoretical floating point performance (which is single precision) so it will be very interesting to see if the driver and firmware differences between the cards makes Quadro faster than it’s rough performance equivalent.
Remember that Quadro (and FirePro) cards are the only officially supported video cards, so in order to fully test the performance differences we had to do a bit of a workaround in order to allow us to compare the two product lines with features such as Realview enabled. However, we want to be clear that this workaround isn’t something we would advise you do on your own machine.
To perform the actual benchmarking, we used a mix of AutoIt scripts and Solidworks macros to set the different quality settings, load the relevant model, and record the average FPS while rotating the model. Note that we tested with different LOD settings, but we found the difference to be marginal so to keep things simple we will only be reporting the results with LOD off (which usually results in a small drop in FPS). There are a number of different ways we could have recorded the FPS, but we opted to simply use a macro with a timer to rotate the model 45 degrees to the left and right for a set number of frames. From the number of frames and the total time it took to render those frames, we are able to determine the average FPS (frames per second). One key factor is that we made sure that every model started with the view set to front-top so that any reflections and shadows would stay in view while the model was being rotated.
For our test models, we chose the following models available from GrabCad.com that will give us results for a range of model complexities based on the total number of parts and number of triangles. These models are:
One note that we would like to make is that if you do not know how many triangles the models you work with have, the easiest method we know of to find out is to simply save the model as an .STL file. During the save process, a window should popup with information about the model including the number of files, the file size, and the number of triangles.
Steam Engine with Horizontal Beam and Centrifugal Pump
To start things off we are going to look at the simplest of our test models which has only 80 parts and .26 million triangles.
With just the “Shaded” view mode the GeForce cards did O.K. although without RealView not even the most expensive GeForce card was any better than even a low-end Quadro. With RealView on the lower end GeForce cards were actually a bit better than the Quadro K620. However, to match the Quadro K2200 you would need a GeForce GTX 980 Ti which is actually a bit more expensive. At the upper end, none of the GeForce cards we tested were able to match the performance of the Quadro M4000.
Where the results get interesting is when we used the “Shaded w/ Edges” view mode. With edges enabled we saw absolutely terrible performance with the GeForce cards. Even a GeForce GTX 980 Ti or Titan X were nowhere near the performance of even a Quadro K620 which less than a third the price.
The Spalker model contains about 4.5 times more parts and twice the number of triangles than the previous model. Our results are somewhat similar to what we saw with the Steam Engine model, although using the “Shaded” view mode the GeForce cards performed much worse than a Quadro K620 without RealView. With RealView, all of the GeForce cards only barely outperformed the K620. Interestingly, all the GeForce cards performed roughly on par to each other with the exception of the GTX 950.
Switching to the “Shaded w/ Edges” view mode, we again see massive performance drops with the GeForce cards to the point that the Quadro K620 handily beats even the highest end GeForce card.
While the Audi R8 model doesn’t have many more parts than the Spalker model, it does have about 3 times the number of triangles making it the most complex model we will be testing.
In “Shaded” mode without RealView, the results are pretty similar to what we saw with the other two models. While the GTX 980 Ti is able to match the Quadro K620, all of the other models see lower performance. With RealView enabled, however, most of the GeForce cards actually performed about on par with the Quadro M4000. Interestingly, after about the GTX 960, we did not see a significant increase in performance.
With “Shaded w/ Edges”, the pattern of poor performance with GeForce cards continues. Once again, even the Quadro K620 will very handily outperform any GeForce card.
Steam Engine with Horizontal Beam and Centrifugal Pump
Even with our most simple model, we start to see a bit better performance out of the GeForce cards at 4K resolutions. While at 1080p we were mostly simply matching the performance of the Quadro K620, now the GeForce cards are often outperforming the Quadro K620 and at times even the Quadro K2200.
Starting once again with “Shaded” mode, all the GeForce cards give performance that is right between the K620 and K2200 without RealView. With RealView enabled, performance is even better with all the GeForce cards outperforming the Quadro K2200 although none were able to match the Quadro M4000
With “Shaded w/ Edges”, even the highest end GeForce card still performs worse than the Quadro K620 without RealView enabled. With RealView on, the performance ranges from a bit worse than the Quadro K620 to a bit worse than the Quadro K2200 depending on the model of GeForce card.
With a bit more complex model, we again see performance higher than the Quadro K620 with the “Shaded” view mode. With RealView disabled, the GeForce cards at best matched the Quadro K2200 with most of them falling somewhere between the K620 and K2200 in terms of performance. With RealView on, however, only the GTX 950 performed worse than the Quadro K2200 with all the other performing a bit better (although still not the the level of the Quadro M4000)
Switching to “Shaded w/ Edges”, the GeForce cards still mostly outperformed the Quadro K620, although none of them were able to match the Quadro K2200.
With the Audi R8 model, the results are a bit different than what we saw previously. Starting again with the “Shaded” view mode, we saw performance that ranged from matching the Quadro K620 to being a bit less than the Quadro K2200 without RealView. With RealView enabled, performance was actually signficantly better with even the GTX 950 outperforming the Quadro K2200. At best, however, even the highest end GeForce card was only able to match the Quadro M4000.
Using “Shaded w/ Edges”, performance with the GeForce cards again takes a nose dive. Without RealView, none of the GeForce cards were able to perform even close to the Quadro K620. With Realview, however, the GTX 970/980 roughly matched the Quadro K620 while the higher end GeForce cards performed right between the K620 and K2200.
One thing is obvious from our results: Quadro cards overall have much better performance in Solidworks than GeForce cards. At 1080p, not even the highest end GeForce card was able to match the very affordable Quadro K620. A GeForce GTX 980 Ti or Titan X should be still be able to give you about 60 FPS for medium models (around .5 million triangles) and about 30 FPS for very complex models (around 1 million triangles), but you would see significantly higher performance with a Quadro card of roughly the same price. If you use a 4K display, however, many of the GeForce cards are able to out-perform the Quadro K620 although given that the K620 only has 2GB of VRAM we would never recommend using it with a 4K display in the first place. Really, the Quadro K2200 is the lowest end card we would recommend for 4K and that card should always outperform even the highest end GeForce card.
If you only use the “Shaded” view mode, the GeForce cards usually give performance somewhere in-between the Quadro K2200 and Quadro M4000 at both 1080p and 4K resolutions. What is interesting is that at 1080p the performance difference between the different GeForce models is actually very minimal. At 4K, however, there is a benefit to having up to a GTX 970, but beyond that the performance does not increase significantly.
Using the Audi R8 model specifically (since that is the most complex model), we can make the following rough estimations of the performance of GeForce cards compared to Quadro cards if we average the results with RealView both on and off:
|Shaded view mode||1080p||4K|
|GeForce GTX 950 2GB||Equivalent to K2200||18% slower than K2200|
|GeForce GTX 960 4GB||5% faster than K2200||10% slower than K2200|
|GeForce GTX 970 4GB||3% faster than K2200||Equivalent to K2200|
|GeForce GTX 980 4GB||3% faster than K2200||5% slower than K2200|
|GeForce GTX 980 Ti 6GB||9% faster than K2200||7% faster than K2200|
|GeForce GTX Titan X 12GB||5% faster than K2200||2% slower than K2200|
|Shaded w/ Edges view mode||1080p||4K|
|GeForce GTX 950 2GB||40% of K620||32% of K2200|
|GeForce GTX 960 4GB||45% of K620||38% of K2200|
|GeForce GTX 970 4GB||54% of K620||52% of K2200|
|GeForce GTX 980 4GB||60% of K620||52% of K2200|
|GeForce GTX 980 Ti 6GB||65% of K620||62% of K2200|
|GeForce GTX Titan X 12GB||66% of K620||62% of K2200|
Overall, with just “Shaded” mode the GeForce cards mostly performed within 10% of the Quadro K2200. However, the big thing we found is that Quadro cards have significant improved performance when using the “Shaded w/ Edges” view mode to the point that even a Quadro K620 will often give significantly better performance than the highest GeForce card.
The reason behind this is not easy to determine, although our best guess is that it has to do with the firmware and driver optimizations used with Quadro cards. Either way, if all you have is a GeForce card then Solidworks should function OK – although you will not be able to use features like RalView – but we would very highly recommend upgrading to an appropriate Quadro card as soon as possible. Especially if you use Solidworks professionally, the extra performance (not to mention simply using a supported card) means that you should almost never consider using a GeForce card instead of a Quadro card.
Puget Systems offers a range of powerful and reliable systems that are tailor-made for your unique workflow.