Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1548
Article Thumbnail

V-Ray Next GPU Roundup: NVIDIA GeForce RTX SUPER Performance

Written on August 9, 2019 by William George
Share:

Introduction

V-Ray Next, from Chaos Group, is made up of a pair of rendering engines: one that uses the CPUs (processors) and another which focuses on GPUs (video cards). We have already tested modern CPUs on V-Ray, using the benchmark utility they make publicly available, but since GPUs can be utilized too it is important to look at how various video cards perform as well.

With the launch of NVIDIA's new GeForce RTX "SUPER" series of video cards, we are taking a look at how the whole RTX lineup performs on the GPU side of V-Ray Next. We are not including the CPU in this testing, so that we can focus on GPU performance alone.

Looking for a Rendering Workstation?

Puget Systems offers a range of workstations that are tailor made for your unique workflow. Our goal is to provide 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 is a list of the hardware we tested V-Ray Next on. Everything except the video card was kept the same throughout all the test runs, to ensure that no other specs could impact performance. The operating system was a fully updated installation of Windows 10 Pro.

Benchmark Details

Chaos Group provides an excellent benchmark for V-Ray Next, which is available for free (registration required) on their website. We ran this benchmark three times per video card, and are using the highest scores for each GPU in the graphs below.

Benchmark Results

Here are the V-Ray Next Benchmark 4.10.06 results (in millions of paths per second) for all of the GeForce RTX video cards, along with the TITAN RTX for good measure, sorted from most to least expensive - with the new SUPER series cards in a darker shade of green:

V-Ray Next Benchmark 4.10.06 NVIDIA GeForce RTX, RTX SUPER, and TITAN RTX rendering performance

And here is another way of looking at the results, as percentages relative to the GeForce RTX 2060's performance:

V-Ray Next Benchmark 4.10.06 NVIDIA GeForce RTX, RTX SUPER, and TITAN RTX rendering performance relative to the RTX 2060

Analysis

V-Ray Next seems to behave differently from other GPU rendering engines we have tested recently, and in particular shows no tangible performance benefit from using the new RTX 2070 SUPER or 2080 SUPER cards compared to the non-SUPER models. However, the RTX 2060 SUPER does show a big improvement over its namesake... and in fact, it performs just as well as the other, more expensive SUPER cards. Considering that it has the same amount of VRAM as well (8GB vs the RTX 2060's 6GB) that makes it a fantastic value!

On the high end, the RTX 2080 Ti and TITAN RTX are less than 10% apart in performance - and since the TITAN costs more than twice as much, and is only available in a dual-fan layout, the 2080 Ti is usually going to be the better choice there. If you need more VRAM than it offers, the Quadro RTX series has options beyond even what the TITAN provides and works great in multi-GPU configurations.

Conclusion: What Are the Best Consumer GPUs for Rendering in V-Ray Next?

Because of the Titan RTX's cooling limitations, and how well V-Ray and other GPU rendering engines can scale across multiple cards, the best choice for many users is going to be the GeForce RTX 2080 Ti. If that is a little too expensive for you, the new GeForce RTX 2060 SUPER also offers great performance for the price - on par with the various 2070 and 2080 models at a much lower price.

Don't forget that V-Ray Next GPU can also utilize the CPU to further boost performance, so pairing a few of those cards with a powerful CPU will give you even faster rendering speeds.

Looking for a Rendering Workstation?

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

Configure a System!

Tags: GPU, Rendering, Benchmark, Performance, NVIDIA, GeForce, RTX, Turing, RTX 2070, RTX 2080, RTX 2080 Ti, Titan RTX, RTX 2060, V-Ray
Tavo Ortiz

Does de RTX 2070 Super supports NVLink?
Can you test two RTX 2070 Super in NVLink? They could be faster than a single RTX 2080 Ti (and a bit cheaper)

Posted on 2019-08-10 22:17:45

Yes, the RTX 2070 Super supports NVLink (the older, vanilla 2070 does *not*). However, NVLink does not directly increase performance - a program has to be coded specifically to utilize it, and even then the main advantage I am aware of (for rendering applications) is the opportunity to pool GPU memory and work on a larger scene size than the cards could individually. That doesn't improve performance, though. Having two cards in a system (with or without NVLink) would improve rendering speed, and I do plan to look at GPU scaling in V-Ray Next at some point, but that will be *without* NVLink because I'm not even sure if the V-Ray Next benchmark is NVLink aware... and even if it is, the test scene it uses does not need additional VRAM.

Posted on 2019-08-12 16:19:02
Mark

Why the 2060 Super better then 2080 super?some thing wrong?

Posted on 2019-08-10 23:27:36
Tim E.

Yes someting very wong

If they did a few more tests and posted the average and the highest the 2080 would probably come out ahead. I hope it would anyways.

Either way it's better than AMD as predictions from tech leaders are saying by 2023 all.basic games will need Ray tracing just to run. I don't know about you but I'm not buying another $499 card in 5 years!

Posted on 2019-08-11 21:34:15
Misha Engel

Guru3D had kind of the same results.
https://www.guru3d.com/index.php?ct=articles&action=file&id=53529
https://www.guru3d.com/articles_pages/asus_radeon_rx_5700_xt_rog_strix_review,25.html

Chaos Group still has a lot of work to do.

Posted on 2019-08-12 15:23:25

Good find! I'm glad to see another site with the same sort of results had. It looks like they are using the older V-Ray benchmark, though, since they got time to render a scene in seconds (which is how the old test gave results) while V-Ray Next's benchmark gives the actual millions of paths per second being traced. If anything, though, that backs these results up as *more* reliable since the same sort of behavior is being observed over two generations of the software.

Posted on 2019-08-12 16:31:46

The higher-end cards certainly do perform better in other GPU based rendering applications. I published articles looking at OctaneRender and Redshift last week, and GPU performance in both of those follows the "normal" price / performance curve as I would have expected:

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

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

Posted on 2019-08-12 16:33:17

Something is strange, at the very least. I double-checked my test results before publishing, and this is definitely the accurate data for the V-Ray Next benchmark. There are a few possibilities I can think of:

1) The way V-Ray Next uses the GPU actually performs better on the RTX 2060 SUPER than on the 2070 or 2080 (either vanilla or SUPER versions). This seems unlikely on the face of it, since the higher-end cards have both more CUDA cores and higher clock speeds... but this is what the data points to at a surface level. If this is correct, then it could be something wrong in the V-Ray render pipeline which a future update might easily fix.

2) There is some sort of bottleneck on V-Ray Next performance given how much RAM is on the GPU. All of the 8GB cards (the 2060 SUPER and both versions of the 2070 and 2080) are in the same ballpark, with the older RTX 2060 6GB slower and the RTX 2080 Ti 11GB and TITAN RTX 24GB both faster. This idea fits the data, but also makes no sense since the scene being rendered doesn't seem that complicated and VRAM usually just limits how complex a scene can be for rendering, rather than directly impacting performance.

3) The particular scene used in the V-Ray Next benchmark could be amiss, and not really representing actual V-Ray Next rendering performance. This seems plausible, but I am still not sure in what way a scene could be set up such that it would give this sort of result. In order to dig into this further, I would need to get my hands on a full license of V-Ray Next and a "better" test scene, and then try to do some separate, manual benchmarking. I'm not sure if / when I might have time to do so :/

I do plan to revisit V-Ray Next in the not-too-distant-future, as I want to look at GPU scaling, but I will be wary because of these results. Maybe I will see if I can get in touch with someone at Chaos Group and see what they say.

Posted on 2019-08-12 16:28:58

Matt and I were just discussing this, and he suggested that it could be memory bandwidth limitations. The 2060 SUPER and both 2070s and 2080s all have very similar memory bandwidth, so if V-Ray Next is bottlenecking on how fast data can be moved into and out of the GPU then that could explain the results we are seeing here. I may try and do a little more testing focused on that, but I've also got messages out to some old contacts at ChaosGroup so hopefully we can see what they say.

Posted on 2019-08-12 17:43:49
@Kv3rul4nt

Are you going to do new tests now that proper GPU acceleration finally has come to Lightroom? I tried it with my 1080 Ti yesterday, and the brush tool actually doesn't lag anymore!

Posted on 2019-08-14 07:21:26

Lightroom testing is on a bit of a pause right now while we update our testing process. We really, really want to figure out a good way to testing things like brush lag and other things in the develop module since that is where most of the performance come from. Things like import/export and generating previews are easy to test, but they actually utilize the hardware in completely different ways than more "active" tasks. So we want to get that testing sorted out first if we can.

Posted on 2019-08-14 16:54:22