Puget Systems print logo

https://www.pugetsystems.com

Read this article at https://www.pugetsystems.com/guides/1559
Article Thumbnail

V-Ray Next Multi-GPU Performance Scaling

Written on August 20, 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 CPU (processor) and another which focuses on GPUs (video cards). One of the big advantages of that type of GPU-based rendering is that you can easily put multiple video cards inside a single workstation. How much benefit does each additional card provide for V-Ray Next, though? We put four GeForce RTX 2080 Ti video cards to the test to find out!

V-Ray Next Logo

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 the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Benchmark Details

To see how increasing the number of video cards in a system affects performance in V-Ray Next GPU, we ran the free benchmark that Chaos Group provides on their website. For hardware, we kitted out a single workstation with four NVIDIA GeForce RTX 2080 Ti video cards - variants with blower-style fans, to ensure that the cooling layout wasn't going to be a problem. Here are the full system specs:

Test Platform
CPU Intel Xeon W-3265
CPU Cooler Noctua NH-U12S DX-3647
Motherboard Asus PRO WS C621-64L SAGE/10G
RAM 12x DDR4-2933 ECC Reg 16GB (196GB total)
Video Card 4 x PNY NVIDIA GeForce RTX 2080 Ti 11GB
Hard Drive Samsung 970 Pro 1TB
Software Windows 10 Pro 64-bit (version 1903)
V-Ray Next Benchmark 4.10.06

In the past we have observed that V-Ray performs differently when run on the primary video card - the one handling display output in Windows - versus a secondary card with no monitor attached. To account for that we tested one, two, and three GPUs with the primary included as well as secondary cards only. For four GPUs we had no choice but to include the primary card in the test. The benchmark was run multiple times on each combination of video cards, and the best result for each combination was used in the charts below.

Benchmark Results

Without further ado, here are the results. We color-coded the graph as well: when the primary video card (with the monitor connected) was included, we used light green - and when only secondary cards were used, a darker shade of green was selected instead.

V-Ray Next 4.10.06 Multi-GPU Performance with 1 to 4 NVIDIA GeForce RTX 2080 Ti Video Cards

And here are the same data points, but shown as relative performance to a single RTX 2080 Ti running as the primary video card:

V-Ray Next 4.10.06 Multi-GPU Relative Performance with 1 to 4 NVIDIA GeForce RTX 2080 Ti Video Cards Compared to a Single 2080 Ti

Analysis

There are two main things I see as take-aways from these results:

  1. V-Ray Next GPU scales very well with additional video cards. Looking just at the results of secondary cards, it is almost perfect scaling from 1 -> 2 - > 3 GPUs. This mirrors what we have see in some other GPU rendering engines, like OctaneRender.
  2. There is a slight penalty from rendering with the primary video card - the one handling output to the monitor(s) in Windows. It isn't huge, but you do lose a few percent off the potential that is there.

Conclusion

Performance in V-Ray Next scales very well across multiple GPUs - but that statement could potentially lead to incorrect conclusions. Doubling the number of video cards in a system effectively doubles rendering performance, but does *not* double the price of the computer. Much of a workstation can stay the same, even as more video cards are added, so you can end up with a bigger percentage increase in performance than the percentage increase in overall system price!

In addition, V-Ray Next GPU can also make some use of the CPU in your system by emulating CUDA on it in software. That isn't as big of a boost as adding another real video card, but with higher-end processors it can certainly add a noticeable amount of performance.

Combining this information with our data on what video cards perform the best in V-Ray will enable you to make the right decisions when it comes to selecting your next rendering workstation. And if you are in the market for one, we would be happy to build it for you!

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 the most effective and reliable system possible so you can concentrate on your work and not worry about your computer.

Configure a System!

Tags: Multi, GPU, Scaling, Rendering, Benchmark, NVIDIA, GeForce, Performance, Intel, Xeon W, Video, Card, RTX 2080 Ti, RTX, V-Ray
Turing

https://www.nvidia.com/en-u...

The article doesn't state what driver version was used, can you please test with the latest Nvidia drivers?

Posted on 2019-08-21 07:56:21

I used the latest NVIDIA Studio Drivers, given their focus on creative professional applications: 431.70

I can't imagine new game-focused drivers would alter this performance substantially.

Posted on 2019-08-21 08:02:16

I should clarify: I can't imagine how new drivers would alter the *scaling* we looked at here. Certainly new drivers could improve (or occasionally reduce) overall performance, though in this case the driver update appears to be focused on game performance rather that CUDA - which is what V-Ray Next GPU uses.

And the testbed system I used has moved on from Labs to our product qualification department, so I don't have access to it at the moment anyhow.

Posted on 2019-08-21 16:30:11