Redshift 2.6.22: NVIDIA GeForce RTX 2080 & 2080 Ti GPU Rendering Performance
Written on September 28, 2018 by William GeorgeIntroduction
Redshift is a production-quality, GPU-accelerated renderer. Traditionally this type of rendering was done on CPUs, but graphics processors (GPUs) are ideal for highly parallel tasks like this - and it is easier to fit multiple video cards in a single computer, to boost performance, than multiple CPUs.
With NVIDIA's recent release of the GeForce RTX 2080 and 2080 Ti, we wanted to see how these new graphics cards stack up to the well-respected 1080 Ti and the top-end Titan V.
Test Setup
At publishing time, the latest version of Redshift is 2.6.22. This version appears to support the new Turing GPU architecture used in the RTX series cards, though the version we had tested most recently before this (2.6.11) does not. As such we ran all new tests with older cards for this article, using the new version of Redshift and the same graphics driver that NVIDIA released to support the new cards.
For our test platform, we wanted to use a high-end processor so that there is no way it is limiting any of the video cards. Since we are looking at single GPUs (sadly we only have one RTX 2080 Ti at this time) we didn't worry about supporting a maximum number of cards. If you would like full details on the hardware configuration we tested on, just click here to expand a detailed list.
Testing Hardware | |
Motherboard: | Gigabyte X299 Designare EX |
CPU: | Intel Core i9 7980XE 2.6GHz (4.4GHz Turbo) 18 Core |
RAM: | 8x Crucial DDR4-2666 16GB (128GB total) |
GPU: | NVIDIA GeForce GTX 1070 Ti 8GB NVIDIA GeForce GTX 1080 Ti 11GB NVIDIA GeForce RTX 2080 8GB NVIDIA GeForce RTX 2080 Ti 11GB NVIDIA Titan V 12GB |
Hard Drive: | Samsung 960 Pro 1TB M.2 PCI-E x4 NVMe SSD |
OS: | Windows 10 Pro 64-bit |
PSU: | Seasonic X-Series 1050W |
Software: | Redshift 2.6.22 Demo Benchmark (Age of Vultures scene) |
Benchmark Results
Here are the Redshift 2.6.22 benchmark render times each of the video cards we tested:
And here is another way of looking at the results, as percentages relative to the GeForce GTX 1080 Ti's performance:
Analysis
It is a little tricky to interpret raw benchmark times into actionable information, as in the first graph above, but the second chart makes things much clearer. The new GeForce RTX 2080 is slightly faster (8%) than the similarly-priced GTX 1080 Ti from the previous generation, though it has less dedicated video memory (8 vs 11GB) - so for folks working with larger scenes it may make sense to stick with the 1080 Ti in that price range. For those with a little more to spend, the RTX 2080 Ti comes in 38% faster than the GTX 1080 Ti - and not very far behind the much more expensive Titan V, while still having a respectable 11GB of VRAM.
Keep in mind that these are only single-card scores, and doubling (or tripling/quadrupling) up on video cards can yield even more impressive results. However, we have found that the dual-fan coolers used by NVIDIA on these new GeForce RTX cards are problematic when used in multi-GPU configurations. For more information about that, check out our in-depth look at that topic.
Finally, it is worth noting that this version of Redshift is not yet able to utilize the new RT cores which NVIDIA added to this generation of GPUs. Those are purpose-built for raytracing, and if Redshift is updated to use them in the future we could see a huge increase in performance. We will be sure to test these cards again if or when that technology is integrated into this rendering engine.
Are the GeForce RTX 2080 and 2080 Ti Good for Rendering in Redshift?
The GeForce RTX GPUs do very well in Redshift when used as individual cards. They outperform the previous generation at the same price point, and there is potential for even better performance in the future if Redshift is updated to use the new RT cores they have. Avoid using dual-fan cards in multi-GPU configurations, though!
Thanks for the article. What's the best card model to stack 3xgpu?
If you are putting any number of these new 2080 / 2080 Ti cards in a system such that they will be next to each other - whether the total number is 2, 3, or 4 GPUs - you will want to use models that have a blower-style fan. We covered why the dual-fan cards, like NVIDIA's Founders Edition, do not work well in these configurations (https://www.pugetsystems.co... ) and then how blower-style cards perform much better (https://www.pugetsystems.co... ). Both of those are worth checking out if you are considering building or buying a system like this. As for what specific model of GPUs to get, the only ones I have had hands-on time with are the Asus Turbo RTX 2080 (https://www.asus.com/us/Gra... ). I'm sure other manufacturers will put out single-fan cards that are similar, but that should be a safe bet at this time.
So, Redshift doesn't work with versions older than 2.6.22?
Thanks a lot for this article!
We didn't test every version, but Redshift 2.6.11 did *not* work with the new GeForce RTX cards - and version 2.6.22 did. There may have been other variants in-between, as Redshift puts out quite a few incremental versions, but if you want to use RTX cards I would make sure you are on 2.6.22 or later.
Using the RTX Graphic Cards*
I am utterly deflated. I just put >$6k into a 18 core i9-7980xe system with GTX 2080 ti card strictly for rendering from Cinema 4D. When using physical or standard in C4D, it renders over twice as fast as my other system (~$4k) geared for After Effects (i7-8700 w/2080ti). HOWEVER, this new C4D system renders SLOWER than my AE system when using Redshift! Unfortunately, I just learned about RS after I had put the order in on my system, and now that is all I plan on using! FML! lol
May I ask if your new C4D system is purely for rendering, or also for modeling and animation? And are you really going to only use Redshift for rendering, or will you want to also use C4D's built-in render engine sometimes?
I ask because if you truly plan to only use Redshift, and you don't do any other heavily threaded work on the system, you might consider just consolidating in one system by moving your second RTX 2080 Ti over to the i7 8700K rig. That CPU is better for stuff like modeling in C4D than the i9 7980XE, and for Redshift having two 2080 Ti cards in a single system (regardless of the CPU) will be a big performance boost. If you *ever* do anything heavily threaded, though, that 18-core processor is going to be far, far faster than the six-core 8700K. If you think that might be the case, maybe just add another 2080 Ti or two to the 7980XE? That will dramatically help with Redshift render times, if your system has the space, power, and cooling to handle multiple GPUs.