by Rob Williams on February 7, 2020 in Processors
AMD’s newest Ryzen Threadripper processor is an absolute beast, and because of that, it’s not for everyone. If you’ve got an insatiable need for more cores than even the 24- or 32-core Threadrippers can offer, then the 64-core 3990X may be right up your alley. Let’s see how it fares across our usual range of workstation workloads.
There are few things we find quite as satisfying as rendering: seeing a bunch of assets thrown into a viewport that turn into a beautiful scene. Rendering also happens to be one of the best possible examples of what can take advantage of as much PC hardware as you can throw at it. This is true both for CPUs and GPUs.
On this page and next, we’re tackling many different renderers, because not all renderers behave the same way. That will be proven in a few cases. If you don’t see a renderer that applies to you, it could to some degree in the future, should you decide to make a move to a different design suite or renderer. An example: V-Ray supports more than just 3ds Max; it also supports Cinema 4D, Maya, Rhino, SketchUp, and Houdini.
Autodesk Arnold
It just takes a little rendering to understand the benefits of such a massive processor. In Autodesk’s Arnold, the renderer has no issue using more than 64 threads, nor should it have issue using even more than that. With both of our renders, the 3990X manages to soar ahead of the rest of the pack.
Blender – CPU
In Blender, we’re not seeing quite as stark gains as we did in Arnold, but it’s not far off. The BMW render in CPU is being handled as fast as a high-end GPU could render it, and saying that, we’d be remiss to say that we’d suggest a CPU like this for Blender unless CPU only is your target. A GPU is going to more often than not deliver better performance for the cost over this CPU, and unlike a solution like KeyShot, you can’t run two instances of Blender at the same time, one with CPU and one with GPU, and continually render in one instance while editing in the other (unless they are different projects, of course).
That said, this bigger CPU could see benefits elsewhere, such as with animation, but we’ve not yet explored that too deeply (yet). From our own testing, a lot of Blender’s physics and baking calculations are at most, two or four threads, which limits scaling; however, we have not looked too deep into compositing. If you use Blender in such a way that the CPU performance is key even though you have a GPU installed, please let us know how you’re using the software!
Blender – CPU + GPU
When you add a GPU into the rendering mix, the 3990X still manages to place on top. The ultimate gains from introducing the GPU isn’t extreme here, but it’s still notable, dropping the Classroom render from 74 to 60 seconds. What those results ultimately prove, though, is that a bigger GPU is still weighed heavier than a bigger CPU, at least in our particular tests.
KeyShot
KeyShot was made for big CPUs, especially since until just a few months ago, the software had revolved entirely around CPU rendering. As we discovered in our dedicated look at KeyShot 9, we found that among our five tested projects, all of them scale virtually the same. There is some slight variation in the charts above, but overall, we think the 3990X is strutting its stuff pretty well. It’s really easy to talk about a product when it pegs the top of a chart!
Chaos Group V-Ray Next – CPU
Interestingly, Chaos Group’s V-Ray becomes one of the rarer renderers to not take proper advantage of the 3990X, at least in all tests. In the Flowers render, gains were seen, but not to the extent that we saw in the other tests. In the more complex Teaset test, the 3990X performed the same as the 3970X. While a V-Ray plugin one version newer is available, we don’t anticipate it improves anything here, based on its patch notes.
Feb 7 Addendum: As it turns out, our projects are a little old, which impacts the scaling a bit. We received an updated Flowers project from Chaos Group, which we quickly tested on the 10980XE and 3990X. We saw better scaling with it than we did with the projects above (257 seconds for 10980XE, and 129 seconds for 3990X). We’ll update the project for our next round of testing.
Let’s see how things fare once a GPU is added in:
Chaos Group V-Ray Next – CPU + GPU
This feels like an odd result, but it’s not unlike what we’ve seen before our of V-Ray. With heterogeneous rendering, the 3990X actually manages to work quite a bit better on the render, and ultimately, the gains seen in CPU+GPU are more impressive than most of the other CPU+GPU we’ve looked at, but it helps when the CPU-only render isn’t scaling as we’d expect.
V-Ray Next Benchmark
What’s intriguing with the result of the standalone V-Ray Bench is that it disagrees with our real-world testing. If we went by this benchmark alone, we’d expect mammoth improvements in render times, but that’s not actually the case, and a reminder as to why real-world benchmarks are important. This is far from the last time we’ll be testing the 3990X, so we’ll continue to monitor software upgrades as they come out, and see what might change.