Can we trust the Vray benchmark

[fraggle]

Hello,
while planning to buy new GPU, I check vray benchmarck like everybody else.
But, regarding the GPU, results are meaningless.
Some system with 4 x 1080 Ti range from 14s to 28s, while other with 7 or even 8 x 1080 Ti are 16s to 22s!
Same for all the type of card, the result are very dangling and the reported time range is very huge.
So at the end it's we can't decided based on this benchmark for GPU.

Please note that the CPU benchmark is a lot more consistent and such CPU benchmark make a lot more sense.

[varosi]

Hello!

The new one - V-Ray Next Benchmark produce proportional (linear) results, so you may actually compare how much a hardware perform rendering with V-Ray Next compared to others.

You may download it here: https://download.chaosgroup.com/?platform=47&product=57

[fraggle]

Thanks, but are you starting a new result chart or will it be mix up with the old one?

[daniel.taskoff]

Hello,
the scores from V-Ray Next Benchmark and V-Ray 3 Benchmark are kept separately, mainly because of the different scoring systems used.
By default, https://benchmark.chaosgroup.com shows the tests submitted with the new scoring system, but you can see the old ones, as well, by following the "< V-Ray 3 Benchmarks" link at the bottom of the page.

Best regards,
Daniel

[fraggle]

Thanks, new benchmarks make a huge difference. AMD was a good horse compare to intel in the previous benchmark, now it look like a bad one.
The ranking are not matching the one from cinebench for sure, a 2990WX outperform any intel CPU, while in vraybench the 2990WX is equivalent to a 8176 intel Xeon! What's the rational explanation for such big gap between the 2 results?

[Karol.Osinski]

I would love to know it too, soon we are going to buy a new workstation for the studio and I wonder if we should stick to something we are using atm such as i9 7900x or give a try to 2990wx (I might have mistaken a models slightly).

[fraggle]

The new AMD 3970x is about to be launch, could be wise to wait for vraybenchmark on this new CPU.
Decision should not be done accordingly to cinebench as the result differ to much from the one on Vraybenchmark regarding AMD.

[varosi]

V-Ray Benchmark Next is measuring how V-Ray Next could perform on a given platform, not the general platform performance. And not other ray-tracer's performance.

[fraggle]

I got this from https://www.cgdirector.com/, so clearly each render engine are more optimized for INTEL or for AMD?

[daniel.taskoff]

fraggle, the results from cgdirector which you posted, are from a very old version of V-Ray Benchmark. For more up-to-date results (using V-Ray Next) with better scaling, please consider checking the V-Ray Benchmark Next results on https://benchmark.chaosgroup.com/next/cpu

As for the difference in performance shown from the two benchmarks, keep in mind, that V-Ray Benchmark Next shows how a particular hardware will render a scene with V-Ray Next, while Cinebench does the same with Cinema 4D.

[fraggle]

Sure but still the benchmark result is very inconsistent, even when targeting only vray benchmark. This result goes against everything we see on other vraybenchmark where the 3970x is almost 2 time faster than any intel brand new CPU. The clock of xeon 6254 is 3.1 vs 3.7 for the 3970x and the 3970x got double core than the 6254. So why is the Xeon ahead!!!

[varosi]

Interesting question! We have to get those processors in here to test more thoroughly to see where the difference come from in V-Ray. Keep in mind that we cannot have in-house all the processors on the market, but we try to profile V-Ray for latest processors as more as we can.