Radeon ROCm 2.0 OpenCL Benchmarks With Linux 5.0 On Ubuntu 18.10 vs. NVIDIA's Linux Driver

[phoronix]

Phoronix: Radeon ROCm 2.0 OpenCL Benchmarks With Linux 5.0 On Ubuntu 18.10 vs. NVIDIA's Linux Driver

With yesterday having posted fresh OpenGL/Vulkan Linux gaming benchmarks for the current NVIDIA GeForce and AMD Radeon graphics cards, in this article is the freshest OpenCL GPU compute data for that set of 14 graphics cards on the very latest Linux graphics driver stack. In the case of AMD Radeon open-source compute, it was tested using the new ROCm 2.0 atop the mainline Linux 5.0 kernel and Ubuntu 18.10.

http://www.phoronix.com/vr.php?view=27478

[Brisse]

Really looking forward to those Radeon VII benchmarks.

[ThoreauHD]

So the Vega 64 is on par with 1080 performance. I'd like to see how the Radeon VII and Radeon Instinct MI60 do when they are released.

[AndyChow]

Radeon VII is MI50 made for consumers, right? I would like to see how it performs these tests, since it`s supposed to be a platform for ML/AI. That said, I`m rather skeptical that the results will be impressive.

[ObiWan]

For those curious what a Vega 56 (165 Watt powerlimit) does with the amdgpu pro opencl backend, instead of ROCm

https://openbenchmarking.org/result/1901304-SP-1901305SP07

[s_j_newbury]

No double precision benchmarks?

[smitty3268]

Radeon VII is MI50 made for consumers, right?

Pretty much, yeah. I think they're bumping some clock speeds a small amount and limiting the fp64 rate vs the pro card.

[Happy Heyoka]

For those curious what a Vega 56 (165 Watt powerlimit) does with the amdgpu pro opencl backend, instead of ROCm

https://openbenchmarking.org/result/...SP-1901305SP07

So if I'm reading that right, comparable to the ROCm on almost everything but memory performance?
I wonder what's going on there and how much whatever that is is contributing to the slight discrepancies here and there...?

[cb88]

kernel latency is also way higher on amdocl...

Basically there is mo reason to not use RoCm on amd cards... it also remains to be seen If the rate limited fp64 can be worked around in the open drivers.

[pal666]

so what is wrong with first batch of tests where amd cards perform much worse than on rest of tests?