Vulkan results show a strong CPU bias

Thomas's Avatar

Thomas

04 Sep, 2019 07:55 PM

I realize OpenCL, CUDA and Vulkan aren't exactly operating at the same abstraction levels nor with the same quality of software support.

But I was quite shocked to see my GTX 1080ti on a 4.8GHz-Turbo 4-core i7-7700k annihilate an RTX 2080ti on a 3.8GHz-Turbo 18-core E5-2696 v3, when on CUDA and OpenCL the CPUs were generally bored I/O slaves and the RTX 2080ti delivered the relative performance improvements the increased silicon budget promised.

Just have a look at the GB5 compute benchmarks at http://browser.geekbench.com/user/abufrejoval and ask yourself about the validity of RTX 2080ti and GTX 1080ti obtaining 57094/69590 on Vulkan and a more realistic GPGPU result of 161609/53054 on CUDA.

Sure there is a bit of a difference between the monocore compute power of those two hosts 710/1218, but such a heavy CPU dependence seems to invalidate a Vulkan compute benchmark.

  1. Support Staff 1 Posted by John on 04 Sep, 2019 08:40 PM

    John's Avatar

    Interesting! These results are certainly unexpected and not something we encountered in our testing. We'll see if we can reproduce these results in our lab.

Reply to this discussion

Internal reply

Formatting help / Preview (switch to plain text) No formatting (switch to Markdown)

Attaching KB article:

»

Attached Files

You can attach files up to 10MB

If you don't have an account yet, we need to confirm you're human and not a machine trying to post spam.

Keyboard shortcuts

Generic

? Show this help
ESC Blurs the current field

Comment Form

r Focus the comment reply box
^ + ↩ Submit the comment

You can use Command ⌘ instead of Control ^ on Mac