Announcement

Collapse
No announcement yet.

Interesting interaction between AMD Radeon Anti-Lag and GPU D3D9 scores

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Interesting interaction between AMD Radeon Anti-Lag and GPU D3D9 scores


    I bought an MSI RX 5700 Evoke OC in April and ran several benchmarks after updating drivers, at the maximum OC allowed by the Radeon software of 1850MHZ and 10% power increase. I did not mess around with the VRAM clock.

    These were the results I obtained at the time:
    DX9: 212
    DX10: 196
    DX11: 285
    DX12: 69
    Compute: 7522
    Total: 16,950

    Recently I reinstalled Windows due to other issues, and got the following:
    DX9: 204
    DX10: 206
    DX11: 299
    DX12: 70
    Compute: 7236
    Total: 17,008

    My questions are:
    1. As suggested in the subject, I have noticed that if I turn on Radeon Anti-Lag in Radeon settings, my DX9 score gets into the 250's but doesn't affect the other tests.
      I don't suppose the gain is real, is it?
      Could it be that enabling it gets around another unidentified issues?
    2. In my original April test, I got that 7522 Compute score rather casually, as I don't recall running the test multiple time. However, now after rerunning the test several times, the most I have been able to score is 7336.
      Could a change in the Radeon drivers since April have affected GPU Compute performance? Could my card have developed an issue since then?
    3. Finally, I have noticed that the scores for Navi cards in the High End Chart here are somewhat deflated.
      The RX 5700 average is reported as 14,517 and even the RX 5700 XT shows only 16,730.
      My card at the meager OC clock speed of 1850MHz achieves ~17,000, which is higher than the RX 5700 XT score on the chart.
      Even at stock settings, my card scores in the 16,100 - 16,200 range.
      I realize that my card is factory overclocked over the reference, even at its stock settings, but so are most 3rd party vendor cards, and I imagine the results people submit are primarily from non-reference cards, since that is what sells the most.
      Could it be that many people penalize their D3D12 scores by not running the test while having a 4K-capable display connected to their PC?
    I apologize in advance for the long read and will appreciate any information that provides answers.

  • #2
    1) We don't know what Radeon Anti-Lag does at a deep technical level. So don't really know the answer. I did find a comment on the AMD web site however that it only works on DX9 and DX11.

    2) Yes, driver changes regularly come out with changes that impact the benchmarks. But it could be other factors, like it is summer now (for us anyway) and ambient temperatures are higher, so thermal throttling could occur sooner.

    3) Yes. Having no 4K would hurt a little. But I think the bigger problem is that there are a lot of systems setup in a sub-optimal fashion. e.g. GPU in the wrong slot, slow CPUs but high end GPU, poor cooling, too much bloatware software running, etc.. Anyone that take s bit of care with optimisation will probably get above average.


    Comment


    • #3
      Thank you for your answers. It all sounds perfectly reasonable.

      In my case I think I can rule out thermal throttling. I have a PC with a big case, with custom fan profiles calling for 100% operation at 80C, and on a cool basement (winter here in Northern US). When I monitor temperatures during the Compute test with HWInfo, my hot spot temperature hits a maximum of 96 or less, and other metrics are under 70C. However, I see your point that many factors external to the GPU could be at play.

      Best regards.

      Comment

      Working...
      X