No announcement yet.

AMD and Nvidia Performance Discrepancies

  • Filter
  • Time
  • Show
Clear All
new posts

  • AMD and Nvidia Performance Discrepancies

    Hello Passmark,

    I am currently in the market for a new Graphics card and have tended to come here to try and determine the best bang for my buck performance wise. Usually AMD tends to come ahead in this in the CPU market and sometimes with Graphics cards. I have noticed that the new line of AMD Graphics cards appear to pale in comparison to the Nvidia cards in the same price bracket to the point where your chart indicates they shouldn't really be considered (Nvidia cards end up with a higher performance vs cost). For example looking at the AMD R9 Fury vs the 980ti, your benchmark indicates the 980 is 40% faster for a similar price. However as I look at benchmarks the two cards appear more neck and neck while giving Nvidia the edge usually.

    For example here:

    It does not appear to demonstrate a 40% performance improvement via the majority of the tests that we have seen. That being said I did read that the 300 line is a rebrand of the 200 line and your software has a hard time determining the difference, however I can't imagine the performance between the benchmark program and game testing wouldn't be reflective of eachother.

    Could you please explain what is causing the sizeable discrepency between your benchmark and other benchmarks that we have seen? Is there some kind of weight that you are multiplying performance numbers that may not be reflective of real world tests?

    Thank you,

  • #2
    We also think some of the new AMD cards should be doing slightly better. (not 40% better, but maybe 10 - 15% better). To maybe put the R9 Fury X on par with the 780Ti. The 980Ti is about $90 more expensive at the moment, so I am not sure I would say they are in the same price bracket.

    We have contacted AMD some weeks ago, and they said they would look into it for us, but as yet haven't provided any detailed response.

    What we suspect has happened is that AMD put a lot of effort into optimising their drivers for a small number of games, maybe just a couple of dozen games. And software that didn't get the special treatment doesn't run quite as well.


    • #3
      Hello David,

      I figured I would circle back around and see if you have updated your benchmark tool yet (and or figure out what the problem was)? With the release of the 300 line of cards, Fury cards, and fairly sizeable performance improvement driver side. I believe we are seeing reviews and performance benchmarks (such as from GPU Boss) place the r9 380 on part with the GTX 960, R9 390 with the GTX 970, Fury on par with the GTX 980, and Fury X on par with the GTX 980ti, and of course with the Pro Duo taking the top spot in performance due to its Crossfire nature. Did you ever figure out the specific issue that was placing your GPU benchmarks so much lower then other review sites (and game benchmarks?) (I assume you aren't factoring in DX 12 performance yet either so that is probably a side issue).

      Thank you,


      • #4
        We don't really think there was anything wrong with our tests. We think it was more of a device driver optimisation issue.

        But having said that, PerformanceTest V9 (beta as of 3/May/16) is now out. It contains a DX12 test and drops one of the DX9 tests.

        We were hoping to get better crossfire / SLI performance, but writing good code for multiple CPUs is extremely hard. If anything the situation has got worse over the last couple of years, with even very well resourced AAA games failing to get good scaling.

        We sent the PT9 beta off of AMD, nVidia and Intel for review / comment a month ago. But so far there has been no feedback on the 3D tests.


        • #5
          I think the benchmarks here are good for a quick overall comparison eg to find comparable GPUs.

          But if you want to compare 2 specific GPUs you need to look at reviews and comparisons.

          eg gtx 980 vs r9 nano: - about the same. But can vary greatly with what exact benchmark (eg game/settings/resolution) you use.

          In my case, I chose: - the nano (especially now that lower prices can be found in Australia)