Announcement

Collapse
No announcement yet.

This Week's Mesa 17.1-dev + Linux 4.11 Radeon Performance vs. NVIDIA

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

  • #11
    Originally posted by Max Payne View Post
    Nvidia still has a huge advantage on csgo at 1920x1080, this should mean mesa drivers have a big cpu overhead when compared to the closed ones.
    Thanks for providing these lower resolution benches instead of just 4K.
    I'm no expert but on a i7-6700k I only experience 30-40% CPU load when playing CS:GO @ 1080p on AMD RX 480.

    I wouldn't call 30-40% CPU load for everything running "big cpu overhead".

    ----

    I wanted to point out 2 ways of looking at these benchmarks.

    #1 - Which card provides maximum performance per $$$$

    #2 - What is the minimum FPS. Is it equal to or above common refresh rates, eg: 60, 74, 120, etc...

    I think the RX 480 is comparable to the 1060 in most areas and is a similar price, I think the kicker is that the open driver isn't even done maturing and the open drivers bring many advantages in stability over the closed drivers. Not to mention buying AMD encourages more open source behavior in the future & discourages closed source drivers to big brands.

    I think that the RX 480 in most cases provides at minimum adequate performance at FHD in most games - this is usually about 60 FPS minimum in my book - 120 FPS prefered (I have a 120hz monitor)

    Nvidia clearly still is and may be for a while superior in performance, but then we're also comparing apples and oranges, a RX 480 will retail $200-300 and is being compared against a $600 GTX 1080, technically in terms of dollars you would have CrossFire RX 480 if you were gonna compare against that price poin.

    Very interesting benchmarks today, great work @mesa-devs, keep the momentum - vega should be really exciting too!

    Comment


    • #12
      Only the load on the thread with the highest usage matters. So when one thread hits 100% the GPU is suppressed and can only process data the same speed this core runs.
      In CS:GO with 1080p resolution and an RX 480 the GPU is definitely limited by the CPU - no matter if 2 cores have zero utilization and another one is doing it's tasks slowly in with the main worker.

      I'm sure the comparison to Nvidias closed source drivers aren't to degrade AMD GPUs but to keep up the improvements, not being satisfied until the last optimization that has a significant performance impact has landed in git.

      For sure the RX 480 is clearly superior over the GTX 1060 on the Linux side just because it can use MESA without hurdles and this is what we like to use on the long term. No one wants to mess around with closed source drivers...
      But as soon as you are too positive about a state you might lose the compulsion to improve it even more. What doesn't kill you makes you stronger so it's the right dose of criticism one has to apply to get the best results in the end.
      And of course the right dose of gratitude - I'm again very happy how the development of MESA moves on.
      Last edited by oooverclocker; 22 March 2017, 05:38 PM.

      Comment


      • #13
        Originally posted by theriddick View Post
        Tomb Raider, seems that's the best game for AMD users under Linux atm
        on normal settings. i heard hair destroys performance

        Comment


        • #14
          Originally posted by oooverclocker View Post
          I'm sure this is perhaps the main reason why Nvidia just can't open source their drivers
          neither amd nor intel open sourced their drivers. they wrote new open source drivers from scratch. while "fuck you" nvidia does not even release hardware docs
          Last edited by pal666; 22 March 2017, 06:03 PM.

          Comment


          • #15
            Originally posted by valici View Post
            lol.. the issue is 215fps vs 280 fps... 1st world problem especially on 60hz monitor.
            Its really just used as a measure to see how the card/driver performs more then realistically playing at that (to the CPU's limit). But yes in this case 1440p and 4k would be better comparisons, however for titles like Civ5 and DeusEx, the 1080p results are important because their so damn low its hardly worth testing higher resolutions.

            Comment


            • #16
              Originally posted by ElectricPrism View Post

              I'm no expert but on a i7-6700k I only experience 30-40% CPU load when playing CS:GO @ 1080p on AMD RX 480.

              I wouldn't call 30-40% CPU load for everything running "big cpu overhead".

              ----

              I wanted to point out 2 ways of looking at these benchmarks.

              #1 - Which card provides maximum performance per $$$$

              #2 - What is the minimum FPS. Is it equal to or above common refresh rates, eg: 60, 74, 120, etc...

              I think the RX 480 is comparable to the 1060 in most areas and is a similar price, I think the kicker is that the open driver isn't even done maturing and the open drivers bring many advantages in stability over the closed drivers. Not to mention buying AMD encourages more open source behavior in the future & discourages closed source drivers to big brands.

              I think that the RX 480 in most cases provides at minimum adequate performance at FHD in most games - this is usually about 60 FPS minimum in my book - 120 FPS prefered (I have a 120hz monitor)

              Nvidia clearly still is and may be for a while superior in performance, but then we're also comparing apples and oranges, a RX 480 will retail $200-300 and is being compared against a $600 GTX 1080, technically in terms of dollars you would have CrossFire RX 480 if you were gonna compare against that price poin.

              Very interesting benchmarks today, great work @mesa-devs, keep the momentum - vega should be really exciting too!
              How many cores? If a driver can take advantage of 1 core only you are into an overhead situation even you have only 1/4 core used, 25% total usage.

              Comment


              • #17
                Originally posted by coder111 View Post
                It looks like RX480 is competitive with 1060, more or less. Ok, maybe several fewer FPSs in some games, equal or more in others, and drivers for AMD will get better with time.

                Overall, it looks like a good deal. Price wise pcpart picker lists RX480 for 200 GBP vs 1060 for 215 GBP. And you get open-source drivers.
                Being performance-competitive with the 1060 is actually an extremely good sign. On Windows, the 1060 and the 480 are often neck-and-neck. If they're that way on Linux, or if they will be that way on Linux soon, it heavily implies that the Linux drivers for AMD are finally getting to where they should be. This is very good news.

                EDIT: The lower FPS in CPU-bound 1080p testing implies that there's still some CPU overhead in the Mesa drivers, but hopefully that will be reasonable to fix as time goes by.

                Comment


                • #18
                  RadeonSI is running very well these days with Tomb Raider and is a nice showcase of the potential open-source AMD Linux gaming performance.
                  Potentional? I remember results of Tomb Raider with Ryzen, where already give crazy fps there and i7700K come very short

                  http://www.phoronix.com/scan.php?pag...ti-ryzen&num=3



                  To me Tomb Raider is good example how game CPU optimizations goes into opposite direction, what is fast on one CPU/GPU became slower on the other I don't think this in mesa being OK but something very unusual happen with that game depending on a CPU+which GPU driver used.

                  He, he, use Ryzen and Tomb Raider to test Linux vs Windows performance... that is good for presentations, like back in year 2012. with LFD2 and nvidia when Valve showed to us how Linux is faster than Windows
                  Last edited by dungeon; 22 March 2017, 09:11 PM.

                  Comment


                  • #19
                    RX480 has equal performance to 1070 with Tomb Raider. Interesting. Wish you tested on High, though.

                    Comment


                    • #20
                      Originally posted by prazola View Post

                      But it's one of the problem if the windows client runs better than the native one...or am I wrong? I don't think any sort of translation layer can boost performances.
                      Nvidia goes better because the driver is better at multithreading.
                      Sure you are right, in that case the translation layer impacts performance, ports like metro lastlight redux which are well done demonstrate that a good quality port can achieve even better performance on linux than windows. Yet, one can not deny that mesa needs a lot of love.

                      Comment

                      Working...
                      X