Announcement

Collapse
No announcement yet.

AMDGPU/RadeonSI Linux 4.10 + Mesa 17.1-dev vs. NVIDIA 378.09 Performance

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

  • #11
    Originally posted by Pepec9124 View Post

    I'd rather have high performance. And as someone else mentioned, deblob your drivers and you'll see how open your stuff really is.
    I'm not aware of anything outside firmware and a decent OpenCL implementation being closed if you wish to sport an RX 480 with the fully open stack, is that correct?
    Nvidia encrypts their firmware and have Nuveau developers wait for them forever. No question they are the king of the hill when it comes down to performance, yet AMD and other related parties actively contribute to the development of the open stack. That's something... Meaningful, if you care about open source. Nvidia, on the other hand, limits its open contributions to the tegra drivers only, if I recall correctly.

    It's ok, I'm still torn between buying an RX480 or a GTX1060, and I'm scared by all the potential headaches that going with AMDGPU will give me eventally, but I'd rather support the company that's friendlier to open source, even if it's tainted by a few blobs here and there.

    Comment


    • #12
      Originally posted by Mike Frett View Post

      ...Unless you support Open Source =p
      Yup, supporting open source is also a choice. But mine remains to buy a 3D accelerator for 3D acceleration.
      I've looked at the RX480, it would have been a worthy alternative on Windows. But on Linux... I'd hate to print a matrix of features, kernels and X servers to determine which driver I need for which scenario.

      Comment


      • #13
        Originally posted by efikkan View Post
        Which still rely on proprietary firmware…
        Nvidia has much better support for open standards, which actually count.
        I'm sorry, but this is misinformation, nvidia supports free standards the least. Nvidia dynamic refresh rate = proprietary, AMD = free, Nvidia technologies (Physix etc.) = proprietary, AMD (TressFX, Mantle etc.) = free.

        Please don't spread misinformation , unless your definition of supporting open standards is somewhat different to what it actually is.

        Comment


        • #14
          I am not willing to buy an expensive graphics card that performs very badly, just because the drivers are open source with the HOPE this might change in the future. – Nope, sorry, AMD. I'll buy NVIDIA in the next few months. Their drivers are feature complete and perform very well.

          Comment


          • #15
            Originally posted by efikkan View Post
            Which still rely on proprietary firmware…
            Nvidia has much better support for open standards, which actually count.
            Nvidia doesn't even have available firmware for their latest cards

            Comment


            • #16
              Originally posted by leipero View Post
              I'm sorry, but this is misinformation, nvidia supports free standards the least. Nvidia dynamic refresh rate = proprietary, AMD = free, Nvidia technologies (Physix etc.) = proprietary, AMD (TressFX, Mantle etc.) = free.

              Please don't spread misinformation , unless your definition of supporting open standards is somewhat different to what it actually is.
              Before accusing others, you should at least get the basics right. Mantle is not free, nor is "AMD".
              Nvidia has the best support for open standards; OpenGL, Vulkan and OpenCL. That what's matters, end of discussion.

              Comment


              • #17
                Originally posted by Steffo View Post
                I am not willing to buy an expensive graphics card that performs very badly, just because the drivers are open source with the HOPE this might change in the future. – Nope, sorry, AMD. I'll buy NVIDIA in the next few months. Their drivers are feature complete and perform very well.
                Here is my subjective experience, so take it as it is, it all depends on definition of "performs very badly". For example, I promised to myself to never buy nvidia product again. The reasons are exactly the same as yours, but our definitions are different. I assume that your definition is playing AAA games or other games that do not require fast reaction time (eg. casual games, strategies, FPS's etc.) and precision, so from that point I can clearly see why would you say such thing that is in my definition completely false, since your aim is highest FPS possible regardless of input lag. My definition is however, very different, I have no interest in playing such games for example, they are not fun to me (quite boring re-hasging over and over again actually, with improved graphics), in games that require fast reaction time and precision, having nvidia GPU (to me) was a nightmare, where every milisecond counts, you just can't do anything how much lag those GPU's (drivers) have, and please don't confuse "lag" with "low FPS or stutter" it's not the same thing, lag is (to others who might confuse it) when button is pressed, the amount of time needed for action to take place in game, it depends on game, but also a lot on drivers and GPU's.

                I don't know what wichcraft happens in nvidia blob to introduce so much lag, but I'm sure that is one of the reasons why their drivers and most of the technologies stay proprietary. However, there is a side effect for those things, among input lag, there is also an effect of support for older hardware, high end GPU, from say 2012 will not preform to expected level because they would not waste resources on those GPU's (since there's new generations availiable), and they will fall back to the real performance of those GPU's (quite lower), and when comparing actual specification, you can clearly see that is ok (AMD GPU's with same specification "preform worse" in terms of FPS), but, due to the driver architecture (again, assumption), it introduce other problems, unsmooth experience with high FPS on non optimized drivers for specific GPU on newer (or less known) games, and complete mess forcing you to upgrade even if you don't have need to do it actually. This is uneaceptable for me, but it makes sense for majority of people, and nvidia plays on that card, and that is smart and very dishonest practise (again in my opinion).

                So there you go, me defending both yours and mine choice in one post .
                Last edited by leipero; 27 January 2017, 12:08 PM.

                Comment


                • #18
                  Originally posted by bug77 View Post
                  So the only AMD card that beats my 1060 is the Fury. And only in two tests. Not a bad buying decision, I reckon.
                  Someone should tell Nvidia that if GTX1080 gives 156fps on Unigine, then the half part GTX1060 should give 78fps and not 100. Nvidia is not to be trusted as always.

                  Comment


                  • #19
                    Originally posted by Steffo View Post
                    I am not willing to buy an expensive graphics card that performs very badly, just because the drivers are open source with the HOPE this might change in the future. – Nope, sorry, AMD. I'll buy NVIDIA in the next few months. Their drivers are feature complete and perform very well.
                    Yep sad but true. As these benchmarks show Nvidia is still much faster and better value for gaming on linux. Both in terms of openGL and Vulkan.

                    I commend AMD for the progress they have made and hope it will continue and start matching Nvidia's performance. Been waiting for that for so many years. Supporting open source is good but we should also hold them to the same performance standards.

                    When it comes to directX performance and Windows drivers it's a different story. AMD is very very good with their drivers on windows. In terms of performance, stability, and features.

                    ps- I look forward to the AMDGPU PRO 16.60 benchmarks as well, especially the Vulkan performance will be much faster than RAD-V which is still slow.
                    Last edited by humbug; 27 January 2017, 12:15 PM.

                    Comment


                    • #20
                      Originally posted by efikkan View Post
                      Before accusing others, you should at least get the basics right. Mantle is not free, nor is "AMD".
                      Nvidia has the best support for open standards; OpenGL, Vulkan and OpenCL. That what's matters, end of discussion.
                      No, it is not end of dicussion, please support your claims, I am not trying to be rude. Ok, Mantle is not open source, but Vulkan is Mantle-derived, so it is partly open source. Again, it depends on your definition of support, what do you mean by support for open standards? Do you mean support for their products in proprietary blob, ora ctual support for open standard, because if you mean about 2nd one, that is simply not the case, and that is misinformation.

                      Comment

                      Working...
                      X