Announcement

Collapse
No announcement yet.

AMDGPU-PRO 17.30 vs. Linux 4.13 + Mesa Git RadeonSI Benchmarks

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

  • #21
    The Superposition results were interesting. Haven't previous tests always placed radeonsi behind the pro driver in that benchmark?

    I wonder if it's due to the new 4.13 kernel?

    Comment


    • #22
      Originally posted by smitty3268 View Post
      The Superposition results were interesting. Haven't previous tests always placed radeonsi behind the pro driver in that benchmark?

      I wonder if it's due to the new 4.13 kernel?
      or LLVM 5.0.

      Comment


      • #23
        These 375 MB drivers without GUI seems broken

        Comment


        • #24
          I am pretty sure, that Michael will update the tests with NVIDIA soon, but so far, OMG, compare this results to the last NVIDIA runs! At least by performance AMD is absolutely competitive now and even wins in some tests clearly. If they finally get audio through HDMI running, we have a clear winner for Linux desktop systems hands down. Buy a card plug it into your PC, turn on and play. A dream comes true! Thank you guys for all your effort, you did a great job.

          Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

          Comment


          • #25
            Originally posted by Qaridarium
            yes very impressive....

            and sad that this zombie driver amdgpu-pro still do have relevance for some people.
            we just need a open source implementation of the "OpenGL Compatibility Profile" stuff and then we can shut down amdgpu-pro forever even for the workstation.
            Workstation things are a bit more delicate than games. AMD must provide always the same rendering output between drivers with the same input, so there are some restrictions which can be relaxed in open drivers which are "not guaranteed" to work the same way with each release.

            Comment


            • #26
              Originally posted by debianxfce View Post
              I wonder how many of the readers do have same cpu and motherboard. More real would test would be used under 100 usd mobo and under 150 usd cpu.
              And when I do those tests unless it's explicitly for a low/budget CPU comparison, readers complain "but the CPU is the bottleneck!", etc
              Michael Larabel
              https://www.michaellarabel.com/

              Comment


              • #27
                Originally posted by Qaridarium

                are you open for a upgrade from the AMD RX-Vega-56 to the full version Vega64(4096 shader) if you get the money for the upgrade?
                Still trying to find out from AMD clearly if I will be receiving any review sample or not, but do hope to get a 64 but regardless any tips are appreciated and if receiving anything from you will be sure to get a 64 if AMD does not send one out.
                Michael Larabel
                https://www.michaellarabel.com/

                Comment


                • #28
                  Originally posted by Qaridarium

                  right now a AMD-RX-VEGA-64 will make ~50$ profit a month. if you have your own power plant(Solar/Wind) even more.
                  How do you know how fast the card is? Is that based on the Frontier edition?

                  Comment


                  • #29
                    Originally posted by oleid View Post
                    70% of the performance of the proprietary driver we will get, they said.
                    What we said was that without a real shader compiler we expected somewhere around 60-70% of proprietary driver performance.

                    We started working on a "real" shader compiler back in 2011, when we hired Tom S.
                    Test signature

                    Comment


                    • #30
                      Pretty sure the GTX 1080 is the mining favorite atm because it has low TDP and high mining performance, pretty sure it beats the Vega cards pretty easily in that department for cost/energy/hashing math. The RX480/580 would probably be better if their prices didn't explode to 1080 levels.

                      The other thing about the AMD RX Vega release is its a forced bundle system, meaning you get bundled to buy ryzen and freesync which is going to REALLY suck for people like me who already have those things........... A big FU from AMD to pre-existing customers.....

                      Comment

                      Working...
                      X