Announcement

Collapse
No announcement yet.

NVIDIA GeForce vs. AMD Radeon Linux Gaming Performance For August 2023

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

  • NVIDIA GeForce vs. AMD Radeon Linux Gaming Performance For August 2023

    Phoronix: NVIDIA GeForce vs. AMD Radeon Linux Gaming Performance For August 2023

    After resorting to buying a NVIDIA GeForce RTX 4060 graphics card to be able to share Linux performance metrics for that more affordable Ada GPU, last month I posted the Radeon RX 7600 vs. GeForce RTX 4060 benchmarks as well as looking at the GeForce GTX 1060 through RTX 4060 GPU compute and renderer performance across several generations of NVIDIA GPUs. For those considering the RTX 4060 for a Linux gaming system or an upgrade to other recently released AMD or NVIDIA GPUs, here is a fresh round of Linux gaming performance metrics on the newest drivers.

    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

  • #2
    Nice. Geometric mean image won't load for me though.

    Comment


    • #3
      Originally posted by phoronix View Post
      Performance For August 2023
      I think it's time for browsers test.

      Comment


      • #4
        I gotta say... 6800 beating 7900 xt in a few benchmarks. this is crazy...

        Comment


        • #5
          For some reason, Firefox outperforms Chromium in some CI tests, but Chromium always outperforms Firefox in the real world.

          Comment


          • #6
            NVidia have better RT, DLSS 2 and 3. Encoders give better quality, too.

            But yes, in traditional rendering (with fake lighting and shadows ) Radeon does a bit better for the same price​

            Comment


            • #7
              Originally posted by peterdk View Post
              Nice. Geometric mean image won't load for me though.
              I don't see it in Chromium and Firefox as well

              Comment


              • #8
                Originally posted by HEL88 View Post
                NVidia have better RT, DLSS 2 and 3. Encoders give better quality, too.

                But yes, in traditional rendering (with fake lighting and shadows ) Radeon does a bit better for the same price​
                Don't forget CUDA. RT, DLSS, and CUDA convinced me to get a 4070 in my new laptop (with a 7940hs). Sorry AMD, I couldn't wait for you anymore. Maybe next refresh.

                Comment


                • #9
                  Something's wrong with the bench results for certain. The 7900 xt is 5/6ths of an XTX. Down to the silicon, bus size, and amount of VRAM. It's an XTX with 1/6th turned off.

                  It will always perform anywhere between 17-23% weaker than an XTX in all circumstances. This isn't the case here at all.

                  Comment


                  • #10
                    Originally posted by panikal View Post
                    Don't forget CUDA. RT, DLSS, and CUDA convinced me to get a 4070 in my new laptop (...)
                    for me and my Vega64 ROCm/HIP works on Fedora 38 out of the box with dnf install rocm* tested with Blender 3.5.1

                    the RT performance does in fact only matter for OptiX in Blender because for real games it is always to slow even with a 4090

                    FSR3 also supports upscalling and can be temporal to like DLSS2 and also has Frame-Generation similar to DLSS3 so i see no point in that. also many people complain about ghosting this means FSR1.0 without Temporal component with frame generation is the better option.

                    for AI workloads the memory wall is one of the biggest problems thats why i ordered yesterday a 7000€ system with a AMD ryzen 7950X3D and a AMD Radeon PRO W7900 48GB (this card alone is 4000€) together with 192GB ram and a 8TB SSD. That will be a nice System for Blender and other workstation work also for AI/deep learning workloads on Linux

                    even a 4090 with only 24GB vram will hit the memory wall very fast...

                    Phantom circuit Sequence Reducer Dyslexia

                    Comment

                    Working...
                    X