Announcement

Collapse
No announcement yet.

AMD Fusion Linux Gallium3D Performance Has Improved A Lot

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

  • #11
    Originally posted by monraaf View Post
    Why is openarena 0.8.8 so much slower than 0.8.5 (even at a lower resolution)?
    It's probably fair to reword the question as "why is openarena 0.8.5 so much faster than everything else ?". IIRC 0.8.5 doesn't use shaders while 0.8.8 does (as do all the other games).

    The entry level GPUs use relatively more silicon for pushing screens around than for running shaders, which matches how they are typically used, but the downside is that the parts become shader-bound more quickly than larger GPUs (which use most of their extra silicon in the shader core).

    I think the Bobcat GPU is comparable to a 5450, but with slower clocks and slower memory, so faster than an RS780/880 or RV610/620 but not much else. That still made it exceptionally fast for an integrated GPU at the time, but it's not exactly a gaming part (at least not for anything with complex shaders).
    Last edited by bridgman; 05 August 2013, 07:15 PM.
    Test signature

    Comment


    • #12
      Originally posted by monraaf View Post
      Why is openarena 0.8.8 so much slower than 0.8.5 (even at a lower resolution)?
      It does many shader-intensive effects, like flares, bloom, etc.

      If you turn all that eye-candy off, it should run about the same as 0.8.5.

      Comment


      • #13
        40% OT: Please, can anybody look into suspension/resume with E-series APUs and DPM? My E-450 with Linux 3.11 rc4 doesn't suspend / resume reliably, that with the latest firmware and the latest Fedora packages. I've crawled through my logs and I found nothing, so, please, help.

        Comment


        • #14
          Originally posted by Alejandro Nova View Post
          40% OT: Please, can anybody look into suspension/resume with E-series APUs and DPM? My E-450 with Linux 3.11 rc4 doesn't suspend / resume reliably, that with the latest firmware and the latest Fedora packages. I've crawled through my logs and I found nothing, so, please, help.
          resume/suspend fails how exactly? can you take a screen/photo of the wake up? motherboard? latest bios? dmidecode?

          i mean too few information to help you, get some of the above and will be better to help you

          Comment


          • #15
            Originally posted by Alejandro Nova View Post
            40% OT: Please, can anybody look into suspension/resume with E-series APUs and DPM? My E-450 with Linux 3.11 rc4 doesn't suspend / resume reliably, that with the latest firmware and the latest Fedora packages. I've crawled through my logs and I found nothing, so, please, help.
            Please file a bug (https://bugs.freedesktop.org). Also, are you only seeing issues with dpm enabled or in general? There are some issues with UVD and suspend/resume that have patches pending what may fix the issues you are seeing. Please provide the details in the bug report.

            Comment


            • #16
              Originally posted by agd5f View Post
              Please file a bug (https://bugs.freedesktop.org). Also, are you only seeing issues with dpm enabled or in general? There are some issues with UVD and suspend/resume that have patches pending what may fix the issues you are seeing. Please provide the details in the bug report.
              I don't know how to properly debug my suspend / resume process. If you can point me to a source or wiki where I can RTFM and document myself about logging / debugging suspension / resuming, you'll have those reports. My issues are with dpm enabled, only.

              Comment


              • #17
                I noticed how SB Vadims shader optimizer makes everything run not only faster, but also more power efficient.
                Just look - SB is dark red, original compiler is orange. SB runs close to catalyst, near half the power consumption!

                Comment


                • #18
                  It's my E-350!!!

                  Thanks for benchmarking my APU Michael! :-)

                  A question:
                  - How to correctly interpret the total frame time graphics? I already read in phoronix people saying the important is lower peaks, even with less time.

                  Comment


                  • #19
                    Originally posted by brosis View Post
                    I noticed how SB Vadims shader optimizer makes everything run not only faster, but also more power efficient.
                    Just look - SB is dark red, original compiler is orange. SB runs close to catalyst, near half the power consumption!
                    Uh? SB is dark red, no SB is grey. Lower is better. And there is no power consumption anywhere.

                    Comment


                    • #20
                      Originally posted by Alejandro Nova View Post
                      I don't know how to properly debug my suspend / resume process. If you can point me to a source or wiki where I can RTFM and document myself about logging / debugging suspension / resuming, you'll have those reports. My issues are with dpm enabled, only.
                      @agd5f: After updating firmware from kernel master git trees, suspending and resuming seems to work. Unless I get more errors, I think I'm going to file a Fedora bug report instead of a fd.o one; Fedora must ship all relevant firmware.

                      Comment

                      Working...
                      X