Announcement

Collapse
No announcement yet.

What Shall We Benchmark Next? Let Us Know!

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

  • #81
    Give readers an idea of how Windows benchmarks translate to linux benchmarks.

    For example, suppose a Windows hardware review site said a certain SSD performs poorly on a certain benchmark. Suppose that Phoronix then revealed that it was just NTFS handling an aspect of that SSD badly and that linux users will not have the same problem because ext4 won't trip over this particular SSD's quirk. That's the kind of thing I'm talking about.

    Comment


    • #82
      Something that would be cool would be to compare a number of different compilers on Linux, for example:

      Proprietary but free for personal use:
      * ICC, intel C compiler
      * SunCC, Sun studio

      BSD-type licence
      * LLVM/Clang
      * LCC
      * PCC

      Copyleft-type licence
      *GCC
      *TCC

      Of the ones above, only ICC, TCC and GCC are reported to be able to compile linux itself, whereas Clang can do FreeBSD, SunCC openSolaris and PCC openBSD.
      Basically, it would be interesting to see how the different compilers quantitatively compare which could influence decisions on which compiler to use based on licence or specific performance.

      Comment


      • #83
        opengl compiz/compiz disabled benchmarks

        Hi.

        I would like to see a benchmark showing the difference in fps when compiz is enabled and disabled.

        For me its about 30-40%...

        Linux desktop developers need to make a way to automatically stop compiz when running a fullscreen 3d app

        The Benchmark would be good for linux and may encourage people to disable the feature when running games and advance the speed of all 3D games.

        Every single game that co-exists for linux and windows I have tried on my system is faster in Linux (i always disable compiz)

        ( I use Nvidia 8500)

        Comment


        • #84
          Wow - phoronix is amazing, your sort of like a genie..

          I suggested both Arch and disabled compiz benchmarks and you delivered both!

          Thanks

          Comment


          • #85
            did I remember to suggest network benchmarks? say network transfer benchmarks or NFS performance between kernel versions. Especially sence we've had some changes and virturalized network changes

            Comment


            • #86
              What about test a linux distro heavy optimized by phoronix (with anything removed but only the stuff let hardware to benchmark work?)

              It could start from turbopup xtreme (puppy linux derivate) or linux from scratch.

              It would be good to test hardware or driver update performance "as basis" of quality.

              Comment


              • #87
                Yes, heavily optimized: is it worth it? Too bad the answer will be no...

                Comment


                • #88
                  Originally posted by LinuxID10T View Post
                  Yes, heavily optimized: is it worth it? Too bad the answer will be no...
                  Thanks for sharing your expertise and highly detailed experience... not.

                  Jokes aside, what could be the problem? it is not about make a new distro with repository, package manager and such: but a kernel which load the very essential stuff to make the stuff you need to benchmark; which is usually: video (tv-out to record the quality without influence framerate), audio, storage and (optionally) networking.
                  For example, you don't need to load the modules that manage anything but ext4, samba, cups, plug&play hardware, [huge list cut]

                  Commercial stuff (the mostly one which need benchmarking) usually avoid too many dependeincies.. this is a big advance for expand the number of application you can benchmark on this "linux from scratch" with little number of package to update.


                  A very light distro heavily optimized for make multimedia work on everyPC (boot from usb): it could even be the definitive linux gaming distro (no eyecandy, no compiz, no cups... just reboot and play (make every PC become, vritually, a console)

                  Comment


                  • #89
                    Originally posted by alexan View Post
                    Thanks for sharing your expertise and highly detailed experience... not.

                    Jokes aside, what could be the problem? it is not about make a new distro with repository, package manager and such: but a kernel which load the very essential stuff to make the stuff you need to benchmark; which is usually: video (tv-out to record the quality without influence framerate), audio, storage and (optionally) networking.
                    For example, you don't need to load the modules that manage anything but ext4, samba, cups, plug&play hardware, [huge list cut]

                    Commercial stuff (the mostly one which need benchmarking) usually avoid too many dependeincies.. this is a big advance for expand the number of application you can benchmark on this "linux from scratch" with little number of package to update.


                    A very light distro heavily optimized for make multimedia work on everyPC (boot from usb): it could even be the definitive linux gaming distro (no eyecandy, no compiz, no cups... just reboot and play (make every PC become, vritually, a console)
                    The truth is, time after time, there aren't many speed differences between system that have been so called "highly optimized" and ones that aren't. There is a reason for that. People that work on GCC work hard so every platform performs well. They put a lot of time and effort into making the compiler and executables faster.

                    Comment


                    • #90
                      Originally posted by LinuxID10T View Post
                      The truth is, time after time, there aren't many speed differences between system that have been so called "highly optimized" and ones that aren't. There is a reason for that. People that work on GCC work hard so every platform performs well. They put a lot of time and effort into making the compiler and executables faster.
                      except when they don't. How much faster could be gcc and gcc generated code without the burden of AIX?
                      http://udrepper.livejournal.com/7326.html

                      Comment

                      Working...
                      X