glxgears benchmark…?

I think it’s funny how often you see the phrase “…I know it’s not a benchmark, but I get XX fps with glxgears…”. If you actually did understand that it’s not a benchmark then you wouldn’t use it as one.  *sigh*

Isn’t it about time someone threw out that fps printf in glxgears?

About these ads

17 Responses to “glxgears benchmark…?”

  1. ZeD Says:

    Can you then point a *valid* benchmark?

  2. Jonas Lihnell Says:

    Actually, it’s a great way to see if you’ve got glx up and running properly. There are a few distros that patch it not to show the fps though, so on those I use glxinfo, but even if it says yes, sometimes it doesn’t work, drivers and stuff.. oh well, just my 2 cents.

    • Harald Says:

      I’m sure it has its uses for testing glx drivers etc. But what’s catching my attention is how people contradict themselves constantly by saying “i know it’s not a benchmark” and go on to compare frame rates all over the place.

  3. Diederik van der Boor Says:

    > Isn’t it about time someone threw out that fps printf in glxgears?

    I’d disagree on that one. It still gives an indication whether things are working or not. So a variation between 800 vs 840 doesn’t say much. But when the fps gets down to 100 you know something is really wrong. I still use glxgears for that purpose.

    • kris Says:

      “But when the fps gets down to 100 you know something is really wrong.”

      Not really. If you have enabled VSync with opengl the fps will be capped to under that. For example when i turn that option on in the nvidia settings tool i get around 72 fps whereas before i get 3200 fps…

      If you have different screen resolution, different DPI, size of glxgears window, etc etc.. everything changes aswell. Do you now agree to what the blogger was saying?

  4. JavierBere Says:

    Glxgears proved to me that opengl was broken on the nvidia driver i was using while glxinfo said it should work. i was getting approximately 5 fps whil one screen and regular hundreds of fps wihle of screen rendering

  5. JS Says:

    When I was a quake player I would use the timedemos, but this is overkill.

    As already said, direct rendering may be on but for whatever reason performance may be sub-par. glxgears is not the tool to assess this, but it is a fairly simple tool that gives a number after a couple of seconds. Maybe we need a tool similar to the windows performance index for KDE ;)

  6. Giovanni Says:

    Use Phoronix Test Suite!

  7. KLM Says:

    The big problem with GLX gears is that softwarerendering can go above 60 fps. Most screens is 60 fps, some goes to 70 and 100, but all is in the 2-3 digit ballpark.

    So no card actually have to render faster than this, to be perfectly usefull for almost any application. The usefullness of the graphic card is not the ‘fps’ but how much that can be drawn in 1/60 of a second.

    So a useful application would be one that tries to draw a little more than glxgears, and using more features like textures, transparency, so it will be maxing a cpu in the 1 digit fps space, but if you have any capable graphic controller it goes to 60 (capped).

    If you want to have a ‘comparable’ number like GLX gears, the program could raise the complexity of the scene it draws until it exactly reaches 60fps.
    I known that people will say this is just another artificial benchmark, but at least it wont measure anything that is guaranteed useless like the fps, but something that could have an impact in other programs.

  8. Diederik van der Boor Says:

    @kris: your comment and the others do make it clear to me we do need a different benchmarking tool…

  9. me Says:

    @ZeD: GLOBS

  10. Top Posts « Says:

    [...] glxgears benchmark…? I think it’s funny how often you see the phrase “…I know it’s not a benchmark, but I get XX fps [...] [...]

  11. HX Says:

    No don’t remove the fps. More info = good.

    If someone sleeps better due to eleventy billion fps in glxgears then thats great. They deserve a pat on the back.

    If they understand that its not a benchmark and still use it as such…then the problem is the person, not the fact that the fps is being displayed.

  12. S Says:

    “Can you then point a *valid* benchmark?”

    Of course!

  13. Jameson Williams Says:

    Of course it’s a benchmark. Benchmarks don’t need to be highly correlated with realworld application performance, but obviously 500 on glx vs. 6000 will be indicative of better graphics performance. Read up on “benchmarks!”

    “There are lies, there are damn lies, and then there are benchmarks.”

    glxgears is important.

  14. Mark Says:

    OK!, I”ve been confussed, concerned, and distressed over the glxgears output for years moving between differant computers and linux versions.
    So what is a simple install and run “benchmark” program to use?

  15. Says:

    “glxgears benchmark? Harald’s hacks” ended up being a really excellent posting, . Keep writing and I’ll keep
    on browsing! Many thanks -Susie

Comments are closed.


Get every new post delivered to your Inbox.

%d bloggers like this: