glxgears frame rate benchmark

[Lennart Sorensen]

On Fri, Dec 14, 2007 at 09:29:25AM -0500, ted leslie wrote:
> I have felt my graphics a little pokey on my new gutsy set up.
> I ran glxgears, and got 11000 FPS on the detault window size.
> I then scaled the window (just pulled on right bottom corner to resize),
> and the FPS changes quite a bit, i.e. 
> goes down to 400 FPS on full screen.
> 
> Now given that these GPU's (I'm using nvidia 7900GT),
> render shaded polygons and such (in card),
> I am very very surprized that the FPS would change that
> much over a window size change.
> 
> I be interested in what others get when they run glxgears
> and alter the window size, and what card it is on,
> i.e. ATI, nvidia,
> 
> I am also using the new nvidia beta driver that is supposed to be the cats ass,
> but having the glxgear FPS go down to 400 FPS on full screen looks to be an 
> issue from what I can see, and maybe i have something set up wrong.

If you go full screen glxgears at 1280x1024, then you are rendering just
over 1 million pixels per frame, so at 400fps you are doing
400Mpixels/s.  That takes a fair bit of memory bandwidth on the card and
even if it is a rather trivial thing to render you are still generating
those pixels.  It would be nice to have a more interesting and complex
model to run as a test that would actually use more of the capabilities
of a video chip and not give a totally uselessly high frame rate (60fps
is plenty for me).

--
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists