[ut2003io] Frame Rate Questions

Andrew Pilley ashridah at icculus.org
Thu Jan 16 07:11:17 EST 2003


On Thu, 2003-01-16 at 17:26, Paul Berger wrote:
> What is the easiest way to determine if poor frame rates is do to video
> driver or game engine overhead?

in ut2k3's case specifically? i'd consider running it in openGL mode
under windows, and comparing the results. as distasteful as that can
sound, it'll give you an idea of the overheads the opengl (and the OS
itself) are imposing. results are to be taken with a grain of salt, and
remember, d3d vs opengl isn't in the slightest bit fair, because of the
extra hacks needed to get the opengl renderer happening.

You could also do some other benchmarks. grab copies of SPECViewPerf 7.0
(from www.spec.org iirc) for windows and linux. keep in mind these are
BIG files, so don't get them unless you've got a reasonable bandwidth
allowance.

http://www.spec.org/gpc/opc.static/opcview70.html

these tests are designed to be taxing, even on a super-heavy-duty SGI
box, so don't be surprised if the results are slow.

http://www.icculus.org/~ashridah/svp has an example of the results from
my system (duron 750, gf4ti4400)

> 
> On indoor levels I usually get 20-30fps, but outdoor levels I get
> 5-10fps at times. Tweaking the detail settings does not appear to have
> an effect, and neither does the screen resolution.
> 
> I running an ATI Radeon 9000 Pro on a 1.4Ghz Athlon with 512MB RAM ...
> not a wimpy system.  The XFree86.0.log says it's using AGP and DRI is
> enabled using the ATI FireGL driver 2.5.1.
> 

hmm. people keep telling me ati's drivers aren't fantastically fast, but
i'm hesitant to believe that right off the bat.

Even if i may come off critical of ati, it's only because it takes a lot
of effort to beat off a two-year-long impression of "screw linux" by
ati. even WITH their spec releases, they still happily turned non-IE/ns4
users away from their site for a while, iirc. that's all fixed now
afaik, but i'm past caring.

Andrew 'ashridah' Pilley


> Thanks for the help.
> Paul
> 
> 
> 
> 





More information about the ut2003 mailing list