Very Computer

Your only option for open source 3D accelerated drivers is the
XFree86 radeon driver for ATI radeon based cards. The XF86
drivers for nVidia and Matrox based cards are only 2d enabled.
Check here for more specifics:

http://www.xfree86.org/4.4.0/radeon.4.html

HTH
Chris.

There is DRI support also for the Matrox cards up to and including G550.
However, the ATI cards have better 3D performance. Please be aware that
open source DRI support is only supported up to and including Radeon 9200.

regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc2(at)uthyres.com Examples of addresses which go to spammers:

Quote:> Now another question. No doubt Matrox has an excellent 2D output, for
> quality and speed. Myself, I'm still using an excellent G400.
> What about the 9200? Is the output quality and speed in 2D up to matrox's
> one?

At home I have a Radeon 9200 in a P4 2.6 GHz.
At work I have a Matrox G550 in a P4 2.8 GHz.

If someone suggests some benchmarks I could run them on these machines to
compare and post the results. Both machines run Slackware 9.1 so both
glxgears and and x11perf is installed. Both machines have HT enabled and
an SMP 2.4.22 kernel. At home I use the opensource dri drivers. At work I
think I use the opensource drivers, but maybe the Matrox binary HAL driver
was installed a long time ago.

I think that I read somewhere that Radeon had bad 2D performance unless
DRI was disabled. However, the 2D performance is enough for my needs.

regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc2(at)uthyres.com Examples of addresses which go to spammers:

Quote:> May I suggest xbench? It is quite old, but frankly I don't have better
> ideas: http://www.artfiles.org/xfree86.org/pub/xbench/

Thanks for the suggestion. I downloaded xbench and did a run on my
home-machine. As any serious benchmark program it does not only deliver
one result number but one number for every part of the test. Unfortunately
this also makes it harder to post the results and to compare results.

At work today I started a run of x11perf to give input to Xmark, but it
took a rather long time and I needed the computer before it was finished.

Instead I tried to run glxgears without and with the environment variable
LIBGL_ALWAYS_INDIRECT set to disable DRI. I was hoping that disabling DRI
would give some kind of a simple 2D benchmark and enabling DRI would give
a simple 3D benchmark.

Both machines where running in 16 bit color depth and the glxgears window
had its original rather small size.

The results where:

G550 no DRI:   400 FPS
G550 with DRI: 670 FPS

Radeon 9200 no   DRI:  560 FPS
Radeon 9200 with DRI: 2925 FPS

I don't think the above benchmark is a very good benchmark for 2D. I would
have expected the Matrox card to have better 2D performance than the ATI
card.

Quote:> Actually I'm more interested in output quality than raw 2D performance.

This is harder to measure. I'm not the right man to judge this. At home I
have a rather cheap 19" monitor, at work I have a more expensive, but
rather old 21" monitor. To make things worse I have a KVM switch at work
which causes shadows in the image.

Quote:> At work I have an nvidia mx440 hooked on a Sony G200. Previously I had a
> matrox G400, and the difference is noticeable (not hard to guess which
> one is better).

Yes, with a good monitor Matrox cards really shines.

Quote:> At home, the output of a Matrox G400 was first driving a sony 17" crt,
> and now a gorgeous sony hx 19" tft. The result is really astonishing
> (vga, not even dvi output!)
> The G400 is quite old actually. What should I do if it dies some day?
> From matrox forum I collect that their last products are a no-go for
> linux: bad proprietary drivers, strange phenomena arise sometimes, no
> 2.6 kernel support.

It is still possible to buy G550 cards. To bad Matrox hasn't got better
support for their new cards, they are already a rather small player on the
graphics market and this will make them lose even more customers to ATI
and nVidia.

regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc2(at)uthyres.com Examples of addresses which go to spammers:

Quote:>> The results where:

>> G550 no DRI:   400 FPS
>> G550 with DRI: 670 FPS

>> Radeon 9200 no   DRI:  560 FPS
>> Radeon 9200 with DRI: 2925 FPS

It now occurst to me that I have done some tweaking of XF86Config at my
home macine with the ATI card:

        ChipID      0x4967
#       Option      "AGPMode" "4"
#       Option      "EnablePageFlip" "on"

The ChipID option is necessary for the 9200 to work with XFree86 4.3.0,
but the other ones only give better performance. As I haven't done any
such tweking on my PC at work with the matrox card I now commented the two
lines out as above. Then I get the following results:

Radeon 9200 no   DRI:  496 FPS
Radeon 9200 with DRI: 2261 FPS

Quote:> Still it's strange to see a 40% ati advantage without dri i.e. in
> software only.

Yes, I'm still a little amazed that the ATI card gave better performance
also without DRI enabled. However, I don't claim this benchmark to be
representative for the overall 2D performance of these two cards. If
someone would suggest another benchmark, ie something like "x11perf
-movetree" I could run that benchmark on both macines to compare. However,
I don't want to run any benchmark that takes longer than a few minutes.

Quote:> But there's a catch. If you focus on the glxgears window,
> you obtain some results, like the following (G400, DRI): 2034 frames in
> 5.0 seconds = 406.800 FPS But if you minimise it then it really flies:
> 7923 frames in 5.0 seconds = 1584.600 FPS Restoring the window, the
> result will be much faster than before: 6241 frames in 5.0 seconds =
> 1248.200 FPS

I didn't do anything like that. On both machines I was running gnome and
the window manager gave focus to the new glxgears window. Then I let it
run until the FPS value seemed stable.

Quote:> I suppose that glxgears is not even a handwaving bench.

No, I don't think it was meant to show some overall benchmarking value.
The only reason that I now used glxgears was that it was a fast and easy
way to get any value at all.

Quote:>> It is still possible to buy G550 cards.
> Is it working at least on vga output?

So far I have only had the opportunity to try the VGA output on matrox
cards. The only card where I have tried the DVI output is a Radeon 9000
Pro, and it worked with XFree86 right out of the box without any changes
to XF86Config.

regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc2(at)uthyres.com Examples of addresses which go to spammers: