> Well the data is at least very clear that if you have a Matrox card, dont
> buy AccelX or MetroX.
>
> XFree86 is also much faster than AccelX with my Mach64, on the same
> hardware. So thats another data point.
>
> Have you benchmarked AccelX versus XFree86 3.3 on your machine, using
> xbench? If so what numbers did you get?
Well, I spent in the range of 7 hours tonight installing and benchmarking
XF86 3.3 vs. AccelX 1.3. Here's what I came up with on my hardware:
Hardware: Pent133, 512K PB Cache, 64MB EDO RAM, Diamond Stealth 64 Video
VRAM (S3968, IBM RGB525 RAMDAC - 220MHz, 4MB VRAM).
Server resolution mode xbench rating
Accel-X 640x480 8bpp 412332 xstones
Accel-X 1152x900 24bpp packed 141681 xstones
XF86-3.3 1152x864 8bpp 268582
XF86-3.3 1152x864 24bpp unpacked 91513 (*)
(*) - I was unable to use 24bpp packed mode since XF86-3.3 doesn't seem to
support that mode on my harware. The XF86-3.3 server completed the
benchmark, but after restarting X around 50 times with slightly different
config options each time, I was unable to eliminate problems with snow and
random lines showing up in the display with XF86, as well as I was unable to
eliminate a pixel stretch problem at the left hand side of the screen that
the Accel-X server does not exhibit. The last version of the config file
that I had has been saved if anyone cares to see where I was headed with it
(or in case some XF86 programmer sees this and wants to know what kind of
problems I was having).
Now, even though my hardware isn't "stellar" by any means (I'm sure the
numbers would go up if I had tried this in single user mode to avoid
background disk transfers from daemon processes and if I had a better CPU),
they are respectable for the mode when you consider that the cards on the
web page Dan pointed out were turning out around 250,000 xstones in a
comparable video mode. In 8bpp modes, they turned out as high as 1.3 or 1.4
million I think I recall, so my smallish 400,000 may not compete there, but
you never know, a PentII 266 should increase my 8bpp mode quite a bit as
well :) Also, I didn't bother setting the resolution down on the 8bpp XF86
test since XF86 doesn't care about resolution in the sense that the virtual
size would have been the same either way and memory usage in the VRAM would
remain constant between both modes with the only difference being that area
that was drawn to and the area the RAMDAC was scanning for pixel information.
-- ***************************************************************************** * Doug Ledford * Unix, Novell, Dos, Windows 3.x, * * dledford@dialnet.net 873-DIAL * WfW, Windows 95 & NT Technician * * PPP access $14.95/month ***************************************** * Springfield, MO and surrounding * Usenet news, e-mail and shell account.* * communities. Sign-up online at * Web page creation and hosting, other * * 873-9000 V.34 * services available, call for info. * *****************************************************************************