HI Jonathan, thanks for the quick reply.
Hi Alex:
On Friday 16 March 2001 19:34, you wrote:
Hi List,
I was wondering whether there is a general intro and/or tutorial somewhere on the web dealing with issues of video hardware acceleration.
I would look in the installed help data base on SuSE 7.1. this is the large life preserver icon.
I already did, and followed their prescribed procedures which led to a working desktop with 3D acceleration, alas not with 1600x1200 resolution. I also looked at the support database and even downloaded NVidia's FAQ. I guess, what I really would like to know is whether my video card has enough RAM to support 1600x1200 in accelerated mode. I got 32 MB, and again, it works fine at 1600x1200, even with 24 bit, if I don't touch 3D. There is a section in the SuSE handbook that says: The maximum possible size of a virtual screen depends on the amount of memory installed on the graphics card and the desired color depth, not on the maximum resolution of the monitor. Since modern graphics cards have a large amount of video memory you can create very large virtual desktops. You should note, though, that you may no longer be able to use 3D functionality, if you fill practically the entire video memory with a virtual desktop. If the card has 16 MB video RAM, for example, the virtual screen can be up to 4096x4096 (!) pixels in size at 8-bit color depth. Especially for accelerated cards, however, it is not recommended to use up all your memory for the virtual screen, since this memory on the card is also used for several font and graphics caches. Well, if that is correct my 32 MB should support 4096x4096 with 16 bit resolution. If I scale it down to 1600x1200 there should be enough memory left to fulfill the needs of the 3D hw acceleration. What gives?
I am trying to educate myself on a little problem that I keep having with my video drivers. I am a little miffed that I can only get 1280x1024 of 16 bit on my 21" monitor using a 32MB AGP2x TNT2 card (Elsa Synergy II) when I turn on hardware acceleration. Gears claims on average 1000 fps in this mode but I am not sure whether it tells the truth. It doesn't look that fast. Also, I find these ugly horizontal bands across the screen. They arise from a black (or dark) area on the left of the screen (e.g., the cursor in an xterm) and stretch all the way across the screen, almost as if the color had been bleached a bit in that area.
See page 251 (of the SuSE Handbook). Alex it seems that the frequency may be too high. Also only 16 bpp works for 3D. I am running 1024 x 768. The video corruption on the left is an indication of a faulty setting. Read /var/log/XFree86.0.log. Entries preceeded by (WW) like so:
WW) The directory "/usr/X11R6/lib/X11/fonts/cyrillic" does not exist. Entry deleted from font path.
are warnings. Usually this file will give you information about what is going wrong. An (EE) is an error and indicates the need for correction.
Yes, I noticed these, too. My log file didn't contain any errors and the warnings I got rid of by editing my XF86Config file, commenting out fontpaths that were not available, and setting the Option "OverridePolarity" "1" in the screen section as it complained about polarity or something. Nonetheless, there hasn't been any improvement whatsoever.
When I turn hardware acceleration off I can enjoy a nice spacious desktop with crispy clean 1600x1200 @ 24bit. Thus, I am not really sure whether the limitation under hw acceleration is hardware limited or driver limited.
I would guess that it is driver limited. Also log into root and run 3Ddiag; it may give some useful info. Again on your computer look at:
/usr/share/doc/packages/nv_glx It discusses limitations of the current NVIDIA driver
I guess, that is probably right. Looking a bit closer at the NVidia FAQ, I found the following line: Late in the process of testing our driver, we found that it is possible to see some image corruption if your physical screen size is set to 1600x1200. We will fix this in the very near term. Unfortunately, their latest driver which I also tried doesn't seem to fix the problem, yet. This remark seems to refer to version 0.9-5 which is currently installed on the system. Thanks again for the pointers, Alex.