Solved! [Re: [Dri-devel] Radeon DRI problems still =(]
On Fri, Apr 13, 2001 at 08:59:13PM -0700, Joseph Carter wrote: On Fri, Apr 13, 2001 at 12:10:11PM -0700, Simon Kirby wrote: Hmm! Where is this option? I can't seem to find it in my BIOS anywhere. Perhaps it's only an option on newer boards. What other things have you tried? P2C/C2P(?) concurrency, turning off AGP-4x mode all together, and other settings. I posted my AGP-related settings about a week ago along with kernel info, system info, etc. Well, I think I just figured out what was doing it. Option "AGPMode" "4" in my X configuration file. I turned it off, and just played Q3 for around two hours in a hot room with no crashes. I'll try AGPMode "2" to see if it crashes with that as well, but it definitely looks like there are some problems on my ASUS P2B-DS (BX chipset) and 4x AGP. Err, maybe it's not even officially supported, which would explain why it doesn't work properly. :) Actually, I'm running the bus at 100 MHz. Dual Slot1 300A Celerons drilled soldered to run at dual 450 MHz, but I left that part out to increase the validity of my report. ;) AMD Duron 800, running 100x8. I don't overclock - my CPU is already 109F as it is! Ahh, that's not too bad. My CPUs were crashing until I made up some custom heatsinks. Two fans on each CPU and duct-tape around the top of all of the fins to keep air inside the sink until near the bottom rather than having the air mostly go out the sides at the top, like Alpha does (but with metal :)). The duct tape actually made a big difference. To reduce ambient, I cut off the stupid air-blocking grills on the exhaust and inlet fan mounts on the case. Then, I took a bunch of old 80mm fans that had died, ripped out guts, and used double-sided tape and duct tape to stick them together to form a "tunnel" through which air flows. This made it suck air out of the case from right above the CPUs, rather than near the edge of the case. Right out the heat sinks and straight out the back -- nice. All of that and a custom dremmel-cut hard drive mount to put the hard drive near the inlet fan (grill also removed) to keep the HDs cool as well as bring in cool air over the PCI cards. Now if only it were a bit more quiet. :) With this double-walled case, most of the noise still goes out the tunnel at the back. Oh well. :) Not bad for about CDN$120 for the CPUs each many years back. One thing I've found is that instead of using screws to mount case fans, use double-sided mounting tape (I use the 3M/scotch stuff). Sometimes it sticks too well, but the padding reduces vibrations which transfer to the case and thus reduces the noise a lot. Also, removing poorly-made grills built into the case often reduce noise a lot, even though air flow increases. This is because both the grill can makes noise, but even more so, most fans are designed to operate quietly at a minimum airflow and become louder with airflow is restricted. Uh, anyway, this is a bit off-topic now. :) Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel
Re: Solved! [Re: [Dri-devel] Radeon DRI problems still =(]
On Sat, Apr 14, 2001 at 02:29:13PM -0600, Harold Oga wrote: On 14 Apr 2001, at 12:54, Simon Kirby wrote: I'll try AGPMode "2" to see if it crashes with that as well, but it definitely looks like there are some problems on my ASUS P2B-DS (BX chipset) and 4x AGP. Err, maybe it's not even officially supported, which would explain why it doesn't work properly. :) Hi, Yes, the BX chipset doesn't support AGP 4x, only 1x and 2x. Is there any way to detect this an at least print an error about it, or deny it by default? I can see this being a problem for others as well... Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] Radeon DRI problems still =(
On Fri, Apr 13, 2001 at 05:02:42AM -0700, Joseph Carter wrote: Actually, what is REALLY freakin' weird is that it seems to work just about to the point that I start typing something, at which point the next thing I type has to start with SysRq because it's locked tight.. It's truly bizarre, and nobody seems to have the slightest idea what could be causing it. I've tried every suggestion I've been given at least once - some with disasterous (but fixable, thankfully) results like changing AGP driving strength. Setting it to the suggested value caused the box not to boot at all. Old ISA video card to the rescue. Hmm! Where is this option? I can't seem to find it in my BIOS anywhere. Perhaps it's only an option on newer boards. What other things have you tried? I suspect we're seeing different problems, and I've pretty much concluded that whatever the hell is wrong, it's specific to my particular setup. It'd most likely take someone with an identical motherboard and video card (possibly also running at 100MHz as I am running it) to duplicate the problems I am having. I'm keeping my eyes open, but at the same time I'm looking into the possibility of a PCI card to replace my Voodoo2 and possibly some sort of setup that way. Actually, I'm running the bus at 100 MHz. Dual Slot1 300A Celerons drilled soldered to run at dual 450 MHz, but I left that part out to increase the validity of my report. ;) Sometimes it will play fine for a few CTF games, but it will usually die somewhere in the game. It seems to die either one of two ways -- either the screen will go blank or become corrupted (for example, scanning the first scan line for the entire screen while my second head on a PCI card remains fine), or the screen will just freeze. In the latter case, the machine will not respond to sysrq. In the former case, it will, and I can unmount and reboot cleanly. [..] I have yet to see anything die while I'm not playing Quake 3, so it seems GL-related. I suppose I should try running UT for a while and see if similar lockups occur. Well, if yu manage to track down the problem, I would be interested in the result. How does it work if you disable the use of the second head? I haven't tried. It's on a PCI board (not the second head of the Radeon VE), so I didn't think it would be affecting it. As I'm running out of other things to try myself, I'll try this in a bit. Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] Radeon lockups with Quake3
On Fri, Apr 13, 2001 at 01:32:39PM -0700, Simon Kirby wrote: I'm trying to compile the glx client but I'm not having much luck. Debian unstable. How are you compiling it over there? Okay, finally got qf-client-glx to compile (library conflicts with DRI tree). It starts up and runs, but I'm not able to find any multiplayer servers and it doesn't let me play single player. I'll search for some in xqf or something. However, in the mean time, there are some strange things happening. As soon as the client starts, my analog flatpanel starts to get all jittery -- pixels are all vibrating left and right. I can _sometimes_ see a single row of pixels jittering normally if the sync is misaligned, but quakeforge isn't even changing the resoution (1024x768) and it just suddenly starts going all funky. Each scanline is different, as if the horizontal sync is changing phase slightly. As soon as I quit, everything returns to normal. I just obscured the window trying to find a multiplayer server, and got a lock up. Sysrq worked, but my mouse was frozen. I hadn't even started playing yet. Trying again... Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] Radeon lockups with Quake3
On Fri, Apr 13, 2001 at 02:07:05PM -0700, Simon Kirby wrote: On Fri, Apr 13, 2001 at 01:32:39PM -0700, Simon Kirby wrote: I'm trying to compile the glx client but I'm not having much luck. Debian unstable. How are you compiling it over there? Okay, finally got qf-client-glx to compile (library conflicts with DRI tree). It starts up and runs, but I'm not able to find any multiplayer servers and it doesn't let me play single player. I'll search for some in xqf or something. However, in the mean time, there are some strange things happening. As soon as the client starts, my analog flatpanel starts to get all jittery -- pixels are all vibrating left and right. I can _sometimes_ see a single row of pixels jittering normally if the sync is misaligned, but quakeforge isn't even changing the resoution (1024x768) and it just suddenly starts going all funky. Each scanline is different, as if the horizontal sync is changing phase slightly. As soon as I quit, everything returns to normal. I just obscured the window trying to find a multiplayer server, and got a lock up. Sysrq worked, but my mouse was frozen. I hadn't even started playing yet. Trying again... Well, I compiled "nuq" instead (the single player version), and I just finished three levels without any problems (other than the weird pixel thing which doesn't happen in q3). I tried obscuring the window, changing virtual desktops, dragging everything over it, etc., and couldn't get it to lock up at all. Before, when I had my G200, I could cause a lockup with any GL program by dragging partially or fully obscuring the GL window. It seems that this isn't causing it, though, or something in the multiplayer client is. Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel
Re: [Dri-devel] Radeon DRI problems still =(
On Sat, Apr 07, 2001 at 08:03:38PM -0700, Joseph Carter wrote: After a few hours, the machine will lock up if I have the drm module loaded. It runs fine for days on end without it, so it's probably AGP related. Lockups happen a lot faster if I run a GLX app that uses more than a few textures. gears and similar run just fine, but something such as QuakeForge will lock up the system usually within a minute. By lockup, I mean that the machine no longer responds to the keyboard and the monitor pops up its no signal message. SysRq commands still work, but remote logins don't. I'm also seeing lockups with Quake3 on my ASUS P2B-DS dual Celeron board with my Radeon VE 32MB DDR. Sometimes it will play fine for a few CTF games, but it will usually die somewhere in the game. It seems to die either one of two ways -- either the screen will go blank or become corrupted (for example, scanning the first scan line for the entire screen while my second head on a PCI card remains fine), or the screen will just freeze. In the latter case, the machine will not respond to sysrq. In the former case, it will, and I can unmount and reboot cleanly. At first I thought it was heat-related, but I've tried with a fan added manually to the VE card's chip and with extreme amounts of air through the entire case, and the lockups seem just as frequent. I've also tried different AGP bus speeds (1, 2, and 4x). No difference. I've also tried "nosmp". No difference. I've also tried 24/32 bpp vs 16 bpp. No difference. This has been happening ever since I got the card working, and over quite a few versions of DRI-CVS. The latest checkout I've tried is from about a week ago. I have yet to see anything die while I'm not playing Quake 3, so it seems GL-related. I suppose I should try running UT for a while and see if similar lockups occur. Simon- [ Stormix Technologies Inc. ][ NetNation Communications Inc. ] [ [EMAIL PROTECTED] ][ [EMAIL PROTECTED]] [ Opinions expressed are not necessarily those of my employers. ] ___ Dri-devel mailing list [EMAIL PROTECTED] http://lists.sourceforge.net/lists/listinfo/dri-devel