| Summary: | r300 crashes enemy territory | ||
|---|---|---|---|
| Product: | DRI | Reporter: | Rui Caridade <waiha_othafa> |
| Component: | General | Assignee: | Default DRI bug account <dri-devel> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | high | CC: | glisse |
| Version: | DRI git | ||
| Hardware: | Other | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
|
Description
Rui Caridade
2006-02-16 23:55:33 UTC
Could you please use instead something like this: Option AGPMode "4" #for me it's more sane option Option ColorTiling "false" #I'm having troble with this Option GARTSize "16" #This is not obrigatory.... It's just to test, i don't know if this afect you but in may case i was afected... P.S:Disable VideoRam override, i think... Thanks.This solved my problems.It isn't as fast but at least it doesn't crashes every 5 min (In reply to comment #1) > Could you please use instead something like this: > > Option AGPMode "4" #for me it's more sane option > Option ColorTiling "false" #I'm having troble with this > Option GARTSize "16" #This is not obrigatory.... > > It's just to test, i don't know if this afect you but in may case i was afected... > > P.S:Disable VideoRam override, i think... (In reply to comment #2) > Thanks.This solved my problems.It isn't as fast but at least it doesn't crashes > every 5 min > > (In reply to comment #1) > > Could you please use instead something like this: > > > > Option AGPMode "4" #for me it's more sane option > > Option ColorTiling "false" #I'm having troble with this > > Option GARTSize "16" #This is not obrigatory.... FWIW, there were known compatiblity/stability issues with some stepping of r300 chips (ae2), some graphic cards and some motherboards with agp 8x. Some of these seem to be fixable by bios updates, but I'm not sure if in some cases agp mode just has to be limited to 4x (in all OS). I'm not sure why ColorTiling would cause lockups, if it does it might be due to bogus cache flushes or something, in which case it would not be directly be due to ColorTiling but just more likely to happen if this is enabled. I haven't done any measurements on r300, but I'd expect performance to suffer quite a bit with tiling disabled (which might also be a reason for the lockups to disappear, as timing is going to be different). Well on glxgears if i activate both EnablePageFlip, ColorTiling and AGP 8x i get almost 4000 points.With them disabled and AGP 4x i get around 1900.My card is a powercolor 9700 pro.I've gotta google for its stepping. (In reply to comment #3) > (In reply to comment #2) > > Thanks.This solved my problems.It isn't as fast but at least it doesn't crashes > > every 5 min > > > > (In reply to comment #1) > > > Could you please use instead something like this: > > > > > > Option AGPMode "4" #for me it's more sane option > > > Option ColorTiling "false" #I'm having troble with this > > > Option GARTSize "16" #This is not obrigatory.... > > FWIW, there were known compatiblity/stability issues with some stepping of r300 > chips (ae2), some graphic cards and some motherboards with agp 8x. Some of these > seem to be fixable by bios updates, but I'm not sure if in some cases agp mode > just has to be limited to 4x (in all OS). I'm not sure why ColorTiling would > cause lockups, if it does it might be due to bogus cache flushes or something, > in which case it would not be directly be due to ColorTiling but just more > likely to happen if this is enabled. I haven't done any measurements on r300, > but I'd expect performance to suffer quite a bit with tiling disabled (which > might also be a reason for the lockups to disappear, as timing is going to be > different). (In reply to comment #4) > Well on glxgears if i activate both EnablePageFlip, ColorTiling and AGP 8x i > get almost 4000 points.With them disabled and AGP 4x i get around 1900.My card > is a powercolor 9700 pro.I've gotta google for its stepping. I'm not sure you can figure that out this way, otherwise identical cards may have been sold with new and old steppings. Though you could certainly figure this out when looking at the markings of the chip. There is also a chip reg which in fact xorg reads to determine the revision, though it doesn't print it out. I believe this information is also displayed in some other OS driver (or maybe you need an external tool there). You could use radeon_dump to figure it out, it's in reg 0x00e0 (RADEON_CONFIG_CNTL) (though I don't know right now how the number there relates to the "named" stepping). In any case, the huge performance loss in glxgears is mostly due to disabling tiling and disabling page flip, and not due to agp 4x mode (though especially page flip helps a lot more there than in reality probably, due to the insanely high framerate). glxgears really is a bad benchmark. Well after a bit more testing i found out that i also need to enable SWcursor.Without it the crashes are much more seldom but still happen.With SWcursor enabled and AGP 8x it takes longer to crash than with SWcursor disabled and AGP 4x. (In reply to comment #5) > (In reply to comment #4) > > Well on glxgears if i activate both EnablePageFlip, ColorTiling and AGP 8x i > > get almost 4000 points.With them disabled and AGP 4x i get around 1900.My card > > is a powercolor 9700 pro.I've gotta google for its stepping. > I'm not sure you can figure that out this way, otherwise identical cards may > have been sold with new and old steppings. Though you could certainly figure > this out when looking at the markings of the chip. There is also a chip reg > which in fact xorg reads to determine the revision, though it doesn't print it > out. I believe this information is also displayed in some other OS driver (or > maybe you need an external tool there). You could use radeon_dump to figure it > out, it's in reg 0x00e0 (RADEON_CONFIG_CNTL) (though I don't know right now how > the number there relates to the "named" stepping). > In any case, the huge performance loss in glxgears is mostly due to disabling > tiling and disabling page flip, and not due to agp 4x mode (though especially > page flip helps a lot more there than in reality probably, due to the insanely > high framerate). glxgears really is a bad benchmark. > (In reply to comment #6) > Well after a bit more testing i found out that i also need to enable > SWcursor.Without it the crashes are much more seldom but still happen.With > SWcursor enabled and AGP 8x it takes longer to crash than with SWcursor > disabled and AGP 4x. This is with mergedfb ? #5376 is the relevant bug report. No i'm not using mergedfb.I've been doing some tests and if i use the ati driver instead of radeon on xorg.conf i can enable EnablePageFlip and play a Enemy Territory even if after a while(random time) it crashes. (In reply to comment #7) > (In reply to comment #6) > > Well after a bit more testing i found out that i also need to enable > > SWcursor.Without it the crashes are much more seldom but still happen.With > > SWcursor enabled and AGP 8x it takes longer to crash than with SWcursor > > disabled and AGP 4x. > > This is with mergedfb ? > #5376 is the relevant bug report. With the drm modules and the radeon from git Enemy Territory no longer crashes. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.