Not all Unhappy campers on ati

[aceman]

What really astonishes me is the stability of the development trees. I'm running kernel 2.6.39-rc4 with drm/mesa/xf86-video-ati from git and everything is completely stable. I haven't submitted a bug report in ages.

Great for you. I am stuck at kernel 2.6.36, because with newer kernels the X server hangs with black screen. Problem reported but no progress so far. I am on the newest (stable) mesa and radeon driver (with HD 4350).

But before this problem, the driver and kernel and mesa combo was quite stable across versions, I am satisfied with it. It just has many graphical bugs (celestia, stellarium, kwin/video tearing), which are slowly getting better.

[darkbasic]

Newer nv cards (dx10+) seem to have better oss drivers than ati cards. Did you check out nouveau lately? There are speed differences but not as extreme like ati. Also more games work - waiting for trine to work with ati/intel.

Can you pleas post some nvidia vs nouveau benchmarks? I'm sorry but I REALLY doubt it.

[pingufunkybeat]

The reason why Trine worked correctly in Nouveau was (ironically) incorrect code. Nouveau did not check a flag and went along, while all other drivers checked the flag and acted accordingly.

Then another bug had the flag set incorrectly, causing all other drivers to fail.

Great for you. I am stuck at kernel 2.6.36, because with newer kernels the X server hangs with black screen. Problem reported but no progress so far. I am on the newest (stable) mesa and radeon driver (with HD 4350).

This is no help for you, of course, but I've run my HD 4350 on .37, .38 and ,39-rc3, with no problems.

Are you absolutely sure you're not making any mistakes when compiling the newer kernels? Can you link the bug report?

[aceman]

This is no help for you, of course, but I've run my HD 4350 on .37, .38 and ,39-rc3, with no problems.

Are you absolutely sure you're not making any mistakes when compiling the newer kernels? Can you link the bug report?

Thanks, at least there is hope Here is the bugreport: https://bugs.freedesktop.org/show_bug.cgi?id=31974. Of course it may depend on wrong compile options or driver options. All starts fine e.g. with NoAccel driver option.

[pingufunkybeat]

Hrm, it might be related to vga-switcheroo or something. I have an HD 4350 (PCI) and an HD 4550 (PCI-x) in this computer and can't remember the last lockup I've had.

Do you compile your own kernels? Are you sure you're using the same config for older and newer kernels?

[aceman]

Yes I compile my kernels and always migrate the same config to newer kernels, answering just new options. I do not have 2 cards in the system. But I could try to disable vga arbitration in the kernel (I have it compiled in if I need it in the future).

However I am not the original reporter of the bug, who has 2 cards. Maybe my problem is different from his just the symptoms are the same.

[Zhick]

Yes I compile my kernels and always migrate the same config to newer kernels, answering just new options. I do not have 2 cards in the system. But I could try to disable vga arbitration in the kernel (I have it compiled in if I need it in the future).

However I am not the original reporter of the bug, who has 2 cards. Maybe my problem is different from his just the symptoms are the same.

I'm the original reporter of that bug. And I also think we're actually hitting different bugs, as since recently my problem has disappeared. Unfortunately I've no idea what exactly fixed it for me, but at least with 2.6.39-rc2 and ati 6.14.1 it's now working for me... at least kinda. The XServer starts up and Xorg-log reports EXA and DRI2 are enabled, but everything is veeery slow... for example I only get ~16FPS on OpenArena.
Anyway I really should go and update that bug-report.

[aceman]

Soooo... I am sure kernel 2.6.38 didn't work. But now I have compiled 2.6.38.3 AND disabled vga arbitration in the kernel. And now everything works fine as before. Yes, even xterm redraws ARE slower but e.g. mplayer and googleearth ARE not noticeably slower. But I will look into it more later and post in the bugreport. Thank you guys for poking into it

[business_kid]

I hardly expected one reply to this thread I started. There was no problem for people to fix - merely an opinion. 30 replies and counting! It has been (and continues to be) a source of

1. Stories of pain and suffering.
2. Tricks, dodges, and wheezes to get past problems.
3. Detailed bug descriptions, indicating hours spent over code.

I clearly touched a nerve. As an electronics hardware guy, I know a lot of the software incompatibility can be down to design choices and timings. Speed everything up 15%, and you _should_ do the sums all over again, but people sometimes don't.

I'm reasonably satisfied with what I have, except for one thing: MyHP Compaq 6715S laptop has a RS690/SB600 chipset, housing, I believe, an RS600 (not R600) GPU. Graphics predictably suck. It's running XFCE on Slackware-13.1 (xorg-server-1.7.7, pixman-0.16.6, libdrm-2.4.20, mesa-7.8.1) with that, and kernel 2.6.33.4. Is there any serious speed advantage to be gained by updating anything there? Software rendering is as fast as "DRI"

[agd5f]

I'm reasonably satisfied with what I have, except for one thing: MyHP Compaq 6715S laptop has a RS690/SB600 chipset, housing, I believe, an RS600 (not R600) GPU. Graphics predictably suck. It's running XFCE on Slackware-13.1 (xorg-server-1.7.7, pixman-0.16.6, libdrm-2.4.20, mesa-7.8.1) with that, and kernel 2.6.33.4. Is there any serious speed advantage to be gained by updating anything there? Software rendering is as fast as "DRI"

Upgrade to r300g built with llvm support if you are not using that yet. It will significantly improve your 3D performance.