I used to have frequent X screen corruption when running under KMS (kernels 2.6.32 to 2.6.37) and the Radeon X driver. I've identified the error in dmesg as a "GPU lockup (waiting for * last fence id *)". About a week ago I dug into the BIOS and found an option to turn off the onboard Sideport memory of my motherboard. Now, over a week later I've yet to experience the GPU lockup. Is this just a coincidence? Has anyone else experienced any problems running KMS with integrated Radeon graphics and Sideport memory? Before last week, I set the onboard graphics to use the 128 MB Sideport memory with various amounts of conventional RAM (which the BIOS calls UMA), ranging from 64 to 512 MB. I would experience the X freeze when running multiple simultaneous instances of software such as Firefox and XPDF, which appear to do something funky to the screen. I avoid or delay the onset of the freeze by running at most two or three of these programs and using only simple Firefox themes. When the freeze happens, I immediately press Ctrl-q (or its equivalent) to exit the corrupted program. Continuing to use the program often leads to a more permanent black screen of death that requires a hard reboot. But since turning off Sideport memory and setting the conventional shared memory to 256 MB, I've had an uncorrupted uptime of 7 days while simultaneously running XPDF, three Firefox profiles, Chromium, Mplayer (paused at -vo gl), and Gnome Terminal doing an IO-heavy file transfer!
- Prev by Date: [volkerdi@xxxxxxxxxxxxx: Re: libdrm 2.4.4]
- Next by Date: problem using G41/i915.
- Previous by thread: [volkerdi@xxxxxxxxxxxxx: Re: libdrm 2.4.4]
- Next by thread: problem using G41/i915.
- Index(es):