Richard Carter

Computer Science & Game Development Student at North Carolina State University

I just wrestled with Arch Linux for several hours before figuring out a stupid mistake.

If the output of glxinfo | grep -i rend includes Software Rasterizer but already has direct rendering: Yes, then it’s very likely that you simply didn’t add your user to the video group. Do so by running usermod -a -G video [username] as root (or with sudo). After logging out and back in (or maybe a reboot), the output should now include Mesa and your video should be running fullspeed.

I also recommend undoing whatever changes you might have made from the ATI page and then going through them again, after you add yourself to the video group. For example I was messing with early and late KMS start but after adding myself to the video group I removed that; I basically undid all my changes, leaving only my custom xorg.conf file, and the video group was indeed the cause of all of my issues.

However, you may want/need to install radeon_ucode and maybe radeon-initrd still despite the video group. Do a search for installing packages from AUR, and follow the simple steps to get those installed in your system, and they will take effect on the next reboot. After radeon_ucode is successfully installed, dmesg | grep firmware should say something like “platform radeon_cp.0: firmware: requesting radeon/R300_cp.bin”; and the last line of dmesg | grep radeon should tell you that it initialized radeon. In any case, I installed those packages and left them in after discovering the video group, and my card is working great. So I don’t think they hurt anything.

Ugh. I can see the use in this feature for a multi-user system, but at the same time, it’s a little annoying that the Beginners’ Guide wasn’t very clear about it and the ATI page had nothing about it. I guess the moral of this story is, debug things as root first; the root user would’ve run video at full speed and then I may have narrowed it down from there.

Also a little sidenote: don’t use glxgears as a benchmark; the FPS can vary widely depending on the size of the window. For a moment I thought xmonad was slowing down my graphics compared to xterm, but instead I was just getting lower glxgears FPS in xmonad because it rendered on the entire left half of the screen, whereas xterm opened glxgears in a tiny window up in the corner of the screen. False alarm! :)

Add A Comment

Sorry for the mess! I recently got hacked and lost my previous theme, so I'm working on reconstructing that. In the meantime, this theme is messy/broken but at least you can read my posts. Sorry!!


Hi, I'm Richard Carter! I use this blog to document particularly difficult-to-solve computer problems. My posts are written for clarity and keywords for search engines to pick up on, so that the next person that runs into the same problem will easily find my solution here and have an easier time than I had! I'm forging a path through the brush, so to speak. So if you came here by search engine, I guess it worked! Enjoy the solution to your problem.