> The first time, when i rebooted one of my dimms was clearly toast. > memtest86 reported numerous errors starting from the first bit of it. > The second time, everything was fine after a power cycle. soft reboot > didn't help. > > I'm wondering if it might have something to do with my having the ati > framebuffer driver loaded, which would be great, since I don't need it > and won't include it in my next kernel build. No I've seen that same weird 'have to power it off completely' behaviour on an Athlon with AMD chipset too. In my case it was XFree 4.0.2 bugs crashing a 3dfx card. What happens to require power cycling, and who is scribbling into the memory or mashing the settings I don't know