Originally my desktop workstation was built with a motherboard which had the VIA KT266A chipset. This configuration with a AMD Athlon XP 2000+ CPU showed problems in that the workstation would freeze occasionally. Dalco, my system vendor, agreed to replace the board.
By the time I got the board replaced, I could fairly reliably reproduce the system freezes. I just had to do a full backup of the contents of my Sony Clie, which contained some 10 MB of data (easily done with plucker and a few e-books from the Project Gutenberg). I did the backup using coldsync over USB. The easiest way to notice the freeze immediately was to have some music played through the sound card. When the system froze, the same snippet of sound would be played over and over again (like 3 or 4 times a second). This means that there was about 0.3 seconds worth of data in the sound card's buffer and no interrupts would be served by the CPU anymore to update that buffer (and anything else).
I tried this with the Linux kernel 2.4.19 (and the corresponding config file) with and without the preemptible kernel patch, and with the Linux kernel 2.4.20. All that varied was the time it took to produce the freezes. About 2 minutes of constant USB data transfer would do it, but sometimes I had to do more than one backup, i.e. more than 5 minutes of USB transfers.
It is not entirely clear to me, why this chipset was causing freezes
when most of the time it worked ok. Under some ill-defined
conditions it caused the CPU to stop. The number of interrupts, the
amount of data transfered, or something like that is part of the
explanation. But then again, the system behaved well under heavy
load (like compiling a programm with make -j5). Also
straight I/O from the ethernet to disk using the full speed of a
private 100 MBit/s switch didn't affect the stability.
Playing a audio CD or reading it out with cdparanoia caused the load to go up like crazy (I had to kill the program) but the system never froze under these conditions. It just almost crawled to a halt and killing the command brought the system back to normal. Viewing a DVD on the other hand was working quite nicely.
In the end I'm puzzled at exactly where the problem lies. All I can say is that changing the main board and using an nVidia chipset solved the problem as far as I can see. So far I didn't manage to freeze the new system like the old.
If you search on google you can find many other reports of similar problems with VIA chipsets. But then again, you will find problems for almost any configuration. Linux has some fixes which are specifically for the VIA chipset but they don't seem to fix the problem completely.
/sbin/lspci
00:00.0 Host bridge: VIA Technologies, Inc. VT8367 [KT266] 00:01.0 PCI bridge: VIA Technologies, Inc. VT8367 [KT266 AGP] 00:06.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 78) 00:07.0 Multimedia video controller: Brooktree Corporation Bt878 Video Capture (rev 11) 00:07.1 Multimedia controller: Brooktree Corporation Bt878 Audio Capture (rev 11) 00:08.0 Multimedia controller: Philips Semiconductors SAA7146 (rev 01) 00:11.0 ISA bridge: VIA Technologies, Inc. VT8233 PCI to ISA Bridge 00:11.1 IDE interface: VIA Technologies, Inc. Bus Master IDE (rev 06) 00:11.2 USB Controller: VIA Technologies, Inc. USB (rev 1b) 00:11.3 USB Controller: VIA Technologies, Inc. USB (rev 1b) 00:11.4 USB Controller: VIA Technologies, Inc. USB (rev 1b) 00:11.5 Multimedia audio controller: VIA Technologies, Inc. VT8233 AC97 Audio Controller (rev 10) 01:00.0 VGA compatible controller: Matrox Graphics, Inc. MGA G550 AGP (rev 01)