On Wed, Nov 07, 2001 at 02:44:23PM +0100, Pavel Jurus wrote: > > I have problems with segmentation faults and almost unreproducable > errors during computations and even with kernel oopses. I'm trying > hard to figure out whether it is hardware problem or bad kernel. > Although it seems to be likely problem with the hardware, I would > like to hear about any experiences with working (or not working) > kernel/hardware configurations for computational intensive tasks > (weather modeling in my case). >
Wild guess: weak power supplies? AMD chips are heavy power users, and want clean power. We had a small AMD K6 cluster that exhibited instability for quite a while before being decommisioned. A few months back one of the former nodes blew a power supply, was replaced with a different brand, and all instabilities vanished. The next cluster is getting higher grade cases/power supplies. -Drake
