On Mon, 2006-03-20 at 18:59 -0500, Sam Varshavchik wrote: > Gilboa Davara writes: > > > > > I'm using Fedora Core on a Tyan Thunder K8WE with two Opteron 270 CPUs. > > AFAIR the Sun Fire uses the same AMD south-bridge as my Tyan board. > > More-ever, I've got a number of HP 385 machines that use the exact > > chipset configuration as the Sun machine, running FC and CentOS; in both > > cases no stability issues what-so-ever. > > > > In short, it should work out of the box. > > I would advise caution when running an Opteron platform. It's been my > impression that, over time, far more problems crop up on x86_64 than on > IA-32 platforms. I'm running both, and, for the last couple of years, I've > many more issues with my Opteron machine than with the Pentium ones. x86_64 > really gets less overall testing than IA-32, and that's the results you get. > > Even though things work fine now, any minor kernel rev can suddenly turn out > to be broken for your specific chipset combination, while other combinations > of Opteron chipsets and hardware work fine. Right now, for example, the > latest kernel I can boot on my dual-Opteron 240 box is 2.6.13. All later > kernels trigger some bizarre bug that apparently results from my specific > combination of RAID-1 and Adaptec SCSI controller. Before I can make much > progress with booting, kernels 2.6.14 and later begin spewing "SCSI > transmission timeout" errors. 9 times out of 10 the arrays get degraded, > and I have to scramble back to 2.6.13 and rebuild. > > The IA-32 box standing next to it, also running RAID-1 and Adaptec SCSI, > happily eats 2.6.15, without complaints. > > -- I had a couple of minor x86_64 only bugs that I've reported. Beyond that, I never had any problems with x86_64 (and I've been using it since 2003). However, YMMV depending on the exact hardware configuration. Gilboa
