[Beowulf] Opteron performance
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Bill Broadley bill at cse.ucdavis.eduWed Dec 8 01:08:46 PST 2004
- Previous message: [Beowulf] Opteron performance
- Next message: [Beowulf] CPU Benchmark Result
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Tue, Dec 07, 2004 at 03:36:30PM -0800, Joel Jaeggli wrote: > On Tue, 7 Dec 2004 pesch at attglobal.net wrote: > >Where can I find a decent comparison betwenn Opteron and Athlon64 > > functionaly the only difference betten the two is memory bandwidth and the > number of ht ports. Er, it's a bit more complicated than that. Athlon 64's are available in 3 flavors: s754 athlon 64s have a 64 bit memory bus (unregistered memory, 1 less cycle) s939 athlon 64s have a 128 bit memory bus (unregistered memory, 1 less cycle) s940 athlon 64s have a 128 bit memory bus (registered, 1 more cycle) The FX is available in both s939 and s940, but are "unlocked" so you can change the multiplier (useful mostly for overclocking). Athlon 64s have 2 cache sizes, 512K and 1024K. Be careful often the same cpu rating (like the athlon 64 3200) are available in 2 cache sizes and 2 clock speeds. So the main performance differences are 1/2 the memory bandwidth for s754 chips, and 1/2 the cache size for some athlon 64's. The effect of either will depend on the application. AFAIK, the s939 athlon 64's have 3 HT's just like the opteron, but none of them are coherent, so you can't run SMP with them. -- Bill Broadley Computational Science and Engineering UC Davis
- Previous message: [Beowulf] Opteron performance
- Next message: [Beowulf] CPU Benchmark Result
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
