In article <38193bdd.34450957 at news.lbl.gov>,
j_martin at munged.lbl.gov (Joel) writes:
> Hi
> I'm looking at what systems would be good for running
> blast/fasta nucleotide searches and am coming up empty in trying to
> figure out what impact large cache would have for these programs. As
> the chip prices are so massively affected by cache size for every
> vendor I've looked at I'm wondering if anyone has information on
> this especially as regards sun ultrasparc II chips and intel Xeons
> in multiprocessor (4+) systems.
We've compared a lot of Suns (anything from an Ultra 5 to an E6000).
NCBI blast doesn't seem to be affected much by cache size, just by
clock speed.
So we're getting quad processor 550MHz P-III Xeons with 2G of RAM
(actually the Dell 6350) running Solaris and they run blast very
nicely. We've got the nice cheap 512K cache variant, as there doesn't
seem any point in paying through the nose for extra cache (although we
haven't checked to see how the performance varies with cache size on
the P-III, but the cost of the 2-meg caches doubles the price.)
--
-Peter Tribble
HGMP Computing Services
http://www.hgmp.mrc.ac.uk/~ptribble/