[BiO BB] remove CTL-M and Buying a bioinformatics workstation
mgollery at unr.edu
Thu Sep 4 12:17:41 EDT 2003
The real key in this performance whitepaper may not be the processor or
operating system. Note that the word size used here was 40, which will lead to
dramatically fewer HSP extensions, and thus a much higher speed on any system.
Of course, you might miss a lot of hits, depending on the data. Due to the
larger word size, one might more accurately compare A/G blast with NCBI's
megablast. On the other hand, if you would like speed with sensitivity check
out PatternHunter from Bioinformatics Solutions, or if you have the money,
TeraBlast from TimeLogic.
Still, I think that Apple is becoming an attractive solution for bioinformatics
due to optimizations with the Altivec. The HMMpfam info looks promising,
although I have not tested them myself.
Quoting Yannick Wurm <idh at poulet.org>:
> I don't have any hands-on experience with large-volume blasting, but
> you might want to have a look at Apple's new G5 computers.
> The numbers shown in the "Performance Whitepaper" of june 2003 (you'll
> find a link at http://www.apple.com/lae/g5/ ) are quite impressive.
> Apple compared the performance of the dual 2GHz Power Mac G5 running
> Apple/Genentech BLAST with a 3GHz Pentium 4-based system and a dual
> 3.06GHz Xeon-based system, both running Red Hat Linux 9.0 and NCBI
> A/G BLAST is an optimized version of NCBI
> BLAST developed by Apple in collaboration
> with Genentech. Optimized for dual PowerPC
> G5 processors, the Velocity Engine, and the
> symmetric multiprocessing capabilities of
> Mac OS X, A/G BLAST makes a wide variety
> of searches available at higher speeds.
> According to the graph they show, using a word length of 40, the Dual
> G5 ran 3 million nucleotides per second whereas the Linux boxes did
> only about 0.75.
> The same paper also states that HMMer runs 4 times faster on the dual
> g5 than on the dual xeon.
> And Microsoft Word runs on it too :)
> Are the results shown surprising? The code for A/G Blast and HMMer seem
> to be optimized for the G5, whereas the standard vanilla versions where
> used on the linux boxes. Could optimization reduce the bias against the
> Yannick Wurm
> \\ http://yannick.poulet.org icq: 22044361
> \\ idh at poulet.org tel: ++220.127.116.11.71.92
> BiO_Bulletin_Board maillist - BiO_Bulletin_Board at bioinformatics.org
Associate Director of Bioinformatics
University of Nevada at Reno
Dept. of Biochemistry / MS330
New phone number! 775-784-7042
This mail sent through https://webmail.unr.edu
More information about the BBB