Path: utzoo!utgpu!watserv1!watmath!att!pacbell.com!ucsd!usc!samsung!sol.ctr.columbia.edu!ira.uka.de!sun1.ruf.uni-freiburg.de!rak
From: rak@sun1.ruf.uni-freiburg.de (Bodo Rak)
Newsgroups: bionet.software
Subject: Re: FASTA speed
Message-ID: <1991Jan29.165916.5272@sun1.ruf.uni-freiburg.de>
Date: 29 Jan 91 16:59:16 GMT
References: <1991Jan25.174726.11329@phri.nyu.edu>
Organization: Rechenzentrum der Universitaet Freiburg, Deutschland
Lines: 37

roy@alanine.phri.nyu.edu (Roy Smith) writes:


>I've often seen comments on these newsgroups, like the one from
>rak@sun1.ruf.uni-freiburg.de (Bodo Rak) today, to the effect that they can
>run FASTA against the whole GenBank database on a PC in 20 minutes.
>Frankly, these numbers astound me; I didn't think PCs were anywhere near
>that fast.

>This brings up the question, however, of what kind of PC are we
>talking about?  I'm used to things like 10 Mhz 286 machines (I do all my

           -          staff deleted       -

Below I will try to give a more precise estimate of the performance of 
Fasta on a quite ordinary AT-type PC uder realistic conditions: 

I have tested the speed of a 20MHz 386 PC by scanning the entire EMBL 
data base (release 24, 37,784 entries comprising 47,354,438 nt) against a 
750 nt long test sequence (ktup=6) with the following results: 
Scan time: 36'50''; total CPU time including the top 50 alignments: 38'16''. 

A 20MHz 386 machine is certainly nothing fancy nowadays. Many of the
laptops are already equipped with this processor. I hope that the above data
encourage those not having acccess to more efficient computers not to 
underestimate the capability of their PCs (or should I say of Fasta even
on a PC?). 

The PC which I used in the test is equipped with a co-processor. I don't 
think, however, that Fasta makes much use of it. 

Bodo Rak
Department of Biology III (Genetics)
University
D-7800-Freiburg, FRG
RAK@SUN1.RUF.UNI-FREIBURG.DBP.DE