Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!unmvax!deimos.cis.ksu.edu!rutgers!rochester!pt.cs.cmu.edu!cadre!pitt!cisunx!spectre
From: spectre@cisunx.UUCP (Robert Sillett)
Newsgroups: comp.sys.tandy
Subject: Re: Accelerator Boards
Message-ID: <17548@cisunx.UUCP>
Date: 14 Apr 89 22:00:43 GMT
References: <1370@cunixc.cc.columbia.edu! <136500014@cpe! <1407@csm9a.UUCP! <3254@ddsw1.MCS.COM> <1411@csm9a.UUCP> <834@jarthur.Claremont.EDU> <1013@nosc.NOSC.MIL> <5264@cs.Buffalo.EDU>
Reply-To: spectre@unix.cis.pittsburgh.edu (Robert Sillett)
Followup-To: comp.sys.tandy
Distribution: na
Organization: University of Pittsburgh
Lines: 26

In article <5264@cs.Buffalo.EDU> ugkamins@sunybcs.UUCP (John Kaminski) writes:
|
|In other words, if your main activities involve data movement to/from/around
|the disks, don't bother.  If they are more computation-intensive, get 
|something like the express order 286 board.  If you do extensive 
|real/floating operations as in many CAD packages, get an 8087.

Another point is that even if you put in the fast 80386 card that you can
buy, you'll still be running the bus a 8 MHz.  That's like driving a
Ferrari only during rush hour.  You'll have the power, but you can't use
it.

I've speed up my 1000 SX about as far as it can go.  I spent $14 for
a NEC V-20 CPU which speeds things up about 10-20%.  I also spent around
$150 or so for an 8087-2 coprocessor.  This has speed my floating point
calculations up around 1000-1100%.  In fact I now preform floating point
math at about 1/2 the speed of an IBM PS/2 Model 50 (10MHz 80286/87).
However this is only for floating point math.  You only _really_ notice
the speed with Lotus or CAD programs.  Since I use math-intensive programs
often, the 8087 is worth it.  It may not be worth the $150 or so to
shave a couple of minutes off a spreadsheet calcuation.

-- 
Robert Sillett -- Univ. of Pittsburgh
Internet: spectre@unix.cis.pittsburgh.edu  [130.49.254.10]
Usenet:   ...!pitt!cisunx!spectre
Bitnet:   spectre@pittunix