archive: Re: SETI re:seti@home <- Correction

Re: SETI re:seti@home <- Correction

Alfred A. Aburto Jr. ( (no email) )
Sun, 16 May 1999 07:03:12 -0700

I have a Pentium II MMX 200MHz system with 96MB RAM
running Windows95 and one data unit takes approximately 88
hours of CPU time (I've completed 5 data units so far using
the longer running code. The first 2 data units I processed ran
in about 10 hours each, but they didn't do as much processing
as the last 5 data units).

I don't think it is the display updating that chews up the time
because the 0.074506 Hz FFT takes an average of 4.2+/-0.2
seconds to do but the shorter FFT's are much faster. The
0.596046 Hz FFT updates the display very fast for example.
It seems that the FFT's under the Windows95 OS are just
pretty slow!

So I don't think it is a display issue.

It is probably an OS or code implementation or both issue. The
LINUX code on the Pentium seems to be much more efficient.

I think I'll try switching to LINUX!

Al

> Corey Ashford wrote:

> I just checked the clock rates on the four solaris
> machines. The fastest is 360 MHz, and the other
> three are running at 296 MHz. SPARC chips are still
> a bit down the curve on process technology, but they
> give a decent bang per clock cycle. I'd be willing
> to bet that a fair amount of processing time is
> wasted on the Windows version with updating the fancy
> screen display. The Unix version have a simple (but
> kinda boring) line-oriented output. It's fine with
> me thought, because it makes adding monitoring tools
> a snap.
>
> The 360 MHz machine finished its first packet in
> 8 hours 26 minutes.
>
> If they make other Unix flavors available (MIPS Irix, PowerPC AIX,
> Alpha OSF, PA-RISC HP-UX) I have machines I can run those on too!
> Just yesterday they had some of those on the web site,
> but they are gone today. Perhaps they are readying
> final versions for the May 17th launch.
>
> --- Corey Ashford <yeroca@yahoo.com> wrote:
> >
> > I now have 6 machines running this code.
> > One Linux machine, P166 w/96 MB of RAM. It took
> > about
> > 48 hours for the first packet.
> >
> > I also have a P233 running it, but has yet to
> > complete
> > its first packet.
> >
> > Here's the good part. I have 4 Sun Solaris Ultra-60
> > workstations
> > running between 300 and 466MHz running
> > this code. I started one today around 2pm, and it's
> > about ready to finish its first packet in the next
> > 20 minutes or so... so it's about 10 hours for that
> > machine. Nice! The rest started a bit later and
> > are a little slower. They'll be finishing up
> > their first packets around 3am or so.
> >
> > This whole thing's kinda fun! I hope it works out.
> >
> >
> >
> >
> > --- "Alfred A. Aburto Jr." <aburto@cts.com> wrote:
> > > Humm,
> > > Pentium II MMX 200MHz, Windows95, takes about 90
> > > hours
> > > to do one data unit.
> > > Al
> > >
> > > > John wrote:
> > >
> > > > 75% done in 15 hours.
> > > > K6 II-350/3D , FreeBSD 2.2.7
> > > >
> > > > -john
> > >
> > >
> >
> > _____________________________________________________________
> > Do You Yahoo!?
> > Free instant messaging and more at
> > http://messenger.yahoo.com
> >
>
> _____________________________________________________________
> Do You Yahoo!?
> Free instant messaging and more at http://messenger.yahoo.com