On Sun, 4 Oct 1998, Robert G. Brown wrote:
> On Sun, 4 Oct 1998, Chris Mauritz wrote:
>
> > > From [EMAIL PROTECTED] Sun Oct 4 13:44:37 1998
> > >
> > > I got the motherboard for $75 and the two PPro 180's for $100 each.. You're
> > > telling me that I can get a dual PII/300 for the same amount? Last I looked,
> > > PII/300's were about $300 each.
> >
> > Ah. You're using PPro-180's. I just built a dual PII-300 machine
> > for about $700, but I used a rather expensive Tyan Tiger motherboard.
> > Last I checked, PII-300's were just over $200 each. So if you're
> > able to get a cheap board and cheap PPro 180's, perhaps it isn't
> > worth it.
>
> Oh, it's "worth" it -- each PII is nearly twice the speed of the PPro
> and there are rather a lot of advantages to having a faster CPU compared
> to having two slower ones that add up to nearly the same speed. Also,
> PPro's are dead technology. Sure they're cheap -- they're cheap for a
> reason.
A PII/400 is 2x the pro, but I've never seen a PII/300 that is 2x
faster... not quite 1.5 in fact, due to the poor L2 cache on the
PII.
The "socket 8" is probably dead, but the "pro" is alive and well in the
new form called "xeon"...
>
> Two configurations that I think are a better idea than a dual PPro right
> now include:
>
> a) a dual PII mobo and just ONE PII -- cost, maybe $120 + 260 = $380.
> Just about $100 more than you are spending, but in six months you might
> be able to add the second PII for just $150-180. Your eventual total
> cost would then be just over twice what you'd spend for the dual PPro,
> but your eventual performance would be between 3 and 4 times as much.
>
I'm not sure how you are going to do this... even a dual PII/400 is not
4x a dual PPro, at least none of the ones I have used are... and I've
tried dual PII/400's with 100mhz bus speeds as well...
> b) a single Celeron 300A. This costs almost exactly what you plan to
> spend -- $100+180 = $280 -- and gives you a 300 MHz cache (albeit a
> slightly smaller one). It might be worth it to up the ante $20-30 and
> get a BX motherboard with 100MHz memory, as this would more than
> compensate for the small clock differential and smaller cache. Although
> I'm sure that one could find some specific parallel task mix for which
> the dual PPro would be better (something just the right size to fit the
> one cache but thrash the other, for example) but on almost any normal
> task mix I'd expect the one Celeron processor to outperform a dual 180
> PPro without the hassle of parallel programming.
>
> Remember that EACH task you run will run close to twice as fast as it
> would on a 180 PPro. Unless ALL of your work is run in near perfect
> parallelism and is just the right size to exploit the PPro's slightly
> larger cache (on slower memory, with more overhead) you lose with the
> dual PPro configuration.
>
> I love smp systems, but I also love optimizing cost-benefit, and it is
> very rare indeed that retail hardware on the edge of obsolescence is
> worth investing in.
>
> rgb
>
no argument there, but I have a quad P6/200 that will hang right in there
with the best dual PII/400's... although the new xeon is quite a machine
and will support 4x machines which will definitely toast a quad P6...
> Robert G. Brown http://www.phy.duke.edu/~rgb/
> Duke University Dept. of Physics, Box 90305
> Durham, N.C. 27708-0305
> Phone: 1-919-660-2567 Fax: 919-660-2525 email:[EMAIL PROTECTED]
>
>
>
