From: Michael S. Lorrey (mike@lorrey.com)
Date: Sat Jul 17 1999 - 13:10:30 MDT
J W wrote:
>
> Mike Lorrey stated:
> >A Celeron, put quite plainly, is a POS. It has a really pissy small
> >cache. A dual PII 450 actually is faster than just a 900 equivalence.
>
> Well, according to a couple of engineering friends of mine who just recently
> took an intel assembly language class, running a dual CPU system (by intels
> scheme) is only theoretically going to reach something like 1.75 times as
> fast, instead of twice. The actuality of it is even worse, depending on
> software.
>
> But, comparing apples to oranges, can you really explain this difference in
> speed just by the cache?
> Oh, I was way off in my speed measurement on my celeron, it took about 63
> hours for a packet. Vs. the 8 hours on the dual PII 450. That's about 8
> times faster. Say that his 450 actually achieved the equivalent of a PII
> 900 (which it can't), scale down by three, a PII 300 should be about 24
> hours. My Celeron 300 is 63 hours, so that makes it equivalent to something
> like a PII 115. Can the small cache really cut my speed down to a third?
Yes it can. Cache is necessary because so many operations, functions,
and information used are repetetive, and having to go back to the RAM or
to the swap file adds immense amounts of dead time. Go cheap on your
chache and you are strangulating the processor.
Mike Lorrey
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:04:30 MST