Hello, been a while, dual CPU mobos

[Lennart Sorensen]

On Thu, Sep 09, 2004 at 01:43:46PM -0400, Michael Laccetti wrote:
> I think my P4 3.0E (being E) is a Prescott chip.  Overclocked nicely, at
> least.  Noticed a decent boost in speed switching from the 2.8C to a 3.0E,
> if only because of the higher overclock.

Long pipelines tend to help the clock speed boosts.  Doesn't help at all
during branch misses (which they have tuned a bit better in the prescott
too, as it is obviously even more important to it).

With the first steppings of the Pentium Pro, the pipeline would have to
be flushed completely and restarted if it ever encountered a 16 bit
aligned read/write, since it was entirely designed for 32bit only.
Intel I guess underestimated how long it would be until people stopped
runing 16bit code.  Later revisions made some changes to allow it to
detect earlier in the pipeline when it was going to have a problem and
deal with it right away to avoid the whole branch being wasted.  As far
as I recall the PPro was around 11 or 13 stages or so.  P2/P3 being
essentially the same design are probably the same length pipeline.  Not
short pipelines by the standards of the time, but short compared to the
P4 and very similar to the athlon 64.

> My dual AMD MP motherboard just fried, and I was looking at getting a dual
> Opteron replacement, but Tyan kindly replaced the thing for me, so it's
> still stuck at MP land.  Maybe sometime next year I'll upgrade my 3.0 to
> something better.  But much will change in a year... 

Well nice when companies replace stuff that fails.  Probably good to
retain customers that way.

Lennart Sorensen
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml