> Greetings all!
>
> We are currently interested in finding out how well Linux works
> with the new Pentium II processors, especially the 300MHz variety. We
> currently have about 50 production servers running Linux-SMP (2.0.27) on
> dual Pentium Pro 200 MHz boxes, but with Intel's announcement that they
> are discontinuing the Pentium Pro line, we need to look to the future.
>
> If you have used Linux with a Pentium II processor we would like
> to hear from you. This is especially true if you have a SMP machine. We
> are interested in answers to the following questions:
There are two issues with the PII that I know of, one major, one
minor.
The minor: Only 2 proc (no quads) SMP machines are supported, a
bummer but not normally a show stopper.
The MAJOR: The PII L2 cache cannot handle more than 512MB due to
a *feature*. :( If you have more than 512MB on board (since at this point
the PPro and PII use the same chipsets most large MBs will go to 1024MB)
with either a single or dual PII your dead in the water.
When writing to memory above the 512MB mark the write will be
mirrored to a location below 512, so if one were to stash some data from
513 - 516, guess what, you just wiped the kernel with a mirrored write in
the first 4 megs. :( The only way around this is to disable L2 cache
totally, or (if bios supports it) disable L2 cache for 513 on. Either
way, past 512 is useless, either you crash or you reduce system speed to
486 levels.
On a benchmark note, for me the PII is ~10% slower at the same
clock rate as a PPro, however since the PPro200 is the top, and the PII300
is out the point is moot. The fastest PII is faster than the fastest
PPro.
So, is the PII great, I don't think so. It does work however.
As long as you NEVER plan on using more than 512MB (a major issue
for some systems) the PII is fine, if you need 512+, oh well, Thankyou
Intel.
----------------------------------------------------------------------------
Christopher E. Brown (CB421) <cbrown@alaska.net> +(907) 357-5680
We are Pentium of Borg. Division is futile. You will be approximated.
----------------------------------------------------------------------------