> > Interesting... our SS10 definitely wedge solid, the processors by
> > themselves were ok.  As I said, I knew the mbus was supposed to allow
> > it but the official word from sun was "not tested".
>
> Indeed, there's a big difference between "supposed to work" and "known
> to be stable and supportable".

It's easy to figure this out - simply look at whether executables will
ever have instructions which may not be on both processors, like SSE, and
if not, then there's nothing to worry about. Every Pentium Pro core (PPro,
P2, P3, all Celerons) has identical user mode binary compatibility; with
the exception of the SSE stuff added for the purpose of marketing, there
is no difference at all once the CPUs are initialised.

> Intel's periodic specification update documents (available from
> developer.intel.com for each of their processor lines) have a "MP
> compatibility matrix" for each stepping and speed.  I'm not sure why
> they bother displaying the full matrix since they only call matched
> speeds "supported".
>
> For example, see http://developer.intel.com/design/PentiumIII/specupdt/

Intel is not a company that makes good products and then sells them;
rather, they make weak products, then spend their energy selling them.
The business people have much more say and control than the scientists.
Therefore, I would not care much about anything that Intel has to say
about whether their processors will work in various combinations.

Remember, after all, that, according to Intel, no Celeron will work in a
dual processor setup.

> Note that, contrary to previous postings on this thread, mixed
> steppings *are* supported; it's just mixed *speeds* which are not
> tested (and therefore should not be assumed to work).

I disagree; they should be assumed to work. The clock speed of the core of
the processor is irrelevant; nothing happens between the motherboard and
the processor at that speed. Only the bus speed is important.

I've had different speed processors in dual CPU systems for years. Doesn't
that qualify as testing?

After all, we're not Radio Shack employees saying, "that cable isn't
tested to work with anything but Windows", or some ISP saying, "we only
support Windows and MacOS". We're not ignorant trained monkeys spitting
out some mantra about the only thing we're trained about.

Therefore, it'd be irresponsible to pretend we didn't know that they work.
We don't have any sort of liability to worry about if some user cannot get
his or her setup to work the same way we claim it should (after all, some
BIOSes might not like it).

We're NetBSD, and we should be happy to say that we support whatever is
thrown our way.

John Klos
Sixgirls Computing Labs