🚀 go-pugleaf

Michael Rios (mikerios@ix.netcom.com) wrote:

: Not even close.  Chips marked with a lower speed in most cases were
: never tested at higher speeds.  The mfr tests at the speed they need to

This would seem to make sense, as possibly destructive testing on every
unit would probably be very bad. ;)

: of the orders are for 20, they'll get marked 20.  Recently Intel ran
: out of 486sx33's.  So they took a bunch of sx25's, and remarked them. 
: The -25 is in the center, the -33 at the bottom.  They sent a letter
: saying that these chips were in every way equal to ones that were
: marked -33 to begin with.  Of course, we knew that; we'd been running

Not that I don't believe you, because I do, but do you have any documented
information about this?  I find it very interesting, and it doesn't surprise
me at all. 

: -25's and one -33 that wouldn't clock up to 40 stably.  And we've never
: had one fail in the field.  (we're talking hundreds, maybe thousands of
: units).

I am currently overclocking my P-90 to 100.  Since on my Triton MB it is just
the difference of moving one jumper I hardly think the MB itself will have
problems.  The CPU has a heatsink/fan combo thingy on it, of course, but it
looks kind of cheap.  I haven't had the slightest problem with it, but in
an earlier post I saw someone (from DEC?) say that they knew someone that had
two of them fail in 2 months.  Can anyone tell me if it is possible for every-
thing to be working fine with no warnings and then, suddenly, have the CPU
catastrophically fail?  This is my only real concern.  Since it is a modern
P-90 (no FDIV bug) I'm pretty sure it is made identically to the P-100 on the
same process.  If you can't tell already, I'm still a little nervous about
this, but the difference seems so small that I think it should be OK. However, 
there is certainly a nice speed difference!  I resisted the urge to try 120,
133, or 150. ;)  They are, after all, on a .35 micron process, and mine isn't.
Please, no opinions!  I've had enough of those already, although I appreciate
them.  I want someone who KNOWS!!  ;)

: this would apply to a very small # of chips.  The vast majority are
: never *tested* at a speed higher than the mark.

So, if a given chip passes some Intel test on a given process and speed it is
shipped at that clock speed?  Would there be any way to duplicate that test
outside of the fab plant just with software?

Knowledgeable replies gratefully accepted!

							--Jonathan