[RC5] Adverse effects of participation?

Zypher rc5 at planetfortress.com
Wed Feb 2 14:03:07 EST 2000


The part that is *new* or *newer* is that it rarely fails QA today...they
(intel and amd) make TOO MANY perfectly good high end chips.

If 650 is etched on the core (for Athlons) it means it DID pass the QA as/at
650, but so many did they needed to sell some 500s. When they fail, they
tend to fail big, and crap out completely.

Another case can be made on testing conditions...a lot of cpus are 'rated'
with no heatsink or a REALLY pathetic one (a couple grams of aluminum, oh
my) at 70-90C, and ambient temp 50C or higher. I doubt the air temp goes
past 150F in most of our homes/businesses. And when it does, the money gets
spent on nice controlled climate industrial racks.

Memory: I doubt they do this anymore, because the difference between
(64mbit) parts is number of 64mbit chips on pcb....it would sure look funny
if two same part # 128mb sdrams from same oem had different amounts of
chips, ie onesided and twosided.

-------------------------------------------------
rc5 at planetfortress.com
www.planetfortress.com/rc5
Join our Gamma Flux team :)
...NOW damnit!
-------------------------------------------------

> >Time to clear up another one: Often overclocking your cpu really
> >isn't...more often nowadays you're buying an *underclocked* cpu ;) For
> >example, the wonderful athlon. If you grab a 500 made in the last few
> >months, and crack the case, the core will actually have been tested to
600
> >or 650. (Its etched on the core, so I'd say thats proof enough) The
> >production is TOO good, they lower enough bins so they can get more
lowend
> >parts, and not screw up their pricing either. Some 550s out in the last
few
> >weeks have actually been 0.18 micron cores...the same ones on the Stock
> 750s
> >and 800s.
>
> This is a rather old practice, but it is often/usually because the part in
> question would not pass the QA tests at the rated speed, so they bump it
> down until it does pass.  They do the same thing with memory, fencing off
> half of a chip that fails, and selling it at the lower level.


--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list