🚀 go-pugleaf

In article <1ht90eINNkiboei0@hpscit.sc.hp.com> matthias@nsr.hp.com (Matthias Kamm) writes:

>This is very possibly wrong.  If this were the case, then the cost of test for
>the 16Mhz chip would be three times that of the 33Mhz chip (assuming 33,25,16
>speed grades for their microprocessors).  Every chip test insertion
>cost bucks, and the 25Mhz chip with two test insertions would cost Intel MORE
>than the 33Mhz chip!  One thing we used to do (when I worked at a semi. co.)
>was to test all chips to the highest speed grade, then mark the chips depending
>on the demand.  This only works if there is extremely high yield at the highest
>speed grade.  In Intel's case, they are probably yielding 100% at 33Mhz for the
>486 (for working chips). 


Well, the way I look at it is what is the opportunity cost of not
testing a 33MHz failure at 25 MHz? You could 1) throw it away or
2) test it at 25 MHz and sell it if it passes. 

Its a matter of which would be cheaper: to make another 25 MHz chip
from scratch _and_ test it or to test the reject. Testing the reject 
is cheaper, usually, depending on how many of the 33 MHz failures 
would pass at 25 MHz. 

Of course, if as you say, their yield is close to 100%, or the failures
at 33 MHz don't have much chance at passing at 25 MHz it would make sense
to just mark some of the 33 MHz parts as 25 MHz.





-- 
-------------------------------------------------------------------------------
Iskandar Taib                          | The only thing worse than Peach ala
Internet: NTAIB@SILVER.UCS.INDIANA.EDU |    Frog is Frog ala Peach
Bitnet:   NTAIB@IUBACS                 !