24 Jan
2013
24 Jan
'13
2:12 p.m.
On 1/24/2013 12:49 PM, Ethan Dicks wrote:
On Thu, Jan 24, 2013 at 3:36 PM, David Riley
<fraveydank at gmail.com> wrote:
I've seen several
references to laser cutting of locations on dies. I
scrapped 10 systems which were built up by a company to do on die
programming, which consisted of a 2 ton slab of optical grade surfaced
granite, a ND laser to zap the dies, and HP micro position sensors, and
other piezo positioners as well as macro positioners to move the dies.
I think the comment was there was 300k in each unit, and they were
"free" if we just would move them by the end of the month.
This was 300k in 2000 era gear, (PC and some embedded running it) so in
80's era scale it would have been more. I wonder how many chips they'd
have to have that were bad to make the economics work, investing say
500k to sell scrap silicon? at minimum. who knows what the NRE was on
either the systems I had or the cost of doing it in the 80's w/o cheap
PCs, probably PDP11 or other workstation grade controlling systems to
develop software for.
On Jan 24, 2013, at 1:42 PM, Fred Cisin <cisin
at xenosoft.com> wrote:
He may be talking about the AMD
second-sourced 486SX, which may have
had a different method of disabling the FPU. My recollection is that
the 486SX was, in fact, the same die as the 486DX (FPU and all) with
a fuse blown that disabled the FPU.
Back in the day, I heard the FPU was
"disconnected" with laser-drilled
holes on dice that failed FPU testing, then the package was labelled
to match. To be clear, a wafer was made up of dice that were intended
to be sold as 486DX chips but failed FPUs (and most likely enough ones
with good FPUs to cover customer demand) were pierced. debacle,
there was software thatcould
completely restore functionality.I think it was called above-board,
NOPE.
which had hardware as well as software,but on
many SX type systems it
was just software Actually if I remember correctly it was the 486SX CPUs
made by AMD
We are talking INTEL. I can't promise what I heard then was true, but it
was commonly
reported as such. It's possible that they did that at first, then
after a die shrink split the lines into straight 486DX and 486SX
wafers and discarded any that failed any test. At first, though, I
find it believable that yields were low enough to want to sell any
parts they could (since retail prices were many hundreds of dollars
per CPU back then).
-ethan