AMD chip yields even better at Chartered than at AMD!

[bbbl67]

http://www.fabtech.org/index.php?option=content&task=view&id=1390&Itemid=2

Yields of AMD chips above 50 percent at Chartered

AMD's first foundry partner, Chartered Semiconductor has already been
able to produce fully working die with yields equal to or even better
than those achieved at AMD, according to a March 15th report from Bear
Stearns.The reports authors said that yields were at 50 percent at
Chartered's Fab7 but sometimes even better than those achieved within
AMD's own fab! Furthermore, AMD would be receiving 2,000wspm by July
2006.

No reference was given to the process node the 50 percent yields were
attributed to. The Bear Stearns report also highlighted that Chartered
has brought forward some tool delivery dates to speed up Fab7
production ramp with 12,000wspm being possible in March and reaching
15,000wspm by mid-year.

The report estimated that about one-third of the revenues from Fab7
would come from the Xbox360 microprocessor orders, though these seem to
be flat since ramping last year, staying at 5 to 6,000wspm.

The delay in Sony's PS3 could boost Chartered's business with
IBM/Microsoft potentially adding more wafer runs this year, the report
stated. Broadcom is also cited as a customer of Fab7 using 110-113nm
processes.

 

[chrisv]

Yields of AMD chips above 50 percent at Chartered

Wow!

50 percent is suposed to be good??

 

[Del Cecchi]

bbbl67 wrote:




Wow!

50 percent is suposed to be good??

yes, sometimes. (dumb investment group trimmed)

 

[YKhan]

Wow!

50 percent is suposed to be good??

They're probably talking about 65nm rather than 90nm. AMD is likely
already above 90% on 90nm since it's a mature process.

Yousuf Khan

 

[The little lost angel]

Wow!

50 percent is suposed to be good??

Seeing that AMD hasn't made any noise about their own 65nm production,
I'll think that this is probably good for 65nm.

 

[Joe Pfeiffer]

Wow!

50 percent is suposed to be good??

I'm remembering somebody I did some consulting for a couple of decades
ago -- I helped put together a course in realtime programming for
their new processor. They were getting a bit concerned, as their
yield at the time was measured in wafers per chip, not chips per
wafer.

(I don't know whether they ever made it into full production)

 

[David Ball]

I'm remembering somebody I did some consulting for a couple of decades
ago -- I helped put together a course in realtime programming for
their new processor. They were getting a bit concerned, as their
yield at the time was measured in wafers per chip, not chips per
wafer.

(I don't know whether they ever made it into full production)

What size were wafers then and how many (attempted) chips would fit on
one wafer?

-- David

 

[Joe Pfeiffer]

What size were wafers then and how many (attempted) chips would fit on
one wafer?

Sorry, I frankly don't remember.

 

[Keith Williams]

Sorry, I frankly don't remember.

It's not unusual for a new process or product to have
good_chips/wafers < 1. It takes a while to tune things.

 

[Del Cecchi]

It's not unusual for a new process or product to have
good_chips/wafers < 1. It takes a while to tune things.

I for sure remember that being the case with a well known IBM processor
of a while back. Probably 200 mm wafers.

 

[Joe Pfeiffer]

It's not unusual for a new process or product to have
good_chips/wafers < 1. It takes a while to tune things.

Now I'm wishing I'd either remember more of the story before posting
or kept my mouth shut. What I do remember is that they were at a
stage of the game where they had been expecting a lot better yield by
then.

 

[Keith Williams]

Now I'm wishing I'd either remember more of the story before posting
or kept my mouth shut.

Nah, it's happened so many times one more example isn't going to
change physics. When you mentioned it I was thinking about how
many examples I could come up with. ...and ran out of digits. ;-)

What I do remember is that they were at a
stage of the game where they had been expecting a lot better yield by
then.

No one *expects* <1% yields. If they do, they certainly don't tell
the execs. ;-)

 

[George Macdonald]

It's not unusual for a new process or product to have
good_chips/wafers < 1. It takes a while to tune things.

So do they have a "search for the culprit" at that stage, like in software
projects?

 

[Keith Williams]

So do they have a "search for the culprit" at that stage, like in software
projects?

Sure, in many senses of "search for the culprit". ;-)

They do use different tools than programmers though. Since the
bugs are smaller the microscopes are much bigger. ;-) A friend was
in a failure analysis group and we used to talk over a brew about
what he was carving up to look see. It's impressive what they can
do with a few megabux worth of toys (and a lot of talent).

BTW, at this stage if we can get *one* good chip, we (the logic
folks) are happy. ...though that it finds its way to a customer
(proof of existence . Partially good chips are quite useful too.

 

[Del Cecchi]

Sure, in many senses of "search for the culprit". ;-)

They do use different tools than programmers though. Since the
bugs are smaller the microscopes are much bigger. ;-) A friend was
in a failure analysis group and we used to talk over a brew about
what he was carving up to look see. It's impressive what they can
do with a few megabux worth of toys (and a lot of talent).

BTW, at this stage if we can get *one* good chip, we (the logic
folks) are happy. ...though that it finds its way to a customer
(proof of existence . Partially good chips are quite useful too.

Especially if you have several that are all bad the same way.

FIB is my candidate for most amazing.

del

 

[Robert Redelmeier]

So do they have a "search for the culprit" at that stage,
like in software projects?

I believe the full order is:

1) Search for the guilty;
2) Punishment of the innocent;
3) Praise and high honors on non-participants.

This drives the denizens to avoid involvement,
and massive CYA when unsuccessful.

-- Robert

 

[George Macdonald]

Especially if you have several that are all bad the same way.

FIB is my candidate for most amazing.

FIB as in "semiconductor device editing or modifications"? It does sound
kinda "amazing".

 

[daytripper]

FIB as in "semiconductor device editing or modifications"? It does sound
kinda "amazing".

No doubt FIB has saved many chip designers careers ;-)

 

[Keith]

Especially if you have several that are all bad the same way.

We prefer them bad in different ways. That way we can put "together" one
chip that works. ...well, at least virtually.

FIB is my candidate for most amazing.

The one that got me was the *picture* of a leaking DRAM cell. They guy
sliced the die along the trench of the leaking cell and the photons ran
right out.

 

[Keith]

FIB as in "semiconductor device editing or modifications"? It does sound
kinda "amazing".

Yep, "Focused Ion Beam". Kinda like a small soldering station. Now
imagine drilling down from the back side of the chip to solder your wires
into the rats nest. Can't do it from the front, there might be ten layers
of wires you'd piss off there! It is amazing stuff, though I'm not sure
for how much longer.