Cell yields 'horrible'

DarkRage

Newcomer
Well, coming from the Inquirer... who knows?

Anybody can share some information about it? 90nm confirmed? Maybe 65nm and having problems?

"SOURCES CLAIM that yields of the Cell microprocessor for the Playstation 3 are not good. A process mole tells us that the Cell yields are far enough below par that the amounts of good dies off the test wafers are "horrible".
The problem is that IBM and Sony designed the chip for some very aggressive process technologies that did not materialise in the way the fab boys were expecting. Because of this, we are told Sony is scrambling to get yields up before it has to make them for real, a fast approaching deadline.

Will it get it done? Who knows. Will it be able to make enough to satisfy initial demand? Not a chance. Then again, in consoles, what else is new?µ"

http://www.theinquirer.net/?article=31443
 
!eVo!-X Ant UK said:
Did'nt IBM say months back that have better yeilds than expected?

No, I recall what they said was that yields were improving better than expected.

By strictly hypothetical example:

If yields improve from 24% to 25% after 1 week of work, if the engineers expected it to take 3 weeks to get that improvement, then yields are "improving better than expected".

However a yield of 25% would still be considered "bad".

Of course it's the Inquirer, so it's probably full of crap anyway.
 
Last edited by a moderator:
If this mole exists and speaks the truth, I would guess it's about getting 65nm working. There shouldn't be any problems with 90nm, especially not after a year+ of production using well established methods.
 
Shifty Geezer said:
If this mole exists and speaks the truth, I would guess it's about getting 65nm working. There shouldn't be any problems with 90nm, especially not after a year+ of production using well established methods.

That would make sense, it probably is the 65 nm process.
EDIT: On the other hand, they may be going for a tougher power envelope for the final production.

This was said in february:"Speaking to Warren Communication News, IBM senior vice president William Zeittler said that the firm is learning how to improve yields of the chip "faster than on any chip we've done.""

http://www.gamesindustry.biz/news.php?aid=14679
 
Last edited by a moderator:
Again, I cannot speak to CELL yields, but traditionally, IBM has been, shall we say, less than impressive with their yields. So an improvement, can only be, well an improvement on their previous attempts. In the end, I don't see how they could really have "bad" yields, if someone is talking about for specific devices I could understand that. But how many SPEs are going to be needed for things like Toshibas TVs, 2, 3, 4? If they don't require all of the SPEs, similar to the PS3, its hard to imagine having bad yields.
 
The 65nm theory is a good one, and would bare out the useage of 'test' wafers, since Cell on 90nm has been being produced 'for real' for some time now. Who knows though, IBM is known to be bad on the yields, but I feel this news would be a little late in coming if we were talking about 90nm. Not that it still might not be - who knows...
 
DarkRage said:
Because of this, we are told Sony is scrambling to get yields up before it has to make them for real, a fast approaching deadline.

Hate to be the one to ask, but just to confirm. If this is regarding the 65nm, which is my opinion also, this fast approaching deadline they're talking about would most likely be 2007, correct, and this would also confirm that 90nm definitely will be in the initial consoles?

(Of course, if this news pertained to the 90nm process, it's a no-brainer that the initial units will have 90nm, but maybe a PS3 in November would be a bit optimistic.)

-aldo
 
aldo said:
Hate to be the one to ask, but just to confirm. If this is regarding the 65nm, which is my opinion also, this fast approaching deadline they're talking about would most likely be 2007, correct, and this would also confirm that 90nm definitely will be in the initial consoles?

(Of course, if this news pertained to the 90nm process, it's a no-brainer that the initial units will have 90nm, but maybe a PS3 in November would be a bit optimistic.)

-aldo
Sony confirmed last week (in CED) that they were launching at 90nm:

amillians said:
It's been a foregone conclusion for some time, but Sony has finally formally confirmed that the 1st generation PS3 will be 90nm. 2nd generation 65nm probably isn't going to happen until late 2007...the lines are on the floor, but they still haven't taken the plastic wrap off of them. Sony's IR department also confirmed that their projected PS3 manufacturing losses "will exceed" JPY100,000,000,000, above and beyond the hit they've already taken on R&D and operations...when pressed if PS3 manufacturing losses would top overall PS2 losses, the IR exec said no comment. They also reconfirmed 6,000,000 units shipped by March 2007

Link
 
I'm not an AVS guy, so although I do see said members assertion (and I'm sure he has all the credibility in the world), I still have to ask where he got his 90nm confirmation from. :)
 
NucNavST3 said:
but traditionally, IBM has been, shall we say, less than impressive with their yields.
Traditionally schmaditionally.

IBM is a huge semiconductor manufacturer, what you're doing is exaggerating and extrapolating a few examples that have made the rounds in computing media and trying to establish some kind of pattern from it. That's not valid reasoning.
 
Mmmkay said:
Consumer Electronics Daily (CED)
http://www.warren-news.com/CEservices.htm

amillians has been a consistent source of CED news.

Hey like I said I wasn't doubting him, I generally trust certain cross-posts on faith alone if they 'feel' right; but at the same time I was just curious what his original source was.

If it's 90nm then it's 90nm!

Let me say though, that the manufacturing loss figure is very similar (in $'s) to the estimated losses of the SCE division overall in the present fiscal year. Makes me wonder if CED got their news from the Sony conference call, and if so, did they mistake division losses for manufacturing losses? If so, we may be able to listen to the conference call and hear 90nm confirmed and perhaps some other interesting tidbits as well...
 
assen said:
If 65 nm is expected in late 2007, that would mean the Inquirer info is about 90 nm, right?
No reason to think that. The Inq just floated a story with a shocking, eye-catching headline, and no intent to explain it to any degree. They probably don't understand it themselves but being the non-journalists they are, they don't waste time getting the facts but spout the rumour and rake in the hits.
 
Shifty Geezer said:
No reason to think that. The Inq just floated a story with a shocking, eye-catching headline, and no intent to explain it to any degree. They probably don't understand it themselves but being the non-journalists they are, they don't waste time getting the facts but spout the rumour and rake in the hits.

I agree but i don't know why anyone would characterize yields on the 65nm process for any chip as 'horrible', given how early it is in the migration to that process. So while they may be sensationalizing the news, i dont see where they could have even gotten such a quote in regards to 65nm?
 
Back
Top