I apologize, I guess my lazy writing style didn't properly clarify my point.
Never, was I arguing your point that increased production volume will come from producing more wafers. We agree here.
What I was disputing was your statement," It never says they getting better yields.". When in fact, the Silicon Strategies' article, as reported by AI, did in fact say yields were increasing. Subsequently, other web articles quoting IBM senior vice president and chief financial officer Mark Loughridge have confirmed his statement to Silicon Strategies, that IBM has improved yields.
Again, sorry for my slopping writing style for any confusion.
I'm a little confused. I just want to understand the process of making the CPUs a little better. So on one wafer more than one process can be produced e.g. 130 vs. 90. Also can more than one cpu core be made 970 vs. some other cpu. That's what I've picked up from some of the posts. If anyone can explain this to me, or give me a link the would explain, that would be great. Thanks
I apologize, I guess my lazy writing style didn't properly clarify my point.
Never, was I arguing your point that increased production volume will come from producing more wafers. We agree here.
What I was disputing was your statement," It never says they getting better yields.". When in fact, the Silicon Strategies' article, as reported by AI, did in fact say yields were increasing. Subsequently, other web articles quoting IBM senior vice president and chief financial officer Mark Loughridge have confirmed his statement to Silicon Strategies, that IBM has improved yields.
Again, sorry for my slopping writing style for any confusion.
It's not your writing it's the mixed terminology again. I'm as much at fault as anyone. It happened when we were talking about dual core vs. dual processors. When in actuality Dual core chips have dual processors in one chip.
It's the same as yields, and yields. Until they, or we have an exact definition to decipher what a yield is in these cases it's going to be hard to determine which is which.
We should start using "wafer yield", and "product yield", or something to make the two more individual when talking about them. Right now we all speak like the blind mice of the mouth. Who's on first (.), or (?)
Just 2 cents worth ... IBM's Fishkill Foundry is primarily a 300mm wafer plant. So it may be safe to assume that the Fishkill plant is ramping-up production (alot of you may be saying DUH!!). Please read Programmer's post to clarify about ramping-up. It's important here. The 300mm wafer ramp-up could be very good news, and the lack of specific information could be intentional. <Read, Apple decreed the announcement to be vague>
Personally, I'm hoping for the 970mp, but will take increased production of all the processors.
Tis a good sign, wish it was sooner, but as was pointed out earlier ... it would be a waste to gripe about it.
We should start using "wafer yield", and "product yield"
Could you define those two for me? My definition of yields (in the context of microprocessor production) is the percentage of working processors at the target clock rate from the total number of fabricated processors that are cut from the wafer. If they image 100 onto a wafer and testing results in 5 working processors, that is a yield of 5%. This is complicated by the fact that some processors may simply not work at higher speeds, but may work acceptably at lower speeds. Unless otherwise specified the current yield is the total percentage of working processors at the minimum speed or above. I'm guessing this is what you mean by "wafer yield"?
The total 300mm, 90nm, 970 *wafer yields* at Fishkill is about 67%, with the following *product yields*:
>2.5 GHz 0.5%
2.5 GHz 2%
2.3 GHz 3%
2.0 GHz 8%
1.8 GHz 16%
1.6 GHz 40%
Lower GHz or Unusable 30.5%
[example with fictional percentages]
Okay, that makes sense but might I suggest we call a spade a spade and use the term "frequency yield" or "clockrate yield"? Product yield is a bit ambiguous.
Again. The reason we need to get more specific on our terminology. . You guy's set it up, and I'll start using it. I'm waaay to tired to think coherently enough to get into this right now.
Comments
Originally posted by Yevgeny
When I get headless G5 for $300?
Hey, that's Wizard69s part
I apologize, I guess my lazy writing style didn't properly clarify my point.
Never, was I arguing your point that increased production volume will come from producing more wafers. We agree here.
What I was disputing was your statement," It never says they getting better yields.". When in fact, the Silicon Strategies' article, as reported by AI, did in fact say yields were increasing. Subsequently, other web articles quoting IBM senior vice president and chief financial officer Mark Loughridge have confirmed his statement to Silicon Strategies, that IBM has improved yields.
Again, sorry for my slopping writing style for any confusion.
Originally posted by Kickaha
Yup.
I'm a little confused. I just want to understand the process of making the CPUs a little better. So on one wafer more than one process can be produced e.g. 130 vs. 90. Also can more than one cpu core be made 970 vs. some other cpu. That's what I've picked up from some of the posts. If anyone can explain this to me, or give me a link the would explain, that would be great. Thanks
Galen
Originally posted by rickag
Onlooker
I apologize, I guess my lazy writing style didn't properly clarify my point.
Never, was I arguing your point that increased production volume will come from producing more wafers. We agree here.
What I was disputing was your statement," It never says they getting better yields.". When in fact, the Silicon Strategies' article, as reported by AI, did in fact say yields were increasing. Subsequently, other web articles quoting IBM senior vice president and chief financial officer Mark Loughridge have confirmed his statement to Silicon Strategies, that IBM has improved yields.
Again, sorry for my slopping writing style for any confusion.
It's not your writing it's the mixed terminology again. I'm as much at fault as anyone. It happened when we were talking about dual core vs. dual processors. When in actuality Dual core chips have dual processors in one chip.
It's the same as yields, and yields. Until they, or we have an exact definition to decipher what a yield is in these cases it's going to be hard to determine which is which.
We should start using "wafer yield", and "product yield", or something to make the two more individual when talking about them. Right now we all speak like the blind mice of the mouth.
Personally, I'm hoping for the 970mp, but will take increased production of all the processors.
Tis a good sign, wish it was sooner, but as was pointed out earlier ... it would be a waste to gripe about it.
Originally posted by onlooker
We should start using "wafer yield", and "product yield"
Could you define those two for me? My definition of yields (in the context of microprocessor production) is the percentage of working processors at the target clock rate from the total number of fabricated processors that are cut from the wafer. If they image 100 onto a wafer and testing results in 5 working processors, that is a yield of 5%. This is complicated by the fact that some processors may simply not work at higher speeds, but may work acceptably at lower speeds. Unless otherwise specified the current yield is the total percentage of working processors at the minimum speed or above. I'm guessing this is what you mean by "wafer yield"?
The total 300mm, 90nm, 970 *wafer yields* at Fishkill is about 67%, with the following *product yields*:
>2.5 GHz 0.5%
2.5 GHz 2%
2.3 GHz 3%
2.0 GHz 8%
1.8 GHz 16%
1.6 GHz 40%
Lower GHz or Unusable 30.5%
[example with fictional percentages]
?
Originally posted by Kendoka
Something like this ?
The total 300mm, 90nm, 970 *wafer yields* at Fishkill is about 67%, with the following *product yields*:
>2.5 GHz 0.5%
2.5 GHz 2%
2.3 GHz 3%
2.0 GHz 8%
1.8 GHz 16%
1.6 GHz 40%
Lower GHz or Unusable 30.5%
[example with fictional percentages]
Okay, that makes sense but might I suggest we call a spade a spade and use the term "frequency yield" or "clockrate yield"? Product yield is a bit ambiguous.