65 nano yields well below 50% for most fabs..? (!)

[the_vorlon]

http://www.fabtech.org/content/view/2462/

The blog is written by Semiconductor Fabtech, Editor-in-Chief, Mark Osborne. He has been covering the semiconductor industry for over ten years. This area is intended to provide insight into topical stories of the time in a more informal manner.

Although Intel Corp. plans to start volume production at the 45nm node later this year, the leading-edge node for the rest of the CMOS logic community will be 65nm. This simple statement lays the foundation for a host of misconceptions about the 65nm node having been ‘cracked,' and that volume production at high yields is a given.

It may come as shock to some that the vast majority of chip manufacturers currently ramping 65nm processes - including some of the major foundries - have less than 50 percent yields!

This statement was issued by John Kispert, President and COO of KLA-Tencor. If anyone should have a good grasp of the bigger picture, it should be John Kispert, head of the largest supplier of metrology solutions in the industry.

 

[BaronMatrix]

hat was an interesting article. The thing I tok away from it ws that AMD and Intel are at the highest yield as they are the only companies producing 65nm in volume.

I do believe that Brisbane taped out more than 8 months ago. Tht should hae given them enugh time to make volume production worth it.

I don't really know what the breakeven point is but I would thnk that AMD is no trying to throw away money so I would say they are getting close to being above 50% as is Intel.

I wonder if they predict the same thing for 45nm.

 

[the_vorlon]

AMD and Intel are at the highest yield as they are the only companies producing 65nm in volume.

You follow AMD more closely than I.. when do they plan to start shipping 65 nano in volume?

:lol: :lol: 8O :twisted: :lol:

But seriously...

Don't want to start a flame war with AMD's #1 advocate.. but....

Two factors indicate strongly to me that AMD's 65 nano still has not fully ramped..

1) - The highest speed grades (5600+, 6000+ FX7x) are all still on 90 nano. Intel had the exact same issue when the rampled 90 nano, and for a while there the fastest pasts were still on 130 nano (the 3.4 Northwoods at the time) - If yield were good, these parts would be on 65 nanos - they are not.

2) - AMD did a very "loose" shrink to 65 nano. If you look at Brisbane, which is virtually identical in terms of transistors to the 90 nano version, it is about 129 mm^2 versus 180 mm^2 for the old 90 nano core. This is a drop of just a shade over 30% in die size versus the +/-48% you wold expect from a "perfect" optical shrink.

AMD clearly had to fairly dramatically relax their layout rules to get 65 nanos to work.

I think if you look at these two facts, plus the still scarce supply of 65 nano AMD parts, and it's pretty clear AMDs 65 nano isn't there yet. I expect they will get there, just as they got 90 nano working very well in the end, I just don't think that they are there yet.

 

[thexder1]

I agree that AMD is not there yet. I know from reading about the industry that the way AMD does it is they will tweak a manufacturing process as they are using it and eventually get it as efficient as possible before going to the next which may not be quite as good initially but after they have tweaked it for a while we will start to see it really take off.

The way Intel does it is they develop a manufacturing process and stick to it not tweaking it like AMD does to get it more efficient. I would assume that Intel does make some changes to the process as they go along but not like AMD does.

It would not surprise me that the yields are that low but they should be picking up as I said soon but I think that Intel is rushing a little too much for the 45 nanometer because I would not be surprised if they are going to have very low yields on it which will make it pretty expensive for them to do it and it would make it expensive to get the CPUs. I think that they may be trying to make up for their failure with netburst

 

[jkflipflop98]

If one of Intel's fabs dropped to 50% yields, the company would fire everyone working in it.

 

[BaronMatrix]

AMD and Intel are at the highest yield as they are the only companies producing 65nm in volume.

You follow AMD more closely than I.. when do they plan to start shipping 65 nano in volume?

:lol: :lol: 8O :twisted: :lol:

But seriously...

Don't want to start a flame war with AMD's #1 advocate.. but....

Two factors indicate strongly to me that AMD's 65 nano still has not fully ramped..

1) - The highest speed grades (5600+, 6000+ FX7x) are all still on 90 nano. Intel had the exact same issue when the rampled 90 nano, and for a while there the fastest pasts were still on 130 nano (the 3.4 Northwoods at the time) - If yield were good, these parts would be on 65 nanos - they are not.

2) - AMD did a very "loose" shrink to 65 nano. If you look at Brisbane, which is virtually identical in terms of transistors to the 90 nano version, it is about 129 mm^2 versus 180 mm^2 for the old 90 nano core. This is a drop of just a shade over 30% in die size versus the +/-48% you wold expect from a "perfect" optical shrink.

AMD clearly had to fairly dramatically relax their layout rules to get 65 nanos to work.

I think if you look at these two facts, plus the still scarce supply of 65 nano AMD parts, and it's pretty clear AMDs 65 nano isn't there yet. I expect they will get there, just as they got 90 nano working very well in the end, I just don't think that they are there yet.


Tha's not the way to look at it. AMD has 3 Fabs. One is producing 65nm and if the word from AMD is correct it is up to 10000 wspm. Fab 7 Chartered is at 3000 so that's about 1/3 of Fab 36(not counting process size differences).

As soon as Fab 36 isat 20000 WSPM, then you will see Brisbane moved to Chartered and Fab 36 will do mainly Barcelona.

WHen Fab 38 gets 300mm wafers they can shut down around 1/3 of the Fab for 45nm OR 65nm equipment.

I mean just like Intel couldn't just stop making NetBurst, AMD can't just stop making 90nm. I mean if Core2 was just equal to X2 then none of you wld have a problem.
I just hope that after Barcelona AMD and Intel stay within 5% of each other. Then you guys can have wars about somethng else.

Leave my PC alone!

 

[the_vorlon]

Tha's not the way to look at it. AMD has 3 Fabs. One is producing 65nm and if the word from AMD is correct it is up to 10000 wspm. Fab 7 Chartered is at 3000 so that's about 1/3 of Fab 36(not counting process size differences).

As soon as Fab 36 isat 20000 WSPM, then you will see Brisbane moved to Chartered and Fab 36 will do mainly Barcelona.

WHen Fab 38 gets 300mm wafers they can shut down around 1/3 of the Fab for 45nm OR 65nm equipment.

I mean just like Intel couldn't just stop making NetBurst, AMD can't just stop making 90nm. I mean if Core2 was just equal to X2 then none of you wld have a problem.
I just hope that after Barcelona AMD and Intel stay within 5% of each other. Then you guys can have wars about somethng else.

Leave my PC alone!

You did a good job of NOT answering my contentions.

If yields are good, why no high speed grade parts? - If it works, it works on 1 fab or 20 fabs.. There are no high speed grade parts on 65 nano. If the process is working, why are all the high speed bins still on 90 nano?

AMD got only a +/- 30% die size reduction versus +/- 48% for a perfect shrink (FWI - factoring cache size and logic changes, Intel's Prenym shinks 47% going from 65 to 45 BTW) - If the AMD 65 nano process is working, why did they have to relax the layout rules so much to get it to work?

 

[darious00777]

Could be because the current architecture was designed two process generations ago. Just a thought.

 

[CaptRobertApril]

If one of Intel's fabs dropped to 50% yields, the company would fire everyone working in it.

:trophy:

It seems you're winning a lot of these today. All well deserved!

 

[WR]

That statement about yields shouldn't come as a surprise. 65nm is new to most of the industry who were on 80 or 90nm just a few months ago. Intel ramped up a year ago because of aggressive in-house research - the report doesn't cover them. AMD started to ramp up recently and is probably getting decent yields now because of their continual process refinement and the fact that Brisbanes are generally in stock at retail (report also doesn't cover them, only Chartered).

If the majority of the industry had 65nm mature, then it would be an embarassment for the current Cell processor to be at 90nm, or for the vast majority of DRAM modules to consist of 80 or 90nm chips.

 

[the_vorlon]

If one of Intel's fabs dropped to 50% yields, the company would fire everyone working in it.

If one of Intel's fabs dropped to 50% yields, the company would shoot everyone working in it.

 

[RandMcnally]

I wouldn't find it hard to believe amd's 65nm sucks. I mean that lightly, but I think that this was truly the last leg that AMD could go with out immersion, notice the abnormally large die size. This process sounds like it's sucking for AMD. PR team at work here. (not supporting intel, or bashing AMD, this has just been my thought since AMD made such a fuss about 45 nm that they were not happy with 65nm, trying to divert people's attention.) EDIT:But this is another article based on 100% speculation considering I don't know a fab company that publicly releases yields.

 

[BaronMatrix]

Tha's not the way to look at it. AMD has 3 Fabs. One is producing 65nm and if the word from AMD is correct it is up to 10000 wspm. Fab 7 Chartered is at 3000 so that's about 1/3 of Fab 36(not counting process size differences).

As soon as Fab 36 isat 20000 WSPM, then you will see Brisbane moved to Chartered and Fab 36 will do mainly Barcelona.

WHen Fab 38 gets 300mm wafers they can shut down around 1/3 of the Fab for 45nm OR 65nm equipment.

I mean just like Intel couldn't just stop making NetBurst, AMD can't just stop making 90nm. I mean if Core2 was just equal to X2 then none of you wld have a problem.
I just hope that after Barcelona AMD and Intel stay within 5% of each other. Then you guys can have wars about somethng else.

Leave my PC alone!

You did a good job of NOT answering my contentions.

If yields are good, why no high speed grade parts? - If it works, it works on 1 fab or 20 fabs.. There are no high speed grade parts on 65 nano. If the process is working, why are all the high speed bins still on 90 nano?

AMD got only a +/- 30% die size reduction versus +/- 48% for a perfect shrink (FWI - factoring cache size and logic changes, Intel's Prenym shinks 47% going from 65 to 45 BTW) - If the AMD 65 nano process is working, why did they have to relax the layout rules so much to get it to work?


First, how do you know why the chip isn't half the size? Second, why not use the proven, tested process to get the higher speeds. Besides, with all of the 5200+/5400+/5600+/6000+ there is only 400MHz difference (and for the most part only 200MHz).

There is a Brisbane 4800+ and the 5000+ is due soon.


As far as why they are holding off on higher clocked parts I would say it's because Fab36 is only ONE Fab and they have to start ramping Barcelona.

That means keeping more volume at 90nm. Again, if Core 2 was equal rather than faster you PROBABLY wouldn't say any of this.

 

[BaronMatrix]

Because AMD has published the specs, the 512x2 Windsor core is 183 mm^2, the 512x2 Brisbane is 126 mm^2, this is less than 1/2

Links:
EDIT( Ooops pressed submit before preview Smile )
http://www.techarp.com/showarticle.aspx?artno=337
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2893&p=2

I really liked the line they fed TR:
Quote:
On the subject of the die size of current 65nm parts, they say the relatively small reduction in die area from 90nm to 65nm is not the result of added L2 cache being placed into silicon and then deactivated in the 4800+ and 5000+ models we tested. Instead, the modest reduction in die size has origins in the esoterica of process technology and AMD's model of continuous, gradual improvement to its manufacturing techniques. We've asked for additional detail on this subject, but that's all we can report at this time.


http://www.techreport.com/onearticle.x/11490

You are so easily fooled. Mostly because I believe you do not understand.

Second, why not use the smaller more cost effective process to maximize your margins... this point is just plain funny.... proving yet again you know nothing about the dynamic or forces behind driving this industry.

And here I was hoping you wouldn't be back. That still doesn't tell me how Vorlon knew why it wasn't half. As soon as you come around it's like "the a-hole of the century who has made several different CPU SKUs" is back.

I could have continued paying for those white papers but I'm a SW developer. I am not fooled because I don't care that much. That's why you don't like me. I am only impressed by the people who WRITE THE WHITE PAPERS, not wannabe's who pay for them to impress the Brood.


Jumping Jack, King of The Brood. Don't mess with him, he'll post a link to someone else's words.

 

[tamalero]

http://www.fabtech.org/content/view/2462/

The blog is written by Semiconductor Fabtech, Editor-in-Chief, Mark Osborne. He has been covering the semiconductor industry for over ten years. This area is intended to provide insight into topical stories of the time in a more informal manner.

Although Intel Corp. plans to start volume production at the 45nm node later this year, the leading-edge node for the rest of the CMOS logic community will be 65nm. This simple statement lays the foundation for a host of misconceptions about the 65nm node having been ‘cracked,' and that volume production at high yields is a given.

It may come as shock to some that the vast majority of chip manufacturers currently ramping 65nm processes - including some of the major foundries - have less than 50 percent yields!

This statement was issued by John Kispert, President and COO of KLA-Tencor. If anyone should have a good grasp of the bigger picture, it should be John Kispert, head of the largest supplier of metrology solutions in the industry.

I remember reading in fabtech about Chartered going way over the 40%
yield on 65nm from AMD around middle of last year.. so I wonder.. is that report acurate?
you think they lied?
what do you think Jack?

 

[pausert20]

If the majority of the industry had 65nm mature, then it would be an embarassment for the current Cell processor to be at 90nm, or for the vast majority of DRAM modules to consist of 80 or 90nm chips.

The Cell processor is being manufactured at 65nm or so says the link below.

ExtremeTech Link

For now, Sony is already cranking out chips on a 65nm production process. The shift to 65nm production allows Sony to produce chips that are 40% smaller than their 90nm counterparts. The move could also be beneficial in promoting a price cut for the PS3 which has gotten off to a slow start in North America.

DailyTech Link