GlobalFoundries outlines roadmap

[Idontcare]

Process roadmap

*Q1 2010. The company plans to ramp a 32-nm process based on SOI and a high-k/metal-gate scheme.

*Q2 2010. It plans to ramp a 45-/40-nm low-power process. The bulk technology will not include SOI or high-k.

*Q4 2010. It plans to offer a 28-nm generic bulk process, based on high-k. It will not include SOI.

*Q1 2011. It hopes to offer a 28-nm low-power bulk process, also based on high-k. The 28-nm processes will ramp in the Dresden fab. The so-called Module 2 plant is capable of making 25,000 wafers per month.

http://www.eetimes.com/news/se...cleID=218500937&pgno=2

The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

 

[alyarb]

pussies.

 

[taltamir]

4x nm?

 

[exar333]

I think he was referring to the following reference in a generic manner...

*Q2 2010. It plans to ramp a 45-/40-nm low-power process. The bulk technology will not include SOI or high-k."

 

[KingstonU]

What are AMD's 32nm plans? A die shrink of Shanghai/Deneb in Q2 2010?

Will bulldozer be on the 28nm process then since it's slated for 2011?

 

[Phynaz]

Originally posted by: alyarb
pussies.

?

 

[magreen]

Originally posted by: Idontcare
The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

That is interesting. I'm assuming no cpu foundry has done that before (in the recent past), making a separate process just for their energy efficient lines?

If so, then it seems you're right to suspect graphics/chipsets/other customers, and not power efficient cpus, which are usually cherry-picked samples of the regular cpus that can work at low voltages, if I'm not mistaken (pls correct me if I'm wrong)

 

[ilkhan]

32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

 

[magreen]

Originally posted by: ilkhan
32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

well, that depends how you define "ramping." I'm sure idc can fill us in, but they may have ramped the 45nm phII years before it was released this year.

 

[drizek]

Ya, im having a hard time believing that. I would love it if it is true though. By the end of 2010 they will be ahead of intel in manufacturing if they can get to 28nm.

 

[KingstonU]

Originally posted by: drizek
Ya, im having a hard time believing that. I would love it if it is true though. By the end of 2010 they will be ahead of intel in manufacturing if they can get to 28nm.

I don't think anyone would expect AMD to ever be ahead on manufacturing process over Intel. Has it ever been so in the past?

 

[drizek]

No, but they have been on par with intel for the most part. Now they are more than a year behind, but if they can get to 32nm in early 2010 then they will only be a few months behind.

They have much more incentive to improve their manufacturing now, because they are selling to both AMD and ATI divisions, as well as potentially nvidia and other TSMC customers.

 

[Idontcare]

Originally posted by: magreen

Originally posted by: ilkhan
32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

well, that depends how you define "ramping." I'm sure idc can fill us in, but they may have ramped the 45nm phII years before it was released this year.

Yeah remember this is GlobalFoundries process tech roadmap, not AMD's product roadmap.

I would expect at least a 3-6month lag between GlobalFoundries announcing the availability of a node versus a customer announcing having received samples from the node.

(consider tapeout to first silicon to debug/verification to respin and repeat before product is released takes around 6 months at best and first silicon can't be effectively generated until the node itself is nearly production ready)

 

[Atechie]

Originally posted by: drizek
No, but they have been on par with intel for the most part. Now they are more than a year behind, but if they can get to 32nm in early 2010 then they will only be a few months behind.

They have much more incentive to improve their manufacturing now, because they are selling to both AMD and ATI divisions, as well as potentially nvidia and other TSMC customers.

So being ~1 year behind is what you call "on par"?

 

[TemjinGold]

Originally posted by: Idontcare

Process roadmap

*Q1 2010. The company plans to ramp a 32-nm process based on SOI and a high-k/metal-gate scheme.

*Q2 2010. It plans to ramp a 45-/40-nm low-power process. The bulk technology will not include SOI or high-k.

*Q4 2010. It plans to offer a 28-nm generic bulk process, based on high-k. It will not include SOI.

*Q1 2011. It hopes to offer a 28-nm low-power bulk process, also based on high-k. The 28-nm processes will ramp in the Dresden fab. The so-called Module 2 plant is capable of making 25,000 wafers per month.

http://www.eetimes.com/news/se...cleID=218500937&pgno=2

The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

Plan... plan... plan... hope? If we dig further down the map, do we get pray?

 

[Fox5]

Originally posted by: magreen

Originally posted by: Idontcare
The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

That is interesting. I'm assuming no cpu foundry has done that before (in the recent past), making a separate process just for their energy efficient lines?

If so, then it seems you're right to suspect graphics/chipsets/other customers, and not power efficient cpus, which are usually cherry-picked samples of the regular cpus that can work at low voltages, if I'm not mistaken (pls correct me if I'm wrong)

They could probably do an atom competitor on that line. (scale the x2's down)
But it's not the first time a company has made a low-power process. AMD has typically specialized in high performance silicon, so this is just new to GF.

But AMD is ramping up very fast if this roadmap is true. Unexpectedly fast I'd say, when is Intel doing 32nm and 28nm?

Also, sounds like GF may have ditched SOI, and possibly high performance silicon all together.

 

[drizek]

I don't think GF ditched SOI. Their 32nm process will have it, and I assume they will bring it back for the next big step to 22nm. Hopefully that will be in or before Q4 2011.

 

[magreen]

Originally posted by: TemjinGold

Plan... plan... plan... hope? If we dig further down the map, do we get pray?

Dreams about :laugh:

 

[ilkhan]

Originally posted by: Fox5

Originally posted by: magreen

Originally posted by: Idontcare
The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

That is interesting. I'm assuming no cpu foundry has done that before (in the recent past), making a separate process just for their energy efficient lines?

If so, then it seems you're right to suspect graphics/chipsets/other customers, and not power efficient cpus, which are usually cherry-picked samples of the regular cpus that can work at low voltages, if I'm not mistaken (pls correct me if I'm wrong)

But AMD is ramping up very fast if this roadmap is true. Unexpectedly fast I'd say, when is Intel doing 32nm and 28nm?

Intel is going 32nm->22nm, and they figure 22nm will be released with ivy bridge, ETA Q4/2011.

Originally posted by: Idontcare

Originally posted by: magreen

Originally posted by: ilkhan
32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

well, that depends how you define "ramping." I'm sure idc can fill us in, but they may have ramped the 45nm phII years before it was released this year.

Yeah remember this is GlobalFoundries process tech roadmap, not AMD's product roadmap.

I would expect at least a 3-6month lag between GlobalFoundries announcing the availability of a node versus a customer announcing having received samples from the node.

(consider tapeout to first silicon to debug/verification to respin and repeat before product is released takes around 6 months at best and first silicon can't be effectively generated until the node itself is nearly production ready)

So figure early 2011 for AMD 32nm? That sounds about right from what I remember.

And because its relevant, isn't 22nm a "minor" shrink for intel, whereas 32nm is a "major" shrink for AMD? (forgive the terms, new equipment or something).

 

[drizek]

So that means AMD is still about year behind, Right? At least we know they are on schedule.

 

[Fox5]

Originally posted by: ilkhan

Originally posted by: Fox5

Originally posted by: magreen

Originally posted by: Idontcare
The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

That is interesting. I'm assuming no cpu foundry has done that before (in the recent past), making a separate process just for their energy efficient lines?

If so, then it seems you're right to suspect graphics/chipsets/other customers, and not power efficient cpus, which are usually cherry-picked samples of the regular cpus that can work at low voltages, if I'm not mistaken (pls correct me if I'm wrong)

But AMD is ramping up very fast if this roadmap is true. Unexpectedly fast I'd say, when is Intel doing 32nm and 28nm?

Intel is going 32nm->22nm, and they figure 22nm will be released with ivy bridge, ETA Q4/2011.

Originally posted by: Idontcare

Originally posted by: magreen

Originally posted by: ilkhan
32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

well, that depends how you define "ramping." I'm sure idc can fill us in, but they may have ramped the 45nm phII years before it was released this year.

Yeah remember this is GlobalFoundries process tech roadmap, not AMD's product roadmap.

I would expect at least a 3-6month lag between GlobalFoundries announcing the availability of a node versus a customer announcing having received samples from the node.

(consider tapeout to first silicon to debug/verification to respin and repeat before product is released takes around 6 months at best and first silicon can't be effectively generated until the node itself is nearly production ready)

So figure early 2011 for AMD 32nm? That sounds about right from what I remember.

And because its relevant, isn't 22nm a "minor" shrink for intel, whereas 32nm is a "major" shrink for AMD? (forgive the terms, new equipment or something).

I thought AMD already transitioned to new tech for 45nm, whereas Intel is waiting until 32nm.

 

[KingstonU]

Originally posted by: Fox5
They could probably do an atom competitor on that line. (scale the x2's down)

AMD is not going to try and compete in the Atom market, because there is no profit there. Intel has publicly stated that they regret making the Atom because of the very lower profit margins.

 

[soccerballtux]

Originally posted by: Fox5

Originally posted by: ilkhan

Originally posted by: Fox5

Originally posted by: magreen

Originally posted by: Idontcare
The 4X nm low-power bulk-Si process was news to me, I wonder what customers they have lined up for this process. ATI mobile chips? (GPU and chipset)

That is interesting. I'm assuming no cpu foundry has done that before (in the recent past), making a separate process just for their energy efficient lines?

If so, then it seems you're right to suspect graphics/chipsets/other customers, and not power efficient cpus, which are usually cherry-picked samples of the regular cpus that can work at low voltages, if I'm not mistaken (pls correct me if I'm wrong)

But AMD is ramping up very fast if this roadmap is true. Unexpectedly fast I'd say, when is Intel doing 32nm and 28nm?

Intel is going 32nm->22nm, and they figure 22nm will be released with ivy bridge, ETA Q4/2011.

Originally posted by: Idontcare

Originally posted by: magreen

Originally posted by: ilkhan
32nm in Q1? Are they serious? That would be...amazing. Isn't Phenom II their 45nm shrink? If thats true they'll have gone 65->32 in ?5 quarters.

well, that depends how you define "ramping." I'm sure idc can fill us in, but they may have ramped the 45nm phII years before it was released this year.

Yeah remember this is GlobalFoundries process tech roadmap, not AMD's product roadmap.

I would expect at least a 3-6month lag between GlobalFoundries announcing the availability of a node versus a customer announcing having received samples from the node.

(consider tapeout to first silicon to debug/verification to respin and repeat before product is released takes around 6 months at best and first silicon can't be effectively generated until the node itself is nearly production ready)

So figure early 2011 for AMD 32nm? That sounds about right from what I remember.

And because its relevant, isn't 22nm a "minor" shrink for intel, whereas 32nm is a "major" shrink for AMD? (forgive the terms, new equipment or something).

I thought AMD already transitioned to new tech for 45nm, whereas Intel is waiting until 32nm.

Depends on the tech you're talking about. AMD/GF is implementing Hi-k come 32nm. Intel already has this. Intel is implementing immersion litho come 32nm. AMD already has this.

 

[Viditor]

Originally posted by: ilkhan

So figure early 2011 for AMD 32nm? That sounds about right from what I remember.

And because its relevant, isn't 22nm a "minor" shrink for intel, whereas 32nm is a "major" shrink for AMD? (forgive the terms, new equipment or something).

I believe that AMD is planning on shipping 32nm in mid 2010 still...

 

[Idontcare]

Originally posted by: ilkhan
So figure early 2011 for AMD 32nm? That sounds about right from what I remember.

That's my expectation as well.

Originally posted by: ilkhan
And because its relevant, isn't 22nm a "minor" shrink for intel, whereas 32nm is a "major" shrink for AMD? (forgive the terms, new equipment or something).

The terminology you are reaching for is "half-node" versus "full-node".

For both GlobalFoundries (AMD) and Intel both the 32nm and 22nm nodes are considered "full nodes".

For TSMC both 32nm and 22nm are considered (by their own statement of such) as being half-nodes; 32nm is the half-node sibling to the current 40nm full-node, and 22nm will be the half-node follow-up to the 28nm node.

(the distinction between half-nodes and full-nodes is a very gray area, more subjective than technical...historically the half-node moniker was reserved for nodes in which the only significant change in the underlying process technology in going from full-node to half-node geometries were changes restricted to the lithography dept, that hasn't truly been the case recently since circa the 80nm half-node as most litho changes now result in dramatic changes in hardmask and resist integration as well as the etch and cleans processes along with implant conditions and general xtor improvements and enhancements)

In all cases new equipment (to some non-zero quantity) is involved in all process shrinks, half-node or full.