TSMC Shows Path to 16nm, Beyond

[Fjodor2001]

Already discredited.
http://forums.anandtech.com/showthread.php?t=2344229

The info in the article I linked to is correct. It clearly states that "Taiwan Semiconductor Manufacturing Co. (TSMC) has become the world's largest integrated circuit (IC) chip supplier in terms of the final market value of its sales". I agree that it may be an unusual way of measuring sales though...

The other article under discussion in the thread you linked to does not make that clarification regarding what sales they mean though, so I agree that article is confusing. But don't mix that article up with the one I linked to, even though they happen to discuss the same topic.

 

[ShintaiDK]

The info in the article I linked to is correct. It clearly states that "Taiwan Semiconductor Manufacturing Co. (TSMC) has become the world's largest integrated circuit (IC) chip supplier in terms of the final market value of its sales". I agree that it may be an unusual way of measuring sales though...

The other article under discussion in the thread you linked to does not make that clarification regarding what sales they mean though, so I agree that article is confusing. But don't mix that article up with the one I linked to, even though they happen to discuss the same topic.

Thats like writing Intel is the biggest company in the world due to final sales.

 

[BallaTheFeared]

Process technology lead is intact. Roadmap technology lead is shrinking fast!

Thats like writing Intel is the biggest company in the world due to final sales.

:thumbsup:

Quality posts, keep up the good work fellas.

 

[Idontcare]

Thats like writing Intel is the biggest company in the world due to final sales.

And they would be, if you computed "final sales" for Intel's chips by the same means in which the TSMC spokesperson computed TSMC's "final sales" figures.

I don't particularly find it interesting that spokespersons for companies intentionally mangle and mix their words and definitions, the conflict of interest is blatantly obvious and everyone knows to expect it from them.

Rather, what I personally find intriguing is the blatantly obvious confirmation bias that is so clearly in play when unaffiliated members champion their favored company, or the downfall of their most loathed company.

The psychology involved in that equation is what I find interesting about these discussions. The "enemy of my enemy is my new best friend" posts. Something really drives that motivation, and confirmation bias rears its head, but why do people gravitate towards such obvious false constructs? Is it unavoidable?

 

[Fjodor2001]

Thats like writing Intel is the biggest company in the world due to final sales.

Not at all. As I understand it by "final market value" TSMC means what their customers (e.g. Qualcomm) in turn sell their chips for to their customers (e.g. Samsung or Sony).

Saying that this compares to the sales of all products that have an Intel CPU in them is just ridiculous, because then you are adding the sales value of all other components in those products (e.g. SSD, motherboard, RAM, etc).

 

[ShintaiDK]

Rather, what I personally find intriguing is the blatantly obvious confirmation bias that is so clearly in play when unaffiliated members champion their favored company, or the downfall of their most loathed company.

The psychology involved in that equation is what I find interesting about these discussions. The "enemy of my enemy is my new best friend" posts. Something really drives that motivation, and confirmation bias rears its head, but why do people gravitate towards such obvious false constructs? Is it unavoidable?

Seems unavoidable as long as the people involved dont understand that none of the companies have any interest in them at all. Or that none of the companies will/can give them what they want, if they champion or loathe them enough. I guess it could be defined as a mental illness as obsessions.

I can understand your local soccer team or something. But these companies....

 

[Fjodor2001]

Seems unavoidable as long as the people involved dont understand that none of the companies have any interest in them at all. Or that none of the companies will/can give them what they want, if they champion or loathe them enough. I guess it could be defined as a mental illness as obsessions.

I can understand your local soccer team or something. But these companies....

Funny to hear you talking about yourself in third person like that...

 

[krumme]

And they would be, if you computed "final sales" for Intel's chips by the same means in which the TSMC spokesperson computed TSMC's "final sales" figures.

I don't particularly find it interesting that spokespersons for companies intentionally mangle and mix their words and definitions, the conflict of interest is blatantly obvious and everyone knows to expect it from them.

Rather, what I personally find intriguing is the blatantly obvious confirmation bias that is so clearly in play when unaffiliated members champion their favored company, or the downfall of their most loathed company.

The psychology involved in that equation is what I find interesting about these discussions. The "enemy of my enemy is my new best friend" posts. Something really drives that motivation, and confirmation bias rears its head, but why do people gravitate towards such obvious false constructs? Is it unavoidable?

I stopped reading tsmc roadmap talk years ago. I sometimes read the gf cartoons because they are weird and funny.

So in this case i didnt read it either. Its impossible for me at least to get any meaningfull information from it. But i read the posts in this thread. I think they were good. Its extremely difficult to get a good discussions from the usual pr bs from tsmc about future processes. So i think you guys did okey

 

[Arachnotronic]

I stopped reading tsmc roadmap talk years ago. I sometimes read the gf cartoons because they are weird and funny.

So in this case i didnt read it either. Its impossible for me at least to get any meaningfull information from it. But i read the posts in this thread. I think they were good. Its extremely difficult to get a good discussions from the usual pr bs from tsmc about future processes. So i think you guys did okey

I think anybody who follows this industry and TSMC's previous timelines will know not to fall for their "roadmaps" ;-)

Hey, IDC or anybody else, what in the hell does "risk production" mean? As far as I can tell, at guys like Intel, "risk production" means running wafers before a design is qualified for sale. What does it mean in TSMC parlance?

 

[krumme]

I think anybody who follows this industry and TSMC's previous timelines will know not to fall for their "roadmaps" ;-)

Hey, IDC or anybody else, what in the hell does "risk production" mean? As far as I can tell, at guys like Intel, "risk production" means running wafers before a design is qualified for sale. What does it mean in TSMC parlance?

You seem to be interested after all?

Well i can assure you; whatever it means it does not give any insight because the details and the nuances is hidden anyway. The end result is just as unclear. Just because you can understand dog language its not a guarantie to get any valuable intellectual insight. (Thats gon a be the best comparison ever)

 

[krumme]

Well what is clear is that Intel is reducing capex.
Tsmc is increasing it.
Intel profit looks to stagnate or fall.
Tsmc profit looks to go the other way.

The scene is obviously changing fast these years and right now. Tsmc must have lower wages than Intel. Intel on the other hand have an advantage in years and competence.

Perhaps the process gab between Intel and Tsmc will not change in the next year or two. But the brute investment force from tsmc will show at some time. Their business model will work for them untill then. That is unless something unforseen changes.(And that always happen lol)

 

[dahorns]

Well what is clear is that Intel is reducing capex.
Tsmc is increasing it.
Intel profit looks to stagnate or fall.
Tsmc profit looks to go the other way.

The scene is obviously changing fast these years and right now. Tsmc must have lower wages than Intel. Intel on the other hand have an advantage in years and competence.

Perhaps the process gab between Intel and Tsmc will not change in the next year or two. But the brute investment force from tsmc will show at some time. Their business model will work for them untill then. That is unless something unforseen changes.(And that always happen lol)

Capex means nothing. You have to look at total expenditure on R&D. The numbers aren't close. I believe tsmc is around $500 million per quarter. Intel is at $2,000 Million per quarter. Brute force is on Intel's side. The question is will all that R&D become too expensive due to diminishing returns.

 

[ShintaiDK]

Well what is clear is that Intel is reducing capex.
Tsmc is increasing it.
Intel profit looks to stagnate or fall.
Tsmc profit looks to go the other way.

The scene is obviously changing fast these years and right now. Tsmc must have lower wages than Intel. Intel on the other hand have an advantage in years and competence.

Perhaps the process gab between Intel and Tsmc will not change in the next year or two. But the brute investment force from tsmc will show at some time. Their business model will work for them untill then. That is unless something unforseen changes.(And that always happen lol)

Your rationale is severely flawed. You are essentially saying Intel should keep building new fabs because thats the only way to keep a high capex.

Intel already build 3 new 450mm ready fabs that will run 14nm. And overall Intel fabs are pretty much up to date.

When you look on TSMC however, its a whole other case. They mainly own obsolete 200mm fabs with no upgrade path and only 3 fabs with 300-450mm capacity.

 

[krumme]

Capex means nothing.

Ahemm. Perhaps a little exageration when you think about it? Lol.

Intel r&d is much, much more than process development. Intel is not so f.....g one dimensional as most like to beliewe. Its a huge business besides designing producing and selling cpu. I agree there is a strong correlation from r&d but capex is a good single number to describe the situation for future process development.

 

[Revolution 11]

TSMC roadmaps are always accurate. I just add 2 years to any estimates they come up with. :awe:

 

[ShintaiDK]

TSMC roadmaps are always accurate. I just add 2 years to any estimates they come up with. :awe:

Haha :biggrin:

 

[Arachnotronic]

TSMC roadmaps are always accurate. I just add 2 years to any estimates they come up with. :awe:

TSMC's roadmaps are precise but not accurate

 

[CHADBOGA]

Thats like writing Intel is the biggest company in the world due to final sales.

Strike 3. :awe:

 

[dahorns]

Ahemm. Perhaps a little exageration when you think about it? Lol.

Intel r&d is much, much more than process development. Intel is not so f.....g one dimensional as most like to beliewe. Its a huge business besides designing producing and selling cpu. I agree there is a strong correlation from r&d but capex is a good single number to describe the situation for future process development.

Maybe over time Capex may be useful in analyzing future process development, but a capex differential in a single year is meaningless. Different companies are on different development timelines. Intel has already built their fabs.

R&D is a far better indicator of future progress of a high-tech company. You're right that Intel is diversified. That being said, I'm confident that the vast majority of its R&D goes to its bread and butter.

 

[Khato]

Since when was it something you could do with zero capability to design anything for the process?

If expecting nothing more than the small space heater result, yeah, taping out a design by no means requires a functional toolset. (Note that I never claimed they had zero capability, no, I fully expect that they had their theoretical design rules available at the time.) I wouldn't at all be surprised if ARM and TSMC did that tapeout PR stunt such that both of them would appear closer to their respective finished products to those who don't understand what a tapeout actually is.

That's not to say that it was necessarily a useless exercise. No, I'd imagine that they decided to tapeout an A57 core in addition to the usual logic test blocks because, well, why not? It'll give TSMC at least as much feedback for tweaking the process and design toolset, possibly even a bit more than just their typical logic test blocks would.

 

[krumme]

Maybe over time Capex may be useful in analyzing future process development, but a capex differential in a single year is meaningless. Different companies are on different development timelines. Intel has already built their fabs.

R&D is a far better indicator of future progress of a high-tech company. You're right that Intel is diversified. That being said, I'm confident that the vast majority of its R&D goes to its bread and butter.

And what is bread and butter for Intel really then?

 

[Exophase]

If expecting nothing more than the small space heater result, yeah, taping out a design by no means requires a functional toolset. (Note that I never claimed they had zero capability, no, I fully expect that they had their theoretical design rules available at the time.) I wouldn't at all be surprised if ARM and TSMC did that tapeout PR stunt such that both of them would appear closer to their respective finished products to those who don't understand what a tapeout actually is.

That's not to say that it was necessarily a useless exercise. No, I'd imagine that they decided to tapeout an A57 core in addition to the usual logic test blocks because, well, why not? It'll give TSMC at least as much feedback for tweaking the process and design toolset, possibly even a bit more than just their typical logic test blocks would.

I'm not sure if the distinction is very clear here.. I'm talking about the difference between having a broken design and not even having the capability to start the design. You may not have claimed zero capability but Phynaz seemed to.

Given the timeline, I'm not going to assume the A57 chips actually worked, but I think if they're calling it an A57 they at least had an attempted design with the appropriate tools and IP macros necessary to build it. That's very different from saying that they'd only be able to start the physical design phase by the end of the year.

I know TSMC, GF, Samsung, etc all produce roadmaps that are overly optimistic, misleading, or outright wrong, but I'm not buying the rationale that we won't see a 16nm product until Q4 2016 at the very earliest. This may end up being the case but I don't see a reason to extrapolate this from the currently available information. It'd be like saying that the first 20nm products won't hit until 2015.

 

[krumme]

TSMC roadmaps are always accurate. I just add 2 years to any estimates they come up with. :awe:

I dont think its that simple.

I often wondered what it was about. Is the reason strategically, tactical or more or less cultural? I mean it cant be delays always lol.

 

[krumme]

From the op eetimes link. Go and read the comments to the article. There are really some good at especially page 2 and 3.

 

[Phynaz]

With old school methods but not with semi automated design....

Then what are they now?