TSMC on track for volume production of 7nm chips for iPhone, iPad in 2018

AppleInsider

Apple chip manufacturer TSMC is reportedly on schedule to receive the first designs for chips using the 7nm FinFET process in the second quarter of 2017, paving the way for new power-sipping A-series processors for future iPad and iPhone models.

According to supply chain reports from the Commercial Times, the first prototype chips in the 7nm FinFET process will be available in the second quarter just after "tape out" is complete, with full volume production in early 2018, in time for that fall's refresh of the iPhone.

"Tape out" is the last step in the design process for the chip, just prior to mass manufacture. Specifically, the photo-mask of the chip is done, and ready to be sent to the fabrication facility. Additional revisions are often made after tape out is complete, and before the chips are manufactured at scale.

Besides just Apple, other customers slated for the technology include Qualcomm, Xilinx, and Nvidia. TSMC is said to already have 15 customers for chips using the process, but is seeking 20.

The timeline for production of 7nm chips is on track, according to previous goal-posts placed by TSMC

Apple's A10 Fusion chip as found in the iPhone 7 family utilizes TSMC's 16nm FinFET process. Apple's A9 in the iPhone 6S and iPhone SE, and A9X processors found in the 12.9-inch iPad Pro use the same size die.

The Fall 2017 line of iPhones are expected to utilize a 10nm process chip.

Two otherwise identical chips under the same load with different die sizes have less power consumed by the chip manufactured with the smaller die. As a result, in most cases the generated heat is less, giving better performance-per-watt metrics.

A more in-depth update on the progress of manufacturing is expected at TSMC's investor conference on Jan. 15.

wood1208

TSMC is said to already have 15 customers for chips using the process, but is seeking 20. How about Samsung ?

melgross

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

and as we saw with the 16nm process from TSMC vs the 14nm process from Samsung, we can also see that a smaller process can actually be significantly less efficient.

the only true 14nm process out is Intel's. What ever these are, they won't be what they want people to think they are.

slprescott

Is there a lower-limit on  FinFET size?  At 7nm, are we approaching that limit?

cali

I just want Samsung to feel the burn. That's all.

blastdoor

melgross said:

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

The GPU guys tend to lag a year behind Apple in adopting a new process. Part of that's about finances, but a big part of it is that Apple's SOCs are smaller than GPU chips, which makes Apple's SOCs easier to manufacture on a new process. In order to get decent yields on big GPU chips, TSMC needs to gain experience with and refine a new process. 

So it's not like the 10nm fabs will be sitting idle after Apple moves to 7nm. They will be cranking out GPUs (among other things). 

Also, Apple does still sell older SOCs in older products. For example, the A8 is still being used in the AppleTV and the A8 is a 20nm chip. 

jdb8167

slprescott said:

Is there a lower-limit on  FinFET size?  At 7nm, are we approaching that limit?

Yep, I think 7 nm is it for FinFET. But the actual feature size now is mostly marketing. So maybe someone will get to 5 nm FinFET because marketing says 5 nm is better than 7 nm.

When Intel is having yield problems on 14 nm, you can be pretty sure everyone else is too. So I'm skeptical of even 10 nm much less 7 nm.

dick applebaum

blastdoor said:

melgross said:

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

The GPU guys tend to lag a year behind Apple in adopting a new process. Part of that's about finances, but a big part of it is that Apple's SOCs are smaller than GPU chips, which makes Apple's SOCs easier to manufacture on a new process. In order to get decent yields on big GPU chips, TSMC needs to gain experience with and refine a new process. 

So it's not like the 10nm fabs will be sitting idle after Apple moves to 7nm. They will be cranking out GPUs (among other things). 

Also, Apple does still sell older SOCs in older products. For example, the A8 is still being used in the AppleTV and the A8 is a 20nm chip. 

Never thought much about it -- older processes continue high-yield production while newer processes ramp up....  What you say makes perfect sense!

Kinda' like GM in the old days: Chevy, Pontiac, Buick, Oldsmobile, Cadillac -- one line would get major makeover, while other lines would add chrome and fins.

melgross

slprescott said:

Is there a lower-limit on  FinFET size?  At 7nm, are we approaching that limit?

Yes, that's it.

melgross

blastdoor said:

melgross said:

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

The GPU guys tend to lag a year behind Apple in adopting a new process. Part of that's about finances, but a big part of it is that Apple's SOCs are smaller than GPU chips, which makes Apple's SOCs easier to manufacture on a new process. In order to get decent yields on big GPU chips, TSMC needs to gain experience with and refine a new process. 

So it's not like the 10nm fabs will be sitting idle after Apple moves to 7nm. They will be cranking out GPUs (among other things). 

Also, Apple does still sell older SOCs in older products. For example, the A8 is still being used in the AppleTV and the A8 is a 20nm chip. 

This has nothing to do with Apple. Apple can only use what's available. Last time, for the first time, Apple took a chance on a process change with an architecture change. That's risky for anyone. It worked for Apple. Maybe they were lucky, it's hard to tell.

Apple's chips aren't all that small. They're the largest SoC out there, and are about the same size as a number of Intel's chips.

melgross

jdb8167 said:

slprescott said:

Is there a lower-limit on  FinFET size?  At 7nm, are we approaching that limit?

Yep, I think 7 nm is it for FinFET. But the actual feature size now is mostly marketing. So maybe someone will get to 5 nm FinFET because marketing says 5 nm is better than 7 nm.

When Intel is having yield problems on 14 nm, you can be pretty sure everyone else is too. So I'm skeptical of even 10 nm much less 7 nm.

There are still many in the chip industry who are skeptical that 5nm is possible. So am I. FinFet dies after 7nm. There is yet no replacement for it.

wizard69

melgross said:

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

and as we saw with the 16nm process from TSMC vs the 14nm process from Samsung, we can also see that a smaller process can actually be significantly less efficient.

the only true 14nm process out is Intel's. What ever these are, they won't be what they want people to think they are.

The Yield problems, at least in the case of TSMC appear to be complete BS. Someplace (I've yet to find the original) TSMC denied the recent report in Digitimes. Samsung is in a position to deliver 10nm chips very early this years so I really don't see TSMC being that far behind. Beyond all of that this is a story about risk production with full scale production for 2018.

wizard69

melgross said:

blastdoor said:

melgross said:

Bah! I find this hard to believe. Everyone now attempting to produce 10nm has run into delays from yield problems. Considering the cost of these factories, there is just no way they are going to produce 10nm for less than a year before jumping to 7nm. I don't believe it.

The GPU guys tend to lag a year behind Apple in adopting a new process. Part of that's about finances, but a big part of it is that Apple's SOCs are smaller than GPU chips, which makes Apple's SOCs easier to manufacture on a new process. In order to get decent yields on big GPU chips, TSMC needs to gain experience with and refine a new process. 

So it's not like the 10nm fabs will be sitting idle after Apple moves to 7nm. They will be cranking out GPUs (among other things). 

Also, Apple does still sell older SOCs in older products. For example, the A8 is still being used in the AppleTV and the A8 is a 20nm chip. 

This has nothing to do with Apple. Apple can only use what's available. Last time, for the first time, Apple took a chance on a process change with an architecture change. That's risky for anyone. It worked for Apple. Maybe they were lucky, it's hard to tell.

Apple's chips aren't all that small. They're the largest SoC out there, and are about the same size as a number of Intel's chips.

Try explaining that Apple can only use whats available to people demanding a better Intel processor in a MBP.     Yeah I know  senseless but people need to understand technology a bit better.  


As for the risks Apple takes that is part of why they can charge a premium for the iPhone.    At least they didn't take a risk with the battery like some companies we know.   As for Apples large SoC, people just don't seem to get it, these are highly innovative chips that are very much on the bleeding edge of what is possible.   Not only are they as large as Intel solutions but in many cases more dense and certainly have more functionality embedded.   Apples A series is one of the reasons I cringe when I here people talk about Apple not innovating, The A series is perhaps the most innovative SoC out there right now.

kongwee

time for graphene cpu.