Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel)

[DisEnchantment]

TSMC's N7 EUV is now in its second year of production and N5 is contributing to revenue for TSMC this quarter. N3 is scheduled for 2022 and I believe they have a good chance to reach that target.


N7 performance is more or less understood.


This year and next year TSMC is mainly increasing capacity to meet demands.

TSMC Radically Boosts CapEx to Expand Production Capacities, To Reach $14B For 2019

www.anandtech.com

For Samsung the nodes are basically the same from 7LPP to 4 LPE, they just add incremental scaling boosters while the bulk of the tech is the same.

Samsung is already shipping 7LPP and will ship 6LPP in H2. Hopefully they fix any issues if at all.
They have two more intermediate nodes in between before going to 3GAE, most likely 5LPE will ship next year but for 4LPE it will probably be back to back with 3GAA since 3GAA is a parallel development with 7LPP enhancements.

Samsung’s Aggressive EUV Plans: 6nm Production in H2, 5nm & 4nm On Track

www.anandtech.com

Samsung's 3GAA will go for HVM in 2022 most likely, similar timeframe to TSMC's N3.
There are major differences in how the transistor will be fabricated due to the GAA but density for sure Samsung will be behind N3.
But there might be advantages for Samsung with regards to power and performance, so it may be better suited for some applications.
But for now we don't know how much of this is true and we can only rely on the marketing material.

This year there should be a lot more available wafers due to lack of demand from Smartphone vendors and increased capacity from TSMC and Samsung.
Lots of SoCs which dont need to be top end will be fabbed with N7 or 7LPP/6LPP instead of N5, so there will be lots of wafers around.

Most of the current 7nm designs are far from the advertized density from TSMC and Samsung. There is still potential for density increase compared to currently shipping products.
N5 is going to be the leading foundry node for the next couple of years.

For a lot of fabless companies out there, the processes and capacity available are quite good.

 

[qmech]

With Intel’s buyout of Tower dead, they’ve announced a new partnership in its place. If this goes well, I expect it may expand in the future.

Intel Foundry Services and Tower Semiconductor Announce New US Foundry Agreement

www.intc.com

So Tower will invest (up to) $300 million in equipment that it will then install (but still own) at Intel's New Mexico facility.

The skeptic in me feels like this is Tower basically just going through with the expansion plans they had already made when they were going to be part of Intel. The $300 million also just happens to be what Intel paid Tower as a termination fee.

It probably makes good sense for both parties. Tower gets to expand as planned and Intel doesn't have to find a new use for the space at the NM facility that was allocated to Tower.

 

[DrMrLordX]

People tend to cling to all sorts of reasons why China will fail, and the West will succeed, while the evidence to the contrary keeps piling on and on.

Oh please. China is quick to toot its own horn. Meanwhile their bread & butter industries are decamping for other countries with cheaper labor forces that aren't imploding.

The worst possible way to solve the population decline is by importing the most desperate people of the globe, with 8 years of education and IQ of 80.

IQ of 80? Seriously? No. Not even on the worst of days.

Sadly any further point-by-point response threatens to devolve into another P&N-type post so I'll just summarize: TSMC doesn't have much to fear from SMIC anytime soon. The performance of their N+1/N+2 7nm process doesn't seem that good, at least based on that one phone using the N+2 node. Maybe they will flood the market with cheap cryptocurrency ASICs. Kaspa, Bitcoin, and even RandomX can benefit from that (assuming they don't fork RandomX again). Which would be ironic considering that it isn't even legal to operate cryptocurrency ASICs in China.

 

[A///]

I'd say it already is there. I'm bowing out for now just because I wouldn't touch this now radioactive mess with a barge pole.

 

[sdifox]

Kirin 9000s is likely the last batch of chips TSMC fabbed for Huawei before the embargo.


Given the billions scammed by Chinese semiconductor startups, I don't see how they could have achieved what they are claiming.


Examples

https://www.chinatalk.media/p/billion-dollar-heist-how-scammers

https://www.telecomtv.com/content/access-evolution/head-of-china-s-homegrown-microprocessor-industry-fund-arrested-in-fraud-scandal-45135/

 

[Joe NYC]

Kirin 9000 is likely the last batch of chips TSMC fabbed for Huawei before the embargo.

It seems like there should be an easy answer to this mystery...

 

[Joe NYC]

Kirin 9000s is likely the last batch of chips TSMC fabbed for Huawei before the embargo.


Given the billions scammed by Chinese semiconductor startups, I don't see how they could have achieved what they are claiming.


Examples

https://www.chinatalk.media/p/billion-dollar-heist-how-scammers

https://www.telecomtv.com/content/access-evolution/head-of-china-s-homegrown-microprocessor-industry-fund-arrested-in-fraud-scandal-45135/

Apparently, the chip is real:

TechInsights Finds SMIC 7nm (N+2) in Huawei Mate 60 Pro | TechInsights

TechInsights, the semiconductor industry's leading information platform, conducted an analysis of the Huawei Mate 60 Pro smartphone and uncovered groundbreaking evidence of SMIC's 7nm (N+2) technology, marking a significant milestone in China's domestic design and manufacturing capabilities.

www.techinsights.com

 

[sdifox]

Apparently, the chip is real:

TechInsights Finds SMIC 7nm (N+2) in Huawei Mate 60 Pro | TechInsights

TechInsights, the semiconductor industry's leading information platform, conducted an analysis of the Huawei Mate 60 Pro smartphone and uncovered groundbreaking evidence of SMIC's 7nm (N+2) technology, marking a significant milestone in China's domestic design and manufacturing capabilities.

www.techinsights.com

That is some monumental effort to get there with DUV. Wonder what kind of yield are we talking about.

https://www.washingtonpost.com/business/2023/09/04/huawei-chip-shows-us-curbs-are-porous-not-useless/289cc9f4-4b00-11ee-bfca-04e0ac43f9e4_story.html

 

[ikjadoon]

PR: MediaTek Successfully Develops First Chip Using TSMC's 3nm Process, Set for Volume Production in 2024

Arm acknowledged in June that its Cortex-X4 (and presumably all the over Arm v9.2 cores it exclusively partners with) had been taped out on N3E:

“In this new generation of CPU designs, we are taking our long-standing partnership with TSMC a step further through taping out the Cortex-X4 on the TSMC N3E process – an industry first,” says Bergey, “this ensures that our ecosystem is ready to maximise the PPA benefits of our processor technologies once they are taped out." Source

Ideally (for the data of course, lol), we'll have X4 on a '4nm' class node in the same time frame for a real head-to-head showdown.

 

[poke01]

PR: MediaTek Successfully Develops First Chip Using TSMC's 3nm Process, Set for Volume Production in 2024

Arm acknowledged in June that its Cortex-X4 (and presumably all the over Arm v9.2 cores it exclusively partners with) had been taped out on N3E:

Ideally (for the data of course, lol), we'll have X4 on a '4nm' class node in the same time frame for a real head-to-head showdown.

That whole press release is lie if Apple releases a 3nm TSMC SoC next week in 2023.

They mention 3NE, so for the PR statement to be accurate then Apples is TSMC 3NB.


OR they mean it's MediaTek's first 3nm chip.

 

[lightisgood]

PR: MediaTek Successfully Develops First Chip Using TSMC's 3nm Process, Set for Volume Production in 2024

Great.
It is rumored that TSMC's N3 (N3B) is suffering from ~60% low yield and high wafer cost.
I guess that MediaTek is going to be a de facto winner of TSMC's 3nm process because N3E is bug fixed 3nm.

 

[poke01]

Great.
It is rumored that TSMC's N3 (N3B) is suffering from ~60% low yield and high wafer cost.
I guess that MediaTek is going to be a de facto winner of TSMC's 3nm process because N3E is bug fixed 3nm.

Let's be honest for Apple 3NB is fine cause they sell the whole product and they have deals with TSMC as well.

Now, I do think only Apple can handle 3NB while everyone else needs 3NE.

The benefit of 3NB is it's out 2H23

 

[A///]

I read it to mean their orders come after apple. for all we know tsmc has been producing n3 derived processors for apple for many months going by that june date posted earlier. why this assumption they're just about to spin up wafers?

some of you treat the gray area as to be ignored when it comes to aura and mystery of tsmc.

 

[lightisgood]

Let's be honest for Apple 3NB is fine cause they sell the whole product and they have deals with TSMC as well.

Now, I do think only Apple can handle 3NB while everyone else needs 3NE.

The benefit of 3NB is it's out 2H23

Disagree.
Apple seems failure to stop loss on the N3.
N3 is expensive, low productivity process and not fit to make smartphone SoC.
I think that optimum products for N3 are H100, Gaudi, MI300, and so on...
So, nobody wants N3 in smartphone or PC market, however, lacking compatibility between N3 and N3E forces Apple to buy N3 wafer.

 

[Ajay]

So, nobody wants N3 in smartphone or PC market, however, lacking compatibility between N3 and N3E forces Apple to buy N3 wafer.

I don't think so. The lack of compatibility between did force Apple into any choice. The decision was made well in advance and Apple just needed new chips to help maintain sales momentum for the iPhone (new this, new that, better this...). Don't recall what A18 will be fabbed on (N3P/X??)

 

[jpiniero]

I don't think so. The lack of compatibility between did force Apple into any choice. The decision was made well in advance and Apple just needed new chips to help maintain sales momentum for the iPhone (new this, new that, better this...). Don't recall what A18 will be fabbed on (N3P/X??)

Rumors are that Apple is going to switch over whatever products use N3B to use N3E ASAP.

I think TSMC offered Apple a deal just so something will use N3B to amortize whatever tools are specific to N3B.

 

[Ajay]

I think TSMC offered Apple a deal just so something will use N3B to amortize whatever tools are specific to N3B.

Well, that and Apple was committed - they needed to get A17 out (for at least the Pro models)and N3E would be too late.

 

[jpiniero]

Well, that and Apple was committed - they needed to get A17 out (for at least the Pro models)and N3E would be too late.

N4P should be comparable in quality to N3B... so they could have just used that instead.

 

[ikjadoon]

That whole press release is lie if Apple releases a 3nm TSMC SoC next week in 2023.

They mention 3NE, so for the PR statement to be accurate then Apples is TSMC 3NB.


OR they mean it's MediaTek's first 3nm chip.

MediaTek and TSMC (TWSE: 2330, NYSE: TSM) today announced that MediaTek has successfully developed its first chip using TSMC’s leading-edge 3nm technology, taping out MediaTek’s flagship Dimensity system-on-chip (SoC) with volume production expected next year.

MediaTek's numbers align with N3E (e.g., +18% perf, +60% logic density), instead of N3 (i.e., +10-15% perf, +70% logic density).

That slide is from Arm, FWIW, and not MediaTek.

 

[Ajay]

N4P should be comparable in quality to N3B... so they could have just used that instead.

No. They targeted N3B well in advance. It just turned out that yields sucked. If they tried a backport, Apple would have missed their target for this year's iPhone.

 

[Doug S]

N4P should be comparable in quality to N3B... so they could have just used that instead.

N3B is marginally better in both power and performance. Not a lot, but a few percent is better than nothing.

TSMC badly needed someone to start depreciating their N3 lines, they have had little or no N3 revenue since it entered "mass production" around the new year. If Apple was not using N3B for A17 then instead of six months of idle lines TSMC would have a full year of idle lines. In a capital heavy industry that's devastating to your bottom line.

Most likely the reason TSMC gave Apple the KGD pricing was to entice them to use N3B. No doubt TSMC made it worth Apple's while, cost wise, to use N3B instead of N4P even ignoring the marginal improvement in power/performance. TSMC would be worse off if Apple paid them more to use N4P, because they'd end up with a capacity crunch there while their N3 lines would have sat idle, or nearly so, for a full year.

 

[DrMrLordX]

No. They targeted N3B well in advance. It just turned out that yields sucked. If they tried a backport, Apple would have missed their target for this year's iPhone.

That might also explain why Apple apparently took all of Intel's N3B allocation when Intel delayed their actual wafer orders (major black eye for Intel since they prepaid for all that capacity).

 

[A///]

That might also explain why Apple apparently took all of Intel's N3B allocation when Intel delayed their actual wafer orders (major black eye for Intel since they prepaid for all that capacity).

I'm sure Apple paid Intel to get that allocation. Would lessen the sting.

 

[Ajay]

I'm sure Apple paid Intel to get that allocation. Would lessen the sting.

WHY?

 

[A///]

WHY?

If Intel prepaid for allocation and time and TSMC gave it to apple without letting Intel have a say just to keep their N3 lines afloat to avoid financial trouble than TSMC undermines their own trustworthiness by doing that. The order from Intel was not small, and it would make other companies think twice before working with TSMC if they're going to reschedule or ignore fabrication holds. While Intel needs TSMC to "survive" TSMC also needs confidence in their motives and they shouldn't make irrational decisions because their N3 line sitting empty for months.

 

[Ajay]

If Intel prepaid for allocation and time and TSMC gave it to apple without letting Intel have a say just to keep their N3 lines afloat to avoid financial trouble than TSMC undermines their own trustworthiness by doing that. The order from Intel was not small, and it would make other companies think twice before working with TSMC if they're going to reschedule or ignore fabrication holds. While Intel needs TSMC to "survive" TSMC also needs confidence in their motives and they shouldn't make irrational decisions because their N3 line sitting empty for months.

Sorry, wasn't clear. Why would Apple pay Intel for wafers that it decided not to use?