Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel) - [2020 - 2025]

Page 210 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,777
6,789
136
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.

1587737990547.png
N7 performance is more or less understood.
1587739093721.png

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

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.


1587739615344.png

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.

---------------------------------------------------------------------------------------------------------------------------------------------------


FEEL FREE TO CREATE A NEW THREAD FOR 2025+ OUTLOOK, I WILL LINK IT HERE
 
Last edited:

511

Diamond Member
Jul 12, 2024
3,707
3,491
106
According to reports from Korean media, TSMC has set the production price of its 2nm process at $30,000 and has established a policy of supplying it without any “discounts.”

TSMC plans to begin trial production within the next 34 months, securing an initial capacity of 30,000 to 35,000 wafers per month, and aims to build a system capable of producing 60,000 wafers per month across four plants in 2026. The initial yield is reported to be in the mid-60% range, while for SRAM it is said to reach over 90%, a level that poses no difficulty for mass production.
 
  • Like
Reactions: Elfear

Win2012R2

Golden Member
Dec 5, 2024
1,146
1,179
96
What about moving your n-1 design to Intel node? You don't suddenly stop making N3 chips when you make N2 node chips
1. Big fixed cost porting stuff - new masks etc
2. By that point N3 will be paid off by early adopters, so discounts might be available, plus newer versions perhaps
3. TSMC won't allow usage of packaging for stuff made elsewhere - can Intel do same stuff as X3Ds?
4. Upsetting TSMC - priceless.
 

511

Diamond Member
Jul 12, 2024
3,707
3,491
106
Big fixed cost porting stuff - new masks etc
one time Intel has the mask shop(IMS Nano fabrication) which is used by TSMC as well so they can get discount.
2. By that point N3 will be paid off by early adopters, so discounts might be available, plus newer versions perhaps
oh really see the post above
3. TSMC won't allow usage of packaging for stuff made elsewhere - can Intel do same stuff as X3Ds?
Well but Intel packages TSMC made dies for Itself and Amazon as for the X3D yes we will se in Q126 in HVM
4. Upsetting TSMC - priceless.
If it's they US government they can't do anything but watch.

If it's fine with fabless to pay 50% more for 15%more than by any means do so just don't complain. TSMC is in the same power someday Intel was maybe slightly below that position.
 
  • Like
Reactions: DKR

Win2012R2

Golden Member
Dec 5, 2024
1,146
1,179
96
one time Intel has the mask shop(IMS Nano fabrication) which is used by TSMC as well so they can get discount.
Yeah and the design rules are 100% compatible with TSMC N3?

If it's they US government they can't do anything but watch
What's they gonna do - kill fabless NASDAQ leaders to make Intel survive? In 3 years many of current bunch will be in jail or bankrupt from civil litigation: all current tariffs are illegal and will be reverted, with damages. In fact this will start in 18 months once Dems take back Congress.
 

DZero

Golden Member
Jun 20, 2024
1,453
530
96
US is still capitalist?

1. Ban China from using TSMC

2. Invest a stake in Intel

3. Put up massive tariffs on imported chips

So the US is saying a foreign competitor can't sell to a major market, and if you want to sell to my market, pay massive tariffs.
Don't forget that Xiaomi is being stopped to use advanced nodes, still uses newest ARM tech.
 

gdansk

Diamond Member
Feb 8, 2011
4,388
7,384
136
You may be surprised to learn that capitalist states have a long history of trying to prevent the spread of technology to other countries and taking stakes in industries they consider important or struggling. For example... who owned Rolls Royce in 1971? It was a high technology company likewise cleft in twain by the government and the turbine engine company is quite successful today.
 
  • Like
Reactions: DKR and 511

gdansk

Diamond Member
Feb 8, 2011
4,388
7,384
136
Yeah, that worked out really well with China.
Well, that's an entirely different political issue. You see, a few years after the Gang of Four was couped by the PLA the US/France/UK/Germany decided to sell the PRC technology. And the rest is off topic.
 

Win2012R2

Golden Member
Dec 5, 2024
1,146
1,179
96
Well, that's an entirely different political issue.
That was just capitalism business (ie: short term profits over long term) - it started 35-40 years ago for greed reasons and China was smart enough to insist on "technology transfer".

Any current limits are too little way too late, maybe they won't be able to train some models, but they sure as heck will be able to steal weights to run inference on cheap locally made hardware. Frankly this I think will push them to finally invade Taiwan, actually not even invade because it is recognised internationally (including by US) as China, sold out by Nixon and Kissinger.

If China decide to retaliate they will stop exporting usual daily stuff - shelves in Wall Mart will be like in late 80s USSR.
 

gdansk

Diamond Member
Feb 8, 2011
4,388
7,384
136
That was just capitalism business (ie: short term profits over long term) - it started 35-40 years ago for greed reasons and China was smart enough to insist on "technology transfer".
No, it started for geopolitical reasons (to exploit the Sino-Soviet split) but it continued as it was profitable.

In any case, I noted an example of so-called capitalist states doing basically what is proposed for Intel to great long-term success. Religious adherence to some mythological concept of "capitalism" (itself defined by the detractors to whatever system of production that has existed in the west) has never been followed at any time by any country. History is littered with examples of meddling (to prevent technology transfer) and bailing out (to prevent accumulated mistakes killing a key industry).
 
  • Like
Reactions: controlflow and 511

511

Diamond Member
Jul 12, 2024
3,707
3,491
106
We all know Intel is too big too fail as it's the only US Company left but the main problem is how will the government save it without making it a monopoly again.
In leading edge Fabrication only tech matters if you have the best tech you are a Monopoly.
 
  • Like
Reactions: DKR

Khato

Golden Member
Jul 15, 2001
1,261
333
136
Data centers need to use lower power GPUs from N2 node.
Except for the fact that NVIDIA's high-end is currently using N4P, and rumors point to the next iteration being N3P. When you have both good design and a captive software ecosystem, the leading edge isn't required.

There's no question that a number of products are sensitive to making use of the best possible process technology. They'll almost certainly be allowed to continue doing so without penalty... so long as investments are made to manufacture the non-sensitive products at Intel. Pretty much the exact same scheme as we've seen keep certain electronics exempt from all tariffs thus far.
 

marees

Golden Member
Apr 28, 2024
1,458
2,050
96
In a new research note shared this week, analyst Jeff Pu said that Apple could be among the customers considering Intel’s upcoming 14A process for future M-series chips.

According to Pu, Intel has already started handing out early 14A process design kits (PDKs) to partners, and Apple is one of the companies showing interest:

The next focus Intel 14A process will incorporate second-gen RibbonFET and PowerDirect, marking a technological evolution built upon the foundation of the PowerVia introduced in Intel’s 18A. Targeting both AI and edge applications, Intel has already provided early versions of the 14A PDK to key customers, with several expressing interest in producing test chips. We anticipate NVIDIA’s gaming GPU (low-end version) and Apple’s M series to emerge as adopters of Intel14A.
This means that Pu believes that both Apple and NVIDIA will be among the first customers to adopt 14A, possibly for lower-end gaming GPUs and future M-series chips, respectively.