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

Page 33 - 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
1,590
5,722
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.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
  • Like
Reactions: Tlh97 and Lodix
Jul 27, 2020
15,750
9,817
106

Is Brooke Asset Management getting screwed by shouldering 49% of the project costs?
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
Is Brooke Asset Management getting screwed by shouldering 49% of the project costs?

If it all goes well, they will share the profits.

That said, I think this development is a big step toward a fully independent IFS, effectively splitting Intel into independent design and manufacture businesses. I think independence is needed to gain full support from the industry (and governments), and hence necessary for success.


PS. There's a poll in the discussion thread above. Please leave your vote.
 

moinmoin

Diamond Member
Jun 1, 2017
4,933
7,619
136
If it all goes well, they will share the profits.

That said, I think this development is a big step toward a fully independent IFS, effectively splitting Intel into independent design and manufacture businesses. I think independence is needed to gain full support from the industry (and governments), and hence necessary for success.
If IFS isn't structurally independent that will make a huge mess of bookkeeping. If Brooke Asset Management is entitled to parts of the profit Intel will (have to) try hard to keep all financials in the clear legally.
 
  • Like
Reactions: Tlh97 and Vattila

RTX

Member
Nov 5, 2020
90
40
61
The next step after hybrid bonding is basically making two die surfaces optically flat and possibly cheaper?flatsurfaces.png
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
If IFS isn't structurally independent that will make a huge mess of bookkeeping. If Brooke Asset Management is entitled to parts of the profit Intel will (have to) try hard to keep all financials in the clear legally.

Nah that shouldn't be a problem. The money wouldn't go directly to the Hedge Fund, it would go to NotIntelFab Inc. Which would get paid per wafer presumably.
 

moinmoin

Diamond Member
Jun 1, 2017
4,933
7,619
136
Nah that shouldn't be a problem. The money wouldn't go directly to the Hedge Fund, it would go to NotIntelFab Inc. Which would get paid per wafer presumably.
That would be a dumb independent fab not even including any fab related research capacity, wouldn't it? Single use, just the node(s) it was built for and, as a NotIntelFab Inc., no capacity to move (further/away) in any direction.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
That would be a dumb independent fab not even including any fab related research capacity, wouldn't it? Single use, just the node(s) it was built for and, as a NotIntelFab Inc., no capacity to move (further/away) in any direction.

There will be some fixed charge per wafer for IP licensing or whatever you want to call it to pay for Intel's R&D and expertise that make the fab able to produce something useful, which will have been agreed upon up front.
 
  • Like
Reactions: Tlh97 and Vattila

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
That would be a dumb independent fab not even including any fab related research capacity, wouldn't it? Single use, just the node(s) it was built for and, as a NotIntelFab Inc., no capacity to move (further/away) in any direction.

The announcement was light on details.
 

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
With the recent passage of a new American law, America intends to feed Taiwan to the wolves (China). With America attempting to reinvigorate their manufacturing industry, they realize defending a foreign supplier is hopeless ...
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
With the recent passage of a new American law, America intends to feed Taiwan to the wolves (China). With America attempting to reinvigorate their manufacturing industry, they realize defending a foreign supplier is hopeless ...

Turns out TSMC gets some of that money to expand here.
 

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
Turns out TSMC gets some of that money to expand here.

Yeap, by bringing TSMC abroad the US is both undermining both Taiwan's leading industry and their own existence ...

America is starting to realize that it ultimately wants to ditch Taiwan since the people over there weren't very interested in participating military service. A part of the reason why the US held "strategic ambiguity" was so that the regime they installed over there could one day develop the capabilities to defend themselves or highly deter their hostile neighbour from invading ...

America recently learned the hard way that they don't want to do nation building for hopeless states so they want the obvious crown jewel (TSMC) here ...
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Yeap, by bringing TSMC abroad the US is both undermining both Taiwan's leading industry and their own existence ...

Not really. It just gives TSMC more facilities and, assuming they still have customers, more profits. That money goes back to Taiwan. It doesn't stay in the US.

Specifically, TSMC wants to produce in the United States to meet the needs of customers here. They do have a lot of them. Apple, AMD, Qualcomm, NVidia.

That's like saying Honda and Toyota (and Nissan) producing vehicles in the United States kills industry in Japan. It doesn't.
 
  • Like
Reactions: igor_kavinski

FangBLade

Member
Apr 13, 2022
199
395
96
Looks like TSMC N3E is making good progress. Risk production appears to be 3 months out.
View attachment 66256
View attachment 66255
So 5nm yiels are excellent, that means despite 5nm is "much" more expensive than 7nm, AMD with chiplet design and that yields have very good margins, they won't have trouble with decreasing price if Intel pushes them, and also AMD biggest advantage when it comes to profit is they use single architecture for all segments.
 
Jul 27, 2020
15,750
9,817
106
Continued from here: https://forums.anandtech.com/thread...-4-genoa-ryzen-7000-etc.2571425/post-40840581

I'll continue this in more detail in a more appropriate thread if you want
Let's suppose Intel 7 and N7 are comparable. Does this mean ADL boosts a lot more aggressively in any given second, giving the CPU less time to cool down? But then how is AMD managing to keep from boosting as often? Is it because Zen 3 cores are so well fed that they are always busy crunching through work so that even at base clock, they get a lot more work done than ADL? This would also mean that ADL P-cores are starved for data and keep waiting, burning power during the waits. So why can't Intel keep them downclocked while they wait? Or is it that their branch mispredictions are so bad that they go through multiple failed branches in the time it takes Zen 3 to successfully execute two or more correctly predicted branches? I know I must not be making sense :D
 

Exist50

Platinum Member
Aug 18, 2016
2,445
3,043
136
Continued from here: https://forums.anandtech.com/thread...-4-genoa-ryzen-7000-etc.2571425/post-40840581


Let's suppose Intel 7 and N7 are comparable. Does this mean ADL boosts a lot more aggressively in any given second, giving the CPU less time to cool down? But then how is AMD managing to keep from boosting as often? Is it because Zen 3 cores are so well fed that they are always busy crunching through work so that even at base clock, they get a lot more work done than ADL? This would also mean that ADL P-cores are starved for data and keep waiting, burning power during the waits. So why can't Intel keep them downclocked while they wait? Or is it that their branch mispredictions are so bad that they go through multiple failed branches in the time it takes Zen 3 to successfully execute two or more correctly predicted branches? I know I must not be making sense :D
So, looking at the link you gave, this situation is very simple. It's a multicore load, in which ADL will be around 241W, and Vermeer around 105W (respective out of box boost limits). Meanwhile, just from looking at a die shot, I'd venture a guess that ADL spends ~130mm2 on CPU cores, while Vermeer spends a combined ~160mm2. Someone could double check those numbers, and I'm likely a bit off, but either way, that's significantly more heat per die area. That's the dominant factor in higher temps.
 
Jul 27, 2020
15,750
9,817
106
I'd venture a guess that ADL spends ~130mm2 on CPU cores, while Vermeer spends a combined ~160mm2. Someone could double check those numbers, and I'm likely a bit off, but either way, that's significantly more heat per die area. That's the dominant factor in higher temps.
So reduction of just 30mm2 in die area can result in a temperature hike of 30 degrees Celsius??? :-O

7 nm lithography process - WikiChip

If you (and other people on this forum) are right, why is WikiChip comparing Intel 4 to N7?
 
Jul 27, 2020
15,750
9,817
106
Sorry, can't edit above post coz having to use VPNside web proxy and it doesn't like to play with AT forum buttons. Forced to use proxy because AT forum URL is not accessible at my workplace (seems to happen from time to time).

ANYWAY, I would think WikiChip people know a bit more about processes than anyone on this forum? (Unless we have real silicon engineers lurking around here who like to mess with our heads :D)