Discussion Intel leading customer for TSMC 3nm?

Page 8 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Thunder 57

Platinum Member
Aug 19, 2007
2,647
3,706
136
I haven't been paying much attention since my last upgrade (2013 lol), but from what I can tell it looks like AMD is in a MUCH better spot today than they were back then. tbh I thought they were on the verge of bankruptcy for years and now they have a $100b stock valuation.

They are, and they were. I don't think it's a stretch to say that Zen literally saved them from bankruptcy. They have been able to iterate on it year after year as well with great results.
 
  • Like
Reactions: Jimzz and Tlh97

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Why would they? TSMC is supposedly ahead of schedule with N4. Who even has a node as far along as N3 that has comparable performance? Samsung's GAA offerings look good on paper, but it's hardly on schedule to be ready ahead of N3. Let's not even start with Intel, with their one developmental Intel4 product.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Who even has a node as far along as N3 that has comparable performance?
Intel 4 should be close enough. Can't rule out Intel's ability to bounce back strong and fast. Any stalling now only plays into Intel's favor, and who knows how long this delay actually takes?
Let's not even start with Intel, with their one developmental Intel4 product.
Why not?
 

clemsyn

Senior member
Aug 21, 2005
531
197
116
Intel 4 should be close enough. Can't rule out Intel's ability to bounce back strong and fast. Any stalling now only plays into Intel's favor, and who knows how long this delay actually takes?

Why not?

I really hope that their new CEO fixes Intel in many ways. Its a tough job but I hope he does it. Competition is always good for us consumers.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Intel doesn't have any EUV equipment to do any kind of real volume of 7 nm/I4 any time near Q1 23.
My remark is actually aimed at whether Apple is going to get a bit shook and start looking at contingency plans, going forward. TSMC's seemingly flawless cadence is now exhibiting a beloved patriot in the armor.

Edit: Why is the matrix interpreting "beloved patriot in the armor" as "beloved patriot"? Tried editing many times without success.

Edit1: I give up.
 

Hitman928

Diamond Member
Apr 15, 2012
5,178
7,629
136

TSMC now says that they aren't going to ship finished 3 nm wafers out until q1 2023.

While not an official delay, this means that they enter HV manufacturing a few months later than probalby most expected.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136

TSMC now says that they aren't going to ship finished 3 nm wafers out until q1 2023.
This article is kind of nuts. It been rumored for a while that N4 was being used for the next SoC after A15 (so, A16, if Apple keeps using this nomenclature). That would put A17 on N3, in time for the 2023 iPhone. Worst case, N3 isn't a very good node, because of poor process performance, and the perf/eff advantage over N4 is substandard.

If that is the case, SS might gain some customers, but it won't gain Apple.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
My remark is actually aimed at whether Apple is going to get a bit shook and start looking at contingency plans, going forward.

They must have some sort of contingency plan; if only because of the hype around the Chinese invasion of Taiwan. I'd have to think that involves Samsung more than Intel.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
Apple may soon cast their silicon nets wider.


Even with the delays TSMC's process is more advanced than Samsung's. Based on what Samsung says about 3GAE compared to their 7nm process, they will be lucky if 3GAE can match up with N4 let alone N3.

This delay probably doesn't even matter to Apple. If TSMC is announcing it now, Apple has been aware of the delay or possibility of the delay for months - they have their own process engineers embedded in TSMC's team. If as many believe Apple had been planning all along to use N4 for A16, it won't matter. Even if they thought at one time they might be able to use N3 they would have found this out in plenty of time to change course for N4.

Perhaps this may even be an advantage for them, as if TSMC started mass production just a bit too late to be used for A16 (and presumably the chips used in the Macs released late summer / early fall next year) then their competition would have access to a better process for nearly a year. The difference between N4 and N3 isn't very large, but Apple is used to having the advantage or at worst parity, they have never been behind in process since they switched to TSMC.

If N3 doesn't enter mass production until sometime in Q1 2023 that shrinks the window for others to get N3 wafers and be ahead, before Apple starts taking tens of thousands of wafers that summer.
 
  • Like
Reactions: Tlh97

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Intel and bleeding edge have been cursed for multiple generations now.

Let's hope for competition's sake it's different this time.
 

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
So TSMC is finally slowing down after all! I guess there is some hope for their competitors to catch up especially Samsung since they have the best possible chance. In the end the industry is going to have to find a way to cope in a post-silicon world very soon since improvements to CMOS technology will likely plateau in a little over 5 years ...
 

remsplease

Junior Member
Oct 22, 2021
16
3
41
There is news that Intel is the co-leading customer on the TSMC 3nm node, with Apple. Did they manage to outflank AMD on process technology? What are they going to make on this node?

https://asia.nikkei.com/Business/Te...become-first-to-adopt-TSMC-s-latest-chip-tech

Both Intel and AMD have had their fair share of in-house FAB growing pains so they're (obviously) now both using TSMC in greater or lesser capacities to keep moving forward with goals, designs, etc.

Intel and Apple have tested, working 3nm and/or 5nm designs in place so they can commit to TSMC node scheduling. My present understanding of AMD's road map is primarily 7nm with 5nm currently in design phase so they haven't purchased 3nm/5nm node-time beyond yield & test runs (yet) but I'm sure they will.

As to what is on the nodes for Intel?..not sure but I would venture a guess it's GPU at 3nm and server/workstation/Alder Lake successor processors at 5nm. Intel stalled their 7nm node sixteen months ago to retool and improve quality/yield so I surmise 7nm production (Alder lake +) will go back to Intel foundries or has already started the move back in-house. Then they'll begin moving 5nm back in-house one they stand-up 5nm nodes and then 3nm with some products continuing to be outsourced.

I believe AMD has some additional growing pains to push through with Global Foundries but the recent shares sell announcement shows the commitment is still there. Coming from the days of unlocking their processors with a graphite pencil (weee!) and then conductive car window defroster repair ink (not so weee!), I'm confident they will get back on track.

Based on a whole lot of odd predictive posts on a lot of boards, I think AMD may be (currently) outflanked in terms of completed, tested, yielding designs being ready to go at 5nm and 3nm. 3D cache seems to be a move in response to Alder Lake but it's going to be at the cost of higher power consumption and higher temps. Same design, same 7nm node, additional cache. Relatively easy to implement. But, we don't see stacking a lot because it's difficult to get past the lower die components being insulated by whatever is on top (more heat + more power) Flat, monolithic dies are still the go-to option. AMD would need to rework and test the current 7nm Zen design to move away from stacking and they don't have the time to do that with Alder Lake around the corner.

I'm a firm believer in healthy *ethical* competition. That having been said, I believe we are going to see a near-term shift back to Intel as the performance leader in the processing space. Very eager to see what ARC has to offer and given the amount of time it's been on Intel's design roadmap, I'd bet a dollar it's going to be competitive. Same with chiplet-based GPUs from AMD. Nvidia seems to be a little more silent these days and appear to be pushing 8nm as far as they can. Interested to see where they're headed.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
That having been said, I believe we are going to see a near-term shift back to Intel as the performance leader in the processing space.

Watch what happens in the server sector and then get back to us. It's about to be Sapphire Rapids vs. Genoa, and Genoa will be launching first (publicly). ODM sale timelines are a different matter, though it looks like Intel isn't even going to be ready for those in Q1, which is a bit shocking.
 

moinmoin

Diamond Member
Jun 1, 2017
4,933
7,619
136
I believe AMD has some additional growing pains to push through with Global Foundries but the recent shares sell announcement shows the commitment is still there.
May I ask what you are referring to with this part? Aside a commitment of buying $1.6 billion worth of wafers (independent of any node) until 2024 as part of the last WSA there are no remaining links between the two.
 

remsplease

Junior Member
Oct 22, 2021
16
3
41
May I ask what you are referring to with this part? Aside a commitment of buying $1.6 billion worth of wafers (independent of any node) until 2024 as part of the last WSA there are no remaining links between the two.

Ability to stand up a 7nm node. The last 14nm 5xxx card I purchased was produced at GF. They wanted 7nm 6xxx cards to come from GF as well.

Still making bank producing and selling stuff with what they have working but I think AMD probably wants to do more. I hope that happens.
 

remsplease

Junior Member
Oct 22, 2021
16
3
41
Watch what happens in the server sector and then get back to us. It's about to be Sapphire Rapids vs. Genoa, and Genoa will be launching first (publicly). ODM sale timelines are a different matter, though it looks like Intel isn't even going to be ready for those in Q1, which is a bit shocking.

Not claiming to have a crystal ball. However, when products are released vs. when they are purchased are apples and oranges when you factor the purchasing and purchased equipment life cycle habits of medium and large companies.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Ability to stand up a 7nm node. The last 14nm 5xxx card I purchased was produced at GF. They wanted 7nm 6xxx cards to come from GF as well.

Still making bank producing and selling stuff with what they have working but I think AMD probably wants to do more. I hope that happens.
I think they sold all their 7N specific equipment. 12N FDX isn’t up yet, IIRC. So, GF doesn’t really have any process of much use to AMD. I suppose AMD can keep making Zen2 IODs for some Low end market once Zen4 is out. Tough to see them using up $1.6B though.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Same with chiplet-based GPUs from AMD. Nvidia seems to be a little more silent these days and appear to be pushing 8nm as far as they can. Interested to see where they're headed.

Nvidia is usually silent - I’m sure they are busily at work getting their 5N product (Lovelace) debugged and yielding well. Not sure what the expected release date is.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
I think they sold all their 7N specific equipment. 12N FDX isn’t up yet, IIRC. So, GF doesn’t really have any process of much use to AMD. I suppose AMD can keep making Zen2 IODs for some Low end market once Zen4 is out. Tough to see them using up $1.6B though.

There's always Dali and Picasso. Plus rumors of low end Zen 3 at GloFo.
 
  • Like
Reactions: Tlh97 and Ajay