Discussion Intel leading customer for TSMC 3nm?

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

naukkis

Senior member
Jun 5, 2002
702
571
136
Article says Intel is getting capacity at the cost of Apple. You don't bat an eye at that?

Apple has financed TSMC's leadind edge nodes because they had to, not because they want to. If Intel co-financed TSMC leading edge nodes that probably fine for Apple. Burning enormous amount money if it isn't absolutely necessarily isn't fine business.
 

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
Who says they are "taking capacity away" from anyone. What is so hard to understand about first come first served? If you come to TSMC tomorrow with a check for a couple billion dollars, you might be first in line for N2 wafers. Since Apple is probably already first in line for N2, more likely you'd be second. If you wait until N2 is around the corner, like N3 is now, you won't be first, or second, or maybe even third.
Read the article please. They say Apple has delayed the adoption of N3 for their phones because of Intel chalking up capacity. Among other 'minor' details, these kinds of remarks are what make the whole article together with its source fishy as fudge.
 

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
Apple has financed TSMC's leadind edge nodes because they had to, not because they want to. If Intel co-financed TSMC leading edge nodes that probably fine for Apple. Burning enormous amount money if it isn't absolutely necessarily isn't fine business.
Personally, I could absolutely buy that. What I couldn't buy is Apple delaying anything because of Intel. The Ego & Arrogance podium is still Apple-Nvidia-Intel and not the other way around - my 2 cents as always.
 

Gideon

Golden Member
Nov 27, 2007
1,608
3,573
136
Personally, I could absolutely buy that. What I couldn't buy is Apple delaying anything because of Intel. The Ego & Arrogance podium is still Apple-Nvidia-Intel and not the other way around - my 2 cents as always.
Agreed, but it might very well be that only that part of the report is inaccurate and Apple simply hasn't that much volume until the rampup of iphone early the year after
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
It will be bad for AMD (and Apple) today and then bad for TSMC once Intel get their processes working as they will get back all this procuction in their own fabs, in the meantime they will stop marketshare bleeding thanks to a complacent and short sighted TSMC...

It is a huge assumption that Intel will necessarily EVER "get their processes working", if by "working" you mean catch up to let alone surpass TSMC.

Now that Intel is buying some of TSMC's leading edge capacity, TSMC benefits from having essentially the entire world customer base for leading edge capacity, other than Samsung of course, being a customer for their leading edge process.

That's going to make it very tough for Intel to compete.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
Read the article please. They say Apple has delayed the adoption of N3 for their phones because of Intel chalking up capacity. Among other 'minor' details, these kinds of remarks are what make the whole article together with its source fishy as fudge.

No, the article does NOT say that. At least not the article linked at the start of this thread.

TSMC has been been saying all along that N3 mass production would happen in "H2 2022". Unless it is at the very start of H2, it is too late for the 2022 iPhone's standard September launch.
 

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
Some wild speculation here. Be gentle.

What do we know.

Cost/transistor has started to increase as node size decreases reversing the previous trend.
Improvements in power, speed are dropping with new node.
TSMC SoIC CoW trails the adoption of a new node by several quarters
Density optimized cache is 2X CPU density

Speculation.

Less processing steps needed for pure cache than CPU die, so even cheaper/transistor.


Putting these together, we can see how designs from AMD and Intel might be diverging.

Traditional way is getting the latest node to stay at the front in the various performance parameters of power, density and cost reduction (no longer true).

Maybe, a different path is now chosen where V-cache and 3D tech allows a better solution.

1)CPU cores are designed traditional way with the cost/transistor now rising slightly. CPU costs increase as even an identical design will cost more at a newer node in spite of the die size falling dramatically.
2)L3 V-cache layers are optimized for density allowing the net total cost/transistor to keep falling relative to designing the combined cores +cache die as one on less dense rules. If 1/2 of your CPU is L3 cache and you can fab it at 2X density, your total silicon area and cost will be 75% of similar transistor count single level design. Assembly, of layers, of course, will eat into that cost saving.

In other words, if you want a CPU to cost X $, you get the highest performance by stacking + optimizing each level with the best libraries for that level, and this works out cheaper (less area and possibly mask steps) than doing everything on one level.

Zen3 was designed before CoW ability was available so it used traditional L3 design for 1st iteration. Future CPUs are designed to use it from the start.

We have seen L3 designed as a concentrated block of cache cells. Is this a design hard wall or a choice based on traditional design? Might we see The entire CPU cores covered by L3 cache with none on the core level? Cache power consumption is relatively low, so is covering the entire cores with L3 cache a problem from a heat dissipation perspective As it might add a few % points to average heat density? Critical hot points would have to accommodated, but this is a small part of your total core area.


Puts some understanding on AMD's statement that they're no longer chasing the newest node at launch. I think we're in store for some amazing products over the next few years.
 

Abwx

Lifer
Apr 2, 2011
10,852
3,298
136
It is a huge assumption that Intel will necessarily EVER "get their processes working", if by "working" you mean catch up to let alone surpass TSMC.

Now that Intel is buying some of TSMC's leading edge capacity, TSMC benefits from having essentially the entire world customer base for leading edge capacity, other than Samsung of course, being a customer for their leading edge process.

That's going to make it very tough for Intel to compete.

It will take a few years eventually but if they have the money it s about sure, they already gave a few hints about the money they ll invest in the coming years.

What they did by outsourcing a part of their production is to keep AMD from getting enough waffers to reach say 50% marketshares in the short term, wich was bound to happen given the huge perf/watt advantage of TSMC s advanced process.
 
  • Like
Reactions: Tlh97 and KompuKare

JasonLD

Senior member
Aug 22, 2017
485
445
136
What they did by outsourcing a part of their production is to keep AMD from getting enough waffers to reach say 50% marketshares in the short term, wich was bound to happen given the huge perf/watt advantage of TSMC s advanced process.

TSMC 3nm won't affect existing 5, 6, and 7nm wafer allocation, so Intel's possible 3nm allocation won't affect AMD.
 

John Carmack

Member
Sep 10, 2016
153
246
116
I'm curious, a lot of people have been posting all from this mindwhatever (some German retail site) and sites talking about how the 5000x is so hard to get because they are selling so well and no one could keep them in stock. How is it that Intel gained marketshare in 1Q 2021 over AMD? You know, cause people for the past two years have been saying Intel was dead and they can't survive. Makes me curious?

*snip*

But I will say I was surprised based on all the comments/threads here on how AMD was selling out everything. Intel still gained marketshare in the Laptop/desktop segment. Congrats to Intel and AMD!

The answer is, it depends.

If you have Intel blinders on: AMD Ryzen 5000 selling out everywhere despite overpriced MSRP and retailer scalping? Fake news! Intel gained marketshare!

If you don't have Intel blinders on then you should look at the Intel 1Q2021 thread.

Datacenter volume down 13% Y/Y and ASP down 14%.

Desktop volume down 4% Y/Y and ASP down 5%.

Notebook volume up 54% Y/Y and ASP down 23%.

AMD was/is supplying 6, 8, 12, 16 core chips to the mid to high end of the market for top dollar while Intel was/is supplying their highest end dual cores and quad cores to the low end notebook market while having to deeply discount Comet Lake to move product.

I don't think Intel is using TSMC for GPUs. If anything, they are using it for 3nm and using it to their advantage. Even if they slipped on 10nm? You'd be dumb not to use every resource possible to gain back your advantage. You can't hate Intel for using TSMC when AMD uses the same company. That's just stupid.

In the AMD 1Q2021 thread you were very much concerned about TSMC's projected 2022 shortages short circuiting growth for AMD so what advantage is there to be gained by Intel if the supplier can't supply?
 

Kuiva maa

Member
May 1, 2014
181
232
116
I do not have any insider info whatsoever. It is plausible that a combination of intel offering copious amounts of cash and TSMC looking for big profits short term direcly from a competing foundry, could outmaneuver AMD in the 3nm timeline. What I find ludicrous is TSMC prioritizing intel over apple. This will never happen, since Apple was the kingmaker to TSMC foundry rule. Apple indirectly funded their R&D that allowed them to be where they are now. And apple is exactly the one company TSMC cannot afford to cross. If they sense that their iphone cadence could be put to risk like that, they may switch foundry partner or put things in motion that will let them fab their own chips longterm. If there is one company that can enter that market need be, that's apple. None of that is worth some extra bucks from intel.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
I do not have any insider info whatsoever. It is plausible that a combination of intel offering copious amounts of cash and TSMC looking for big profits short term direcly from a competing foundry, could outmaneuver AMD in the 3nm timeline. What I find ludicrous is TSMC prioritizing intel over apple.

Doesn't appear that's what happened. The 3 nm node is just simply too late for the 2022 iPhone, but it was in time for some 2023 iPads. So they bought what they needed and Intel bought the rest. If Apple is upset about anything it's that they aren't able to get 3 nm available in time.
 
  • Like
Reactions: Exist50

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
If you have Intel blinders on: AMD Ryzen 5000 selling out everywhere despite overpriced MSRP and retailer scalping? Fake news! Intel gained marketshare!

Intel definitely lost desktop market share but gained a ton of notebook share. Granted most of those notebooks aren't running Windows.
 

Hitman928

Diamond Member
Apr 15, 2012
5,182
7,632
136
Doesn't appear that's what happened. The 3 nm node is just simply too late for the 2022 iPhone, but it was in time for some 2023 iPads. So they bought what they needed and Intel bought the rest. If Apple is upset about anything it's that they aren't able to get 3 nm available in time.

Assuming this is true, what happens when Apple comes calling for those wafers in 1H23?
 

Hitman928

Diamond Member
Apr 15, 2012
5,182
7,632
136
That's a good question. Presumably Apple would get what they want and Intel would be second in line.

That's why I have a hard time believing the rumor with the volume given. What could Intel produce for a couple of quarters, then just stop producing (or drastically reduce volume) that anyone would care about? Maybe I'm just blind to the possibilities here, but it just doesn't make sense to me. If the goal is to prevent AMD from getting those wafers as has been suggested, well, AMD wasn't going to take them to begin with inside that time frame so I don't see how that is a smart move on Intel's part. Again, this is all assuming this isn't part of some long term purchasing agreement from Intel which would make more sense but would also be a huge signal that Intel is going to go through a pretty big shift in their business model.
 
  • Like
Reactions: Tlh97 and moinmoin

Failnaught

Member
Aug 4, 2008
26
25
91
Every time I've heard the idea of "skipping a node" to catch up suggested, either in the recent past or 20 years ago, process engineers have said that such a thing is almost impossible.

I'm trying to learn, I'm not in the industry, please forgive the noob question.

What is the thing that you must learn in mass production that you cannot learn at small volume, R&D stage? To draw the analogy, is it like a clinical trial where you need huge population/test numbers to find the small problems that hurt yield/performance etc (like the blood clot problem with the Oxford/AZ vaccine, which is ~1 in 1M)? But probably you don't need huge volumes to find a problem that's affecting yield measurably (at say 0.1% level).

Related to that, it is claimed (for example here: https://stratechery.com/2018/intel-and-the-danger-of-integration/ ) that Intel's fab problems originates from controlling smaller fab market share, which is all getting concentrated to TSMC/Samsung at the leading edge. How much of this is true? I mean, sure, infinite money helps, and you can do crazy things like have multiple teams of engineers work in parallel on the same problem. But isn't most of the money used to build fabs going to setting up the factories, as opposed to paying + equipping a smallish elite team to figure things out?
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
That's why I have a hard time believing the rumor with the volume given. What could Intel produce for a couple of quarters, then just stop producing (or drastically reduce volume) that anyone would care about? Maybe I'm just blind to the possibilities here, but it just doesn't make sense to me. If the goal is to prevent AMD from getting those wafers as has been suggested, well, AMD wasn't going to take them to begin with inside that time frame so I don't see how that is a smart move on Intel's part. Again, this is all assuming this isn't part of some long term purchasing agreement from Intel which would make more sense but would also be a huge signal that Intel is going to go through a pretty big shift in their business model.

The wafer projection did say TSMC was projected to do 120 kwpm in 2023, which is a ton. I suppose it's possible that TSMC might bring more capacity online by the time Apple needs more.
 

Hitman928

Diamond Member
Apr 15, 2012
5,182
7,632
136
The wafer projection did say TSMC was projected to do 120 kwpm in 2023, which is a ton. I suppose it's possible that TSMC might bring more capacity online by the time Apple needs more.

Where did the 120 kwpm for 3 nm projection come from? That looks quite high to me for 2023, especially 1H23.
 

Doug S

Platinum Member
Feb 8, 2020
2,202
3,405
136
That's why I have a hard time believing the rumor with the volume given. What could Intel produce for a couple of quarters, then just stop producing (or drastically reduce volume) that anyone would care about? Maybe I'm just blind to the possibilities here, but it just doesn't make sense to me. If the goal is to prevent AMD from getting those wafers as has been suggested, well, AMD wasn't going to take them to begin with inside that time frame so I don't see how that is a smart move on Intel's part. Again, this is all assuming this isn't part of some long term purchasing agreement from Intel which would make more sense but would also be a huge signal that Intel is going to go through a pretty big shift in their business model.

Why would Intel "stop producing"? iPad Pro is not a high volume product like iPhone, it can't soak up all of TSMC's N3 like they have been soaking up nearly all of N5 over the past year. Even if they also do an M3 SoC for the Mac in N3 that's still chickenfeed compared to iPhone wafer needs.

If Apple can't get N3 in time for the 2022 iPhone then their next high volume product will be the 2023 iPhone. TSMC will have brought a lot of additional N3 (and maybe N3P depending on when that's ready) capacity online between the launch of the 2022 and 2023 iPhone.