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

[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.

 

[FlameTail]

they really could steal allot from Samsung Logic foundry business. So, the pressure is really on for Intel to deliver. Honestly, I hope they do.

It shouldn't be difficult. Samsung Foundry is... well- if I were to describe it, I'll have to use some not so nice words.

We have discussed at length about it in the previous pages, including the fact that Samsung 2nm has seemingly worse density than even TSMC 3nm!

 

[Doug S]

they really could steal allot from Samsung Logic foundry business

Is there really that much to steal? As far as leading edge nodes, given that Intel isn't even pushing hard for foundry customers until 18A? Most of Samsung's leading edge business is Samsung, and they aren't going to ditch their own foundry for Intel's when it is barely keeping its head above water.

 

[Ajay]

Is there really that much to steal? As far as leading edge nodes, given that Intel isn't even pushing hard for foundry customers until 18A? Most of Samsung's leading edge business is Samsung, and they aren't going to ditch their own foundry for Intel's when it is barely keeping its head above water.

Fair point. I guess just put them out of the running for leading node design wins. Intel's success is predicated on a lot of 'ifs' - as I put in a prior thread.

 

[DavidC1]

I assure you if Intel is seriously ahead, Nvidia definitely, and even AMD eventually will use IFS. Actually Jensen said he will consider it so it's not an impossibility as some are suggesting.

Product quality speaks for itself, and at certain point, sells by itself. All argument about being fanboys and entrenched beliefs are thrown out the window once one product is decisively better.

The question you are asking is whether Intel can reach that point. There are doubts, and it's a reasonable one.

Also, when China takes Taiwan is another instance when Nvidia and AMD will use IFS.

This is a profound statement. Intel may have the best node in the industry, but it's no use if they don't have customers.

There's way more than whether a node is good or not in deciding to use them.

Intel had a 2.5 year lead and no one cared. The process group acted like they were gods and treated equipment vendors horribly. Fabbing was not a priority and likely any potential customers were horribly treated. Design tools were either proprietary and/or extremely difficult to work with.

Like I keep saying, it's the people not the tech!

 

[igor_kavinski]

Curious question: What is the backup plan of TSMC customers? Should TSMC production be "disrupted" indefinitely due to geopolitical reasons, who stands to gain the most? Samsung? Do they have enough capacity to handle all of TSMC customers, even though those customers would get screwed royally due to lack of choice and poor yields, not to mention inferior and problematic silicon? I doubt that any of the major customers (other than maybe QC or Mediatek) have designs compatible for fabbing in Intel foundries.

How many fabs does TSMC have outside Taiwan?

 

[itsmydamnation]

Curious question: What is the backup plan of TSMC customers? Should TSMC production be "disrupted" indefinitely due to geopolitical reasons, who stands to gain the most? Samsung? Do they have enough capacity to handle all of TSMC customers, even though those customers would get screwed royally due to lack of choice and poor yields, not to mention inferior and problematic silicon? I doubt that any of the major customers (other than maybe QC or Mediatek) have designs compatible for fabbing in Intel foundries.

How many fabs does TSMC have outside Taiwan?

if Taiwan is disrupted then so is South Korea and most of SEA, so everyone is boned given how much shipping there is in SEA.

its this exact reason Australia is going nuclear subs , time on station in choke points in SEA.

 

[DavidC1]

I don’t doubt that.

Similarly, it’s pretty clear MTL was half baked when it launched with unfinished microcode. So who’s to say it couldn’t have been launched 3-4 months earlier?

The fabs can and should announce when nodes have qualified for HVM independently of when a product is ready.

These are semantics. Companies reached the level they are with their own unique strategies. And TSMC's lead was propelled by Apple's, which relied on launching a product in September like clockwork. Apple isn't going to suddenly launch it in April just for the sake of accelerating TSMC.

Products are what matters.

Too bad Intel either cancelled every other product on Intel 4 or moved them to a different process.

Sierra Forest on Intel 3 is coming basically 6 months or less from now. If that's the case then process-wise there's no delay.

Sierra Forest launching that early is only possible because it's a totally new product line. Any significant product needs a 1-year cycle, so expecting a successor to Meteorlake 6 months later for example makes zero sense. Whatever successor to Meteorlake is, it's 12 months, and if the process is delayed because of it, so be it.

The Intel discussion thread shows even now Meteorlake isn't fully ready, with firmware giving significant improvements. Had it launched it July, it would have been catastrophic, and Intel likely knew this. Not that July was realistic since RPL released just 6 months ago.

if Taiwan is disrupted then so is South Korea and most of SEA, so everyone is boned given how much shipping there is in SEA.

its this exact reason Australia is going nuclear subs , time on station in choke points in SEA.

Computer chips will be the last thing on anyone's(at least sane ones) minds once it gets that bad.

 

[FlameTail]

Samsung? Do they have enough capacity to handle all of TSMC customers, even though those customers would get screwed royally due to lack of choice and poor yields, not to mention inferior and problematic silicon?

Aha. I am sure the massive influx of volume from all the ex-TSMC customers now going through Samsung Foundry will instantly fix their yield issues! /s

 

[igor_kavinski]

Aha. I am sure the massive influx of volume from all the ex-TSMC customers now going through Samsung Foundry will instantly fix their yield issues! /s

Samsung fab employees searching the wafer for good dies.

Their customers looking at them doing that and going, "Oh dear God! It's gonna be a longggggggg year..."

 

[Doug S]

Curious question: What is the backup plan of TSMC customers? Should TSMC production be "disrupted" indefinitely due to geopolitical reasons, who stands to gain the most? Samsung? Do they have enough capacity to handle all of TSMC customers, even though those customers would get screwed royally due to lack of choice and poor yields, not to mention inferior and problematic silicon? I doubt that any of the major customers (other than maybe QC or Mediatek) have designs compatible for fabbing in Intel foundries.

How many fabs does TSMC have outside Taiwan?

The world economy would be screwed worse than it was with covid if TSMC production in Taiwan was halted via a China incursion/invasion. They will be more resilient once they get the fabs in the US, Japan and (maybe, not sure if it is happening or not?) Germany going, but it will still lead to massive shortages.

So the question of "what would Apple/AMD/Qualcomm/etc. do if they couldn't get their chips from TSMC" is probably academic. Even if they could get their chips from TSMC there would be other stuff coming from Taiwan or from/through China they couldn't get (China isn't going to be able to waltz in and take it over without at minimum massive sanctions basically blocking all western imports/exports from China) so you wouldn't be able to buy iPhones even if Apple had stockpiled a bunch of SoCs. You wouldn't be able to buy Galaxys either. But that's fine, the market for "wants not needs" (like a smartphone/laptop/TV/car/etc. to replace the working one you already had) would disappear.

 

[igor_kavinski]

But that's fine, the market for "wants not needs" (like a smartphone/laptop/TV/car/etc. to replace the working one you already had) would disappear.

And prices on used hardware would go up Up UP!

 

[DrMrLordX]

Sierra Forest on Intel 3 is coming basically 6 months or less from now. If that's the case then process-wise there's no delay.

Are you suggesting that Intel 3 is identical to the Intel 4 we're seeing today in Meteor Lake?

 

[moinmoin]

Reducing the Taiwan issue to TSMC is plain inane. There are plenty companies situated in Taiwan, other significant foundries like UMC, OSATs like ASE Group, suppliers like Unimicron Technology, Nan Ya PCB, and Kinsus Interconnect Technology (remember the substrate shortage?) etc. pp. And that's not even starting with ODMs like Foxconn, Compal Electronics, Pegatron etc. pp.

So PR China recklessly executing what it keeps threatening to do would at once bring the whole worldwide technology industry to a standstill, no ifs and buts. (This will be my first and last post touching P&N, so please go there if you feel this is a discussion worth pursuing.)

 

[Doug S]

And prices on used hardware would go up Up UP!

Yep, just like what happened with used cars during and in the aftermath of covid. Can't build new cars, people will bid up the prices of late model used cars. If Apple couldn't make iPhone 16, prices of used iPhone 14 and 15 would shoot up.

 

[misuspita]

The world economy would be screwed worse than it was with covid if TSMC production in Taiwan was halted via a China incursion/invasion. They will be more resilient once they get the fabs in the US, Japan and (maybe, not sure if it is happening or not?) Germany going, but it will still lead to massive shortages.

So the question of "what would Apple/AMD/Qualcomm/etc. do if they couldn't get their chips from TSMC" is probably academic. Even if they could get their chips from TSMC there would be other stuff coming from Taiwan or from/through China they couldn't get (China isn't going to be able to waltz in and take it over without at minimum massive sanctions basically blocking all western imports/exports from China) so you wouldn't be able to buy iPhones even if Apple had stockpiled a bunch of SoCs. You wouldn't be able to buy Galaxys either. But that's fine, the market for "wants not needs" (like a smartphone/laptop/TV/car/etc. to replace the working one you already had) would disappear.

It would be something we never seen before in this generation. The huge majority of the best consumer electronics are coming from Taiwan and China. Apple, AMD, Nvidia, Qualcomm, all those will halt instantly all production, thus rendering incredible price woes. Not to mention, if those ASML equipment gets destroyed, there's no turning back... Can't fix them. And at the rate new ones are produced... good luck.

 

[jpiniero]

So the question of "what would Apple/AMD/Qualcomm/etc. do if they couldn't get their chips from TSMC" is probably academic. Even if they could get their chips from TSMC there would be other stuff coming from Taiwan or from/through China they couldn't get (China isn't going to be able to waltz in and take it over without at minimum massive sanctions basically blocking all western imports/exports from China) so you wouldn't be able to buy iPhones even if Apple had stockpiled a bunch of SoCs.

See... I don't think there will be sanctions if (when) China invades. And that's why. Since everything is made there, China has the US (and the rest of the world really) by the balls.

No guarantee that the fabs would be harmed during an invasion.

 

[Doug S]

See... I don't think there will be sanctions if (when) China invades. And that's why. Since everything is made there, China has the US (and the rest of the world really) by the balls.

No guarantee that the fabs would be harmed during an invasion.

The fabs would be inoperable because ASML would stop providing support for their machines. Without that they are basically useless. The west has China by the balls as well. If they sanction them and stop importing stuff from China, China's economy will take a massive hit. Tons of unemployment, lots of social unrest. The kind of conditions that lead to revolution.

 

[jpiniero]

The fabs would be inoperable because ASML would stop providing support for their machines. Without that they are basically useless.

ASML would only do so if there were sanctions. Since there won't be... I imagine they will continue to support.

I do think the invasion is why Intel hasn't committed to spinning off the fabs. That being said, despite what Intel says publically, I believe they are still going down the path of a spinoff. Just not committing to it just yet. Whether Intel can actually afford to stay on the leading edge without regaining their 90% server market share is kind of the question.

 

[eek2121]

Intel 20a is more than twice as dense as Intel 4 (> 400 mm2 density on test chips) apparently.

Side effect of Moore's law's death. Transistors not getting any smaller means you need to keep innovating, but at other levels.

Transistors are getting smaller. I don’t know why you think otherwise. Only the speed has not kept up.

2024 Big Race is TSMC N2 and Intel 18A - Semiwiki

There is a lot being said about Intel getting the lead back from TSMC with their 18A process. Like anything else in the semiconductor industry there is much more here than meets the eye, absolutely. From the surface, TSMC has a massive ecosystem and is in the lead as far as process technologies...

semiwiki.com

This is a profound statement. Intel may have the best node in the industry, but it's no use if they don't have customers.

Apple, Qualcomm, Nvidia, AMD are all "whales". Which one of them will most likely jump ship to Intel Foundry?

Whoever stated this has their head in the sand. Intel has had customers for IFS since before it was called “IFS”. They just previously did not have any big customers outside of the biggest one of all…themselves.

I know people are being pessimistic, but ever since they announced their 5 nodes in 4 years initiative, little has changed. They are executing.

I am willing to bet Arrow Lake launches on 20a next year.

 

[desrever]

Intel hasn't even demonstrated they can compete with Samsung at this point, people here think they can compete with TSMC?

People here seem to think they are going to beat TSMC?

Intel's current best node:

• barely competitive with N5
• has barely any products using it
• 3 year later than TSMC N5
• can't produce GPUs

At least Samsung had a 5nm class node 2 years ago with somewhat competitive products. If it wasn't for sanctions Intel would be 4th place behind SMIC today.

 

[Doug S]

ASML would only do so if there were sanctions. Since there won't be... I imagine they will continue to support.

I do think the invasion is why Intel hasn't committed to spinning off the fabs. That being said, despite what Intel says publically, I believe they are still going down the path of a spinoff. Just not committing to it just yet. Whether Intel can actually afford to stay on the leading edge without regaining their 90% server market share is kind of the question.

There already are sanctions preventing China from acquiring EUV scanners, so I think ASML would already be prevented from providing them service for ones they acquired through "other" means.

It is crazy to think there wouldn't be sanctions if China invaded Taiwan. Heck, I think there's a fair chance the US would become involved militarily - though that would be tricky as both sides would want to avoid it escalating into nuclear. At minimum the US would be providing intelligence and arms (which is already being done) It is folly to think that both sides would continue trading as normal if China took such a disruptive step.

 

[eek2121]

Intel hasn't even demonstrated they can compete with Samsung at this point, people here think they can compete with TSMC?

People here seem to think they are going to beat TSMC?

Intel's current best node:

• barely competitive with N5
• has barely any products using it
• 3 year later than TSMC N5
• can't produce GPUs

At least Samsung had a 5nm class node 2 years ago with somewhat competitive products. If it wasn't for sanctions Intel would be 4th place behind SMIC today.

Intel 4 has a density greater than that of TSMC N5 and they are about to beat TSMC to “2nm” class nodes by more than a year. How exactly are they behind again?

EDIT: Intel has crammed 14 cores into something like 40mm2. AMD with Zen 4 needed almost twice that space for 8 cores. I am trying to find the exact density number again, but I am pretty sure it beat TSMC N5 when it comes to released products from
AMD. I have not checked Apple’s listed densities to see if they are better.

 

[FlameTail]

Intel 20a is more than twice as dense as Intel 4 (> 400 mm2 density on test chips) apparently.

Interesting. What is that density for? Logic only? Logic-SRAM-Analog mix? HD cells?

Density of Apple A17 Pro on N3B is 184 mm². This is a mixed density value with SRAM, Logic and Analog.

 

[trivik12]

There are others beyond the big 3 or 4 who could still use Intel IFS and that itself could add sizable revenue. Any OEM making products for US Govt that requires locally sourced chips, Intel is supposedly making a custom product for Amazon for AWS and I am sure they will work with other large cloud providers as well. We will know sometime this year about the customer who prepaid for 18A access. Rumor is IFS customer product will be the 1st one out for 18A ahead of even Clearwater forest or any Intel client product for 18A.

 

[DavidC1]

Are you suggesting that Intel 3 is identical to the Intel 4 we're seeing today in Meteor Lake?

No. Usually delays cascade, but a totally new line avoids that, because it doesn't have to deal with market forces.