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.

AnandTech Forums: Technology, Hardware, Software, and Deals

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

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.

AnandTech Forums: Technology, Hardware, Software, and Deals

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

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.

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


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

 

[mikegg]

Seems odd that in his post he said the external customers with Intel 18A will be HVM in 2028-2030. I’m pretty sure that should be 14A/-E.

That seems to make sense. No one is going to put their flagship products on Intel 18A. That's way too risky. Companies will put their 2nd/3rd tier products on 18A and those can use a non-leading edge node.

 

[jdubs03]

That seems to make sense. No one is going to put their flagship products on Intel 18A. That's way too risky. Companies will put their 2nd/3rd tier products on 18A and those can use a non-leading edge node.

Maybe? That’s where that reputation of reliability comes into play. Clearly, they’re gonna need to prove themselves if they want leading edge output.

But I still will reference this as the baseline scenario:

That to me reads as 1st/2nd tier, depending on the version of the M-series on 18AP. The rumor for A-series in 2028 is base on Intel 14A, Pro on TSMC A14.

 

[Doug S]

That to me reads as 1st/2nd tier, depending on the version of the M-series on 18AP. The rumor for A-series in 2028 is base on Intel 14A, Pro on TSMC A14.

I'm still skeptical of the idea that Apple would go all-in on Intel for nearly half the iPhone market all of a sudden. I suppose they could do that as plan A, but they'd have to have some sort of plan B with TSMC, but that's tricky to do now that TSMC is so capacity constrained. Maybe the backup plan would be they'd do like they've done before and that year's iPhones would use A21 again in the base range and A22P in the Pro.

Apple would have to have agreed upon ways to insulate them from yield risk. They could do some type of "known good die" pricing to insure the cost per working die stays within the agreed upon range. Parametric yield is a bit more tricky. Since it would be the lower end part they could afford to sacrifice frequency and/or power relative to the TSMC Pro SoC - though obviously Intel would want to avoid that as it would be a bad look!

I suspect Apple is kicking the tires with Intel on multiple fronts, but that doesn't necessarily mean they're driving that car off the lot. If Intel is able to demonstrate with 18A shipments over the next six months that they've finally turned the page on their problems then this all sounds a bit more believable. No matter how good 14A test wafers are performing today, committing a major launch that you have to tape out in ~18-20 months seems to need some assurance they can get from the prototype line to mass production at scale and at high yield that is reasonably close to the same stuff they're seeing from TSMC's A14 test wafers.

 

[marees]

Reminds me of the time, my daughter's school said during her admission to kindergarten

"We will raise fees by 10% every year — because your salaries increase by 10% every year"

hello! It doesn't !!!

https://twitter.com/x/status/2005430020249112712

 

[mikegg]

Reminds me of the time, my daughter's school said during her admission to kindergarten

"We will raise fees by 10% every year — because your salaries increase by 10% every year"

hello! It doesn't !!!

https://twitter.com/x/status/2005430020249112712

It does. TSMC N3 is running at max capacity with wait times. They should be increasing prices.

 

[Fjodor2001]

It does. TSMC N3 is running at max capacity with wait times. They should be increasing prices.

Or increase capacity (which they do but apparently not enough). Anyway, let’s hope we get some competition from Intel in the years ahead.

 

[mikegg]

Or increase capacity (which they do but apparently not enough). Anyway, let’s hope we get some competition from Intel in the years ahead.

Fabs take years to build.

 

[jpiniero]

Fabs take years to build.

And a ton of money.

 

[Fjodor2001]

Fabs take years to build.

And a ton of money.

Agreed. And that is all well known.

 

[LightningZ71]

Industry total capacity growth is HARD limited by the fact that there is only one VOLUME manufacturer of the high end lithography equipment needed to make post "7nm" or "5nm muli pattern" products. Volume is a VERY generous word as it's quite low.

Rapidus still doesn't count. No one else is making anything comparable.

 

[Kaluan]

Industry total capacity growth is HARD limited by the fact that there is only one VOLUME manufacturer of the high end lithography equipment needed to make post "7nm" or "5nm muli pattern" products. Volume is a VERY generous word as it's quite low.

Rapidus still doesn't count. No one else is making anything comparable.

So basically one monopoly is propped up by another.

 

[dangerman1337]

Reminds me of the time, my daughter's school said during her admission to kindergarten

"We will raise fees by 10% every year — because your salaries increase by 10% every year"

hello! It doesn't !!!

https://twitter.com/x/status/2005430020249112712

Isn't TSMC N3P/X basically 25K a wafer? Wonder how SF2X and 18A-P per wafer compare to it right now? I mean I find it hard to believe Nvidia would not consider SF2X for RTX 60 considering they where completely okay with Samsung 8nm with RTX 30.

So basically one monopoly is propped up by another.

Better hope that xLight startup works out even if it's unlikely.

 

[511]

Isn't TSMC N3P/X basically 25K a wafer? Wonder how SF2X and 18A-P per wafer compare to it right now? I mean I find it hard to believe Nvidia would not consider SF2X for RTX 60 considering they where completely okay with Samsung 8nm with RTX 30.

N3P is lot higher 18A ramp will take a full year and SF2 IDK

 

[dangerman1337]

N3P is lot higher 18A ramp will take a full year and SF2 IDK

Apparently SF2X HVM is sometime next year so for a RTX 60 release looking to be late 2027, I don't think it's an unreasonable expectation.

 

[maddie]

Industry total capacity growth is HARD limited by the fact that there is only one VOLUME manufacturer of the high end lithography equipment needed to make post "7nm" or "5nm muli pattern" products. Volume is a VERY generous word as it's quite low.

Rapidus still doesn't count. No one else is making anything comparable.

Very true statement. Weakest link argument.

I really think energy supply is another hard constraint. Realistic rates of growth for both Litho and energy are not mainly limited by money constraints, but by physical ones, and skilled human talent scarcities. You need to build the supply chains before or simultaneously building your desired output physical capacity.

In a world with spare capacity, throwing money at a problem works most times. When physical resources are scarce, extra money only causes prices to rise. Years later, the real world has reacted and returned to equilibrium.

I find this reality very hard to explain to financial people. When worlds collide I guess.