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.

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


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

 

[moinmoin]

Interesting. So, what is Qualcomm doing, I would have thought that they, or Samsung, would have the fastest ARM SoC for Android.

They rely on Samsung's foundries instead on leading edge TSMC nodes as MediaTek does. 🐰

 

[Thibsie]

Troll mode (or/not) : Qualcomm has better anti-competitive tactics.

 

[Asterox]

MediaTek?! Yekhhhh....

Do not underestimate MediaTek, Qualcomm has become sloppy and in the near future it will cost them dearly.

Keep in mind, TSMC and Mediatek these are Taiwanese companies.

 

[DrMrLordX]

Qualcomm is releasing the Snapdragon 898 (or whatever they're calling it now) soon. It should be a decent upgrade over the Snapdragon 888. In fact Anandtech has a preview of it:

The Snapdragon 8 Gen 1 Performance Preview: Sizing Up Cortex-X2

www.anandtech.com

I am more hopeful that its thermals/power will be better than the 888. Snapdragon 888 (and Exynos 2100) both were hot enough to throttle pretty aggressively. Was not a good look for 5LPE.

 

[igor_kavinski]

Do not underestimate MediaTek, Qualcomm has become sloppy and in the near future it will cost them dearly.

Keep in mind, TSMC and Mediatek these are Taiwanese companies.

Their current products may have improved but this is what I have:

MediaTek | MT6739
ALCATEL 1X (2019) 5008U - Phone Specifications & Features (whatismyphone.com)

It works fine for 90% of my usage patterns. But when it slows down for weird unknown reasons, it REALLY freezes for multiple seconds. Just yesterday, it froze completely and wasn't responding to any touch input or button presses. I gave up and put it back in my pocket. Checked maybe 30 minutes later and it was back to normal.

 

[FlameTail]

Why isn't anybody talking about Samsung and IBM's collabrative VTFET- verticak transistor effort ?

 

[naukkis]

Their current products may have improved but this is what I have:

MediaTek | MT6739
ALCATEL 1X (2019) 5008U - Phone Specifications & Features (whatismyphone.com)

It works fine for 90% of my usage patterns. But when it slows down for weird unknown reasons, it REALLY freezes for multiple seconds. Just yesterday, it froze completely and wasn't responding to any touch input or button presses. I gave up and put it back in my pocket. Checked maybe 30 minutes later and it was back to normal.

Every other phone SOC manufacturer also have ultra-lowend models with only low-clocked A53 cpus and they are equally awful. For $100 phones Mediatek had already years ago chips like p60 which had 4xA73+4xA53 @2GHz. It was pretty decent SOC for lowend phones.

 

[BorisTheBlade82]

Qualcomm is releasing the Snapdragon 898 (or whatever they're calling it now) soon. It should be a decent upgrade over the Snapdragon 888. In fact Anandtech has a preview of it:

The Snapdragon 8 Gen 1 Performance Preview: Sizing Up Cortex-X2

www.anandtech.com

I am more hopeful that its thermals/power will be better than the 888. Snapdragon 888 (and Exynos 2100) both were hot enough to throttle pretty aggressively. Was not a good look for 5LPE.

The CPU part of the new MediaTek SoC is faster compared to the new Snapdragon 8 Gen 1.
One reason as has been said is using TSMC 4nm compared to Samsung. The other is their implementation of the small cores. They opted for separate caches and FPU per core while QC use the pairwise shared approach.

 

[Ajay]

Why isn't anybody talking about Samsung and IBM's collabrative VTFET- verticak transistor effort ?

I don't know. Why don't you start!

 

[DrMrLordX]

The CPU part of the new MediaTek SoC is faster compared to the new Snapdragon 8 Gen 1.
One reason as has been said is using TSMC 4nm compared to Samsung. The other is their implementation of the small cores. They opted for separate caches and FPU per core while QC use the pairwise shared approach.

Slightly off-topic (granted it is a showdown between TSMC and Samsung, more-or-less), but got any links?

 

[BorisTheBlade82]

Slightly off-topic (granted it is a showdown between TSMC and Samsung, more-or-less), but got any links?

The article you quoted and the AT article about Dimensity https://www.anandtech.com/show/17070/mediatek-announces-dimensity-9000

Inside you will also find the link to the AT article about the ARM architecture for 2022, were Andrei also explains the different little core approaches.

 

[hemedans]

Do not underestimate MediaTek, Qualcomm has become sloppy and in the near future it will cost them dearly.

Keep in mind, TSMC and Mediatek these are Taiwanese companies.

As long as mediatek continue to use Mali Gpu they will never compete with Qualcomm, Mali Gpu drivers are so Bad to the point highend Mali gpu perform worse than Midrange/Lowend Adreno Gpu.

 

[hemedans]

The CPU part of the new MediaTek SoC is faster compared to the new Snapdragon 8 Gen 1.
One reason as has been said is using TSMC 4nm compared to Samsung. The other is their implementation of the small cores. They opted for separate caches and FPU per core while QC use the pairwise shared approach.

Mediatek has faster cpu because of higher clock especially in middle cluster.

-Dimensity 9000 has 3 cortex A710 operating at 2.85 Ghz
-sd 8 gen 1 has 3 cortex A710 operating at 2.5Ghz

Also Cortex X2 in in dimensity 9000 run at 3.05Ghz compare to sd 8 gen 1 which run at 3ghz.

 

[Roland00Address]

Do not underestimate MediaTek, Qualcomm has become sloppy and in the near future it will cost them dearly.

Keep in mind, TSMC and Mediatek these are Taiwanese companies.

How is Mediatek working with supporting their hardware with drivers and so on thus allowing or not allowing Android updates? Same question after that but with unofficial open source where you brew your own if the main company is not supporting it?

 

[FlameTail]

So will VTFET allow them to make 3D Processors? Today's processors are 2D ( ie they are made on a flat silicon. ).

So thanks to VTFET we will get Cube-shaped processors ? [XD]

 

[DrMrLordX]

The article you quoted and the AT article about Dimensity https://www.anandtech.com/show/17070/mediatek-announces-dimensity-9000

Inside you will also find the link to the AT article about the ARM architecture for 2022, were Andrei also explains the different little core approaches.

I can see why you say that, but no benchmarks means meh

Let's wait and see actual devices with it before we jump to any conclusions.

 

[Tuna-Fish]

Why isn't anybody talking about Samsung and IBM's collabrative VTFET- verticak transistor effort ?

It's interesting but still quite far in the future.

So will VTFET allow them to make 3D Processors? Today's processors are 2D ( ie they are made on a flat silicon. ).

So thanks to VTFET we will get Cube-shaped processors ? [XD]

... no, but sort of a little. Present-day transistors are basically a line segment laying on it's side on a plane. VTFET would turn that line segment to point straight up. It's fairly straightforward why this would get you extra density. But you still cannot get more than one layer of those on a chip.

Proper 3d would require the ability to make high-quality transistors on implanted material. This does not work for silicon, because only the original grown crystal is epitaxial (all atoms line up to make up a single perfect crystal). When you implant silicon, it will end up as a jumbled mess. That still works for flash, because they only ever light up a very small portion of all the transistors at the same time, so it's okay for them to leak like a sieve, but is a total no-go for logic. More than a decade ago there were some interesting results about III-V materials that suggested that you could implant a competitive n-type transistor layer on top of a normal SiGe p-type layer, thus resulting in transistors in two planes on top of each other, instead of just one. Sadly, that didn't work out. I wish I knew why.

The interesting part about VTFETs, other than just improved density, is that on finfet processes you can only have one size of fin. To make a transistor that can supply more current, you have to just use more fins. This hurts optimization for many circuits because the jump from 1 fin to 2 fins is just too big. VTFETs essentially remove that constraint and bring us back to how it was with scalar transistors in that with them you can make the gate a bit wider or narrower to fit your needs. Usually, with new processes the design rules have gotten stricter and as a result the utilization has gone down a bit, thus hurting the actually achievable density. VTFETs should reverse that process, if even just a little bit.

 

[DisEnchantment]

I renamed the thread to include Intel discussions as well because of IFS 2.0.

While Intel's struggles may merit its own thread, the whole industry is starting to struggle lately and we can discuss the topics faced by the industry in general and upcoming innovations and so on.
Also going forward, the success of Intel is very important for the health of the industry and to some extent for computing needs of government agencies in the western world. (Also I have college mates at Intel and I wish them success)

Qualcomm seems like a great fit for IFS because of lesser conflict of interest at the moment. Or IBM.
Maybe more customers in the GAAFET era who knows.
So far only Samsung seems to have a "working" PDK for GAAFET with both Cadence and Synopsys having already certified their tool chain/libs/ips for the MBCFET.
But its anybody's guess how things will turn out.

From last IFS update, Intel 20A seems most interesting for me.




One of the good things that happened in the past couple of years is that many of the Pellicle problems are more or less addressed and Intel (and Samsung) can approach AMSL/Mitsui if they want off the shelf solutions to improve their yield and uptime.
(TSMC seems to have solved all of this in house and naturally not really keen on discussing the learnings)
The other is high NA EUV availability.

 

[Ajay]

Oh boy, more good news (not!) - hopefully it's a pessimistic viewpoint:

https://twitter.com/x/status/1473472017790009346

 

[Ajay]

Here is a summary of Intel's presentation at IEDM: Intel Scaling Innovations
One interesting slide is this one:



I had read about Forksheets, but didn't understand the implication for added scaling then. The clear presence here of a high-K dielectric, vs what I thought was poly, changes the narrative. Now, add CFET's to the mix (lower drive currents, it would seem, unless some high mobility exotic materials enter the mix). I'm not sure how CFETs work yet - I'd like to see a 3-D cell. Some of the slides are too small, sadly.

 

[uzzi38]

Plopping this analysis piece here as it gives a good breakdown as to the positions of all of TSMC's largest customers right now and their main issues.

TSMC Wants To Make Intel Dependent On External Manufacturing - Wafer Supply Agreement Insights For AMD, Apple, Broadcom, Intel, MediaTek, Nvidia, and Qualcomm

People all over semiconductor world have been speculating about the TSMC and Intel deal. SemiAnalysis wants to set the record straight on these. We are going to dive into the details for wafer supply agreements including prepayment terms and capacity agreements for AMD, Apple, Broadcom, Intel...

semianalysis.substack.com

 

[DisEnchantment]

Plopping this analysis piece here as it gives a good breakdown as to the positions of all of TSMC's largest customers right now and their main issues.

TSMC Wants To Make Intel Dependent On External Manufacturing - Wafer Supply Agreement Insights For AMD, Apple, Broadcom, Intel, MediaTek, Nvidia, and Qualcomm

People all over semiconductor world have been speculating about the TSMC and Intel deal. SemiAnalysis wants to set the record straight on these. We are going to dive into the details for wafer supply agreements including prepayment terms and capacity agreements for AMD, Apple, Broadcom, Intel...

semianalysis.substack.com

That prepayment from NVDA is bonkers. Probably just NVDA alone is financing 25% of TSMC's total projected capital expenditure in 1 FY.
Probably without that payment NVDA would be last in line for N5.

No wonder TSMC is building a ton of new Fabs at once. AZP1, F18P5/6/7/8, F14P8, F12P8/9 and coming 2H22 F20P1/2. The investment there is mind boggling.
I bet Intel would need to pay an even bigger amount if they are going to want substantial N3 wafers. Basically Intel is going to be partly financing the construction of one of the TSMC Fabs with their N3 down payment.

TSMC need to expand outside of Taiwan, otherwise one natural calamity and the world is going back to 20th century.

 

[Ajay]

I bet Intel would need to pay an even bigger amount if they are going to want substantial N3 wafers. Basically Intel is going to be partly financing the construction of one of the TSMC Fabs with their N3 down payment.

There we comments attributed to TSMC that Intel would have to pay to play. Not sure how that worked out. Very curious about what was discussed in Gelsinger's recent meeting with TSMC execs.

TSMC need to expand outside of Taiwan, otherwise one natural calamity and the world is going back to 20th century.

Yes. I waffled on this at first, but now believe that, in the US, we need to allocation some of our Semicon stimulus money to companies from other counties - so long as they start building plants in the US within a couple of years. Samsung and TSMC would certainly qualify and have already committed to new building new plant(s). Given the national security implications, the US military has been pushing for this. We also need to get more of the supply chain pipeline on US soil.

 

[maddie]

That prepayment from NVDA is bonkers. Probably just NVDA alone is financing 25% of TSMC's total projected capital expenditure in 1 FY.
Probably without that payment NVDA would be last in line for N5.

No wonder TSMC is building a ton of new Fabs at once. AZP1, F18P5/6/7/8, F14P8, F12P8/9 and coming 2H22 F20P1/2. The investment there is mind boggling.
I bet Intel would need to pay an even bigger amount if they are going to want substantial N3 wafers. Basically Intel is going to be partly financing the construction of one of the TSMC Fabs with their N3 down payment.

TSMC need to expand outside of Taiwan, otherwise one natural calamity and the world is going back to 20th century.

Agreed, but the 20th wasn't that bad. Better than the last 2yrs for sure.

 

[NTMBK]

Agreed, but the 20th wasn't that bad. Better than the last 2yrs for sure.

You know the Spanish Flu was in the 20th century, right?