TSMC 2 nm Node to Enter Risk Production in Q4-2024, Mass Production in Q2-2025 if All Goes Well

[btarunr]

The cutting edge 2 nm EUV foundry node by TSMC is expected to enter risk product in Q4 2024, according to a report by Taiwan-based industry observer DigiTimes. 2 nm would be an important milestone for the foundry company, as it would be the first from the company to implement GAA (gates all around) FETs, the technological successor to FinFETs, which drove silicon fabrication node development for almost a decade, from 16 nm to 3 nm. The GAAFET technology will be critical for the foundry's journey between 2 nm and 1 nm.

TSMC is expected to risk-produce chips on its 2 nm node in its new fab at the Baoshan campus in the Hsinchu Science Park, located in northern Taiwan. Should all go well with risk production, one can expect mass production of chips by Q2-2025. Until then, refinements to the company's final FinFET node, the N3 family, will remain the cutting-edge of silicon fabrication. Samsung has a similar 2025 target set for mass production on its 2 nm node, dubbed SF2. Across the Pacific, Intel Foundry Services has its Intel 20A node, which implements GAAFET (aka RibbonFET) technology aiming for similar timelines, including an ambitious 2024 mass production target.



View at TechPowerUp Main Site | Source

 

[ratirt]

TSMC is ramping up quick with the new nodes I see.

 

[Denver]

TSMC is like a bullet train, which only gets faster and faster as they throw money at it.

Intel and Samsung are coal trains that only burn money, but barely move.

 

[AnotherReader]

TSMC is like a bullet train, which only gets faster and faster as they throw money at it.

Intel and Samsung are coal trains that only burn money, but barely move.

TSMC had issues with N3 as well so we'll see if they actually hold to this timeline.

 

[Daven]

I can see products based on TSMC 2 nm by the end of 2025. That’s two years after the introduction of products based on TSMC 3 nm.

Intel on the hand is all over the place with their ambitious node schedule. Intel 3 by the end of 2024 sure. Its just an optimized Intel 4 node. But unless Intel backtracks and redesigns Intel 20A to also be an optimized Intel 4, then I don’t see it happening until the end of 2025, beginning of 2026 as a full node shrink.

 

[Denver]

TSMC had issues with N3 as well so we'll see if they actually hold to this timeline.

It is true that they had initial problems with lower-than-expected yields, Maybe they've already resolved it. But at the same time, the competition didn't even appear in commercial products, from what I saw, TSMC's first 3nm generation is already equivalent or superior to Samsung's third 3nm generation (assuming it appears in some real product).

So, the fact is, the entire semiconductor industry is depending on TSMC to continue moving forward.

 

[AnotherReader]

It is true that they had initial problems with lower-than-expected yields, Maybe they've already resolved it. But at the same time, the competition didn't even appear in commercial products, from what I saw, TSMC's first 3nm generation is already equivalent or superior to Samsung's third 3nm generation (assuming it appears in some real product).

So, the fact is, the entire semiconductor industry is depending on TSMC to continue moving forward.

Without a doubt, they are the leaders. However, Intel was the unquestioned leader for longer and stumbled in the end. This is how Intel's troubles started; slipped timelines for 14 nm and that led to the disaster that was Intel 7.

 

[3valatzy]

Samsung are coal trains only burn money, but barely move.

That's what happened to AMD, in fact. Still stuck on 7nm with Navi 33? Right?

(assuming it appears in some real product)

Samsung has many processes.
7LPP, 6LPP, 4LPE, 4LPP, 4LPP+, 4HPC, 4LPA, 3GAE, 3GAP, 3GAP+. I wish AMD was smarter and designed a low-end GPU using a Samsung process.

Exynos 2400 is fast. And it really shows good gains over the previous Exynos 2200 - up to 45-50% higher performance https://nanoreview.net/en/soc-compare/samsung-exynos-2400-vs-samsung-exynos-2200

Take a look at the available products - https://semiconductor.samsung.com/processor/mobile-processor/exynos-2400/

 

[Denver]

That's what happened to AMD, in fact. Still stuck on 7nm with Navi 33? Right?



Samsung has many processes.
7LPP, 6LPP, 4LPE, 4LPP, 4LPP+, 4HPC, 4LPA, 3GAE, 3GAP, 3GAP+. I wish AMD was smarter and designed a low-end GPU using a Samsung process.

Exynos 2400 is fast. And it really shows good gains over the previous Exynos 2200 - up to 45-50% higher performance https://nanoreview.net/en/soc-compare/samsung-exynos-2400-vs-samsung-exynos-2200

Take a look at the available products - https://semiconductor.samsung.com/processor/mobile-processor/exynos-2400/

It was only a short time ago that Samsung managed to stabilize 4nm yields to an acceptable level for commercial use, AMD would not have time to move some designs to 4nm; it was probably already defined and ready for production when Samsung was still struggling to fix it's mess.

 

[3valatzy]

It was only a short time ago that Samsung managed to stabilize 4nm yields to an acceptable level for commercial use

Probably not applicable because they can sign a contract to pay only for the good dies, not for the junk wafers.

AMD would not have time to move some designs to 4nm; it was probably already defined and ready for production when Samsung was still struggling to fix it's mess.

Yes, let's stick with TSMC 7nm for a few years more. Until then the customers willing to buy something like Navi 10/23/33 will go down to net zero.

 

[Denver]

Probably not applicable because they can sign a contract to pay only for the good dies, not for the junk wafers.



Yes, let's stick with TSMC 7nm for a few years more. Until then the customers willing to buy something like Navi 10/23/33 will go down to net zero.

It's applicable because you don't understand how the chip development cycle works. The decisions were made a long time ago.

Yes, I suspect that sub-$300 GPUs will disappear over time. Nvidia has already abandoned it.
A 5nm design costs almost twice(1.85x) as much as a 6nm design, in addition to the cost per wafer.
Margins are becoming too tight to justify, I doubt chips like N33 will even pay for the development cost. If AMD continues to offer entry-level chips it will only be to gain market share, and it will always be at a node below the flagships.

 

[3valatzy]

It's applicable because you don't understand how the chip development cycle works. The decisions were made a long time ago.

The decisions are wrong. They had to be right when made.
I can't state what the details are about the chip design and manufacturing process (IP qualification -> architecture -> verification -> physical -> software -> prototype -> validation -> sales), but I see the outcome. Stagnation, inflation and things becoming worse by the day.

Yes, I suspect that sub-$300 GPUs will disappear over time. Nvidia has already abandoned it.

Nvidia hasn't abandoned anything.
RTX 3050 6GB 180 bucks.
RTX 3050 8GB 220 bucks.
RTX 4060 8GB 299 bucks.

 

[Denver]

The decisions are wrong. They had to be right when made.
I can't state what the details are about the chip design and manufacturing process (IP qualification -> architecture -> verification -> physical -> software -> prototype -> validation -> sales), but I see the outcome. Stagnation, inflation and things becoming worse by the day.



Nvidia hasn't abandoned anything.
RTX 3050 6GB 180 bucks.
RTX 3050 8GB 220 bucks.
RTX 4060 8GB 299 bucks.

Aside from the inherent challenges in manufacturing processes that inflate costs disproportionately relative to the benefits of increased density and efficiency, inflation can also be attributed to the following factors:

1. TSMC's lack of competitors, resulting in profit margins nearing 60%;
2. ASML's monopoly in supplying EUV equipment to TSMC and other fabs, further exacerbating market conditions.

Today, 3050 is an irrelevant product, it was launched 2 years ago; just like 1630, 6500xt, etc... these GPUs were originally created for laptops, they only came to desktop market to fill this gap; From then on, products under $300 disappeared. If the MSRP of a product is U$ 300 in the US, throughout the rest of the world it will cost much more than that;