TSMC Confirms Suspicions: 2nm PC Chips Won't Arrive Until 2026, Lagging Behind Intel

TSMC Confirms Suspicions: 2nm PC Chips Won’t Arrive Until 2026, Lagging Behind Intel

Cellular Stockpile Tech Articles Leave a Comment

TSMC confirma las sospechas: sus 2 nm para PC no llegarán hasta 2026 y estará por detrás de Intel

It was an open secret and now the company has confirmed it. A new TSMC Roadmap that extends until 2026 shows the progress of the world’s leading semiconductor volume company, and again, delays are evident, except now they are official. Therefore, there will be no changes in 2023, and we will continue with the N3E, while in 2024, 2025, and 2026, TSMC will launch up to 8 different nodes on its Roadmap, focused on various markets.

Yes, TSMC will be late to the fight with Intel for performance in PCs and HPCs, but it will arrive on time for the 2 nm LP to meet Apple’s needs. The problem is that the BSPDN technology has even more delays, and interestingly, it will be optional like Intel, but it will also arrive late, so they have a challenging time competing in performance until 2026.

TSMC Roadmap through 2026: 8 nodes in three years, four of them late

As it is complex to explain and the graphic is very descriptive, we need to consider where each node comes from, that is, its base. Therefore, TSMC claims that the N5 is the starting point for everything that will arrive since the N3E as the main node comes from it, even though it is born from the N3, which was discarded and is omitted almost everywhere.

Hence, TSMC claims that the N3E represents a +18% speed improvement, over a 30% density enhancement, and a -32% energy reduction compared to the N5. These figures are more impressive than saying that the N3E only surpasses the original N3 by 5% in speed, logically. This is what we will see in 2023, but in 2024, things only slightly improve.

The most notable is not the N3P, which gets another improvement over the N3E: +10% speed (frequency), +4% density, and a 5%-10% energy reduction. The most notable is the N3AE, or N3 Automotive Early, which is the node aimed at the automotive sector in its early version, focused on the needs of electric cars. And with this, ladies and gentlemen, next year will conclude for TSMC, with no more changes. What will Intel have? Well, Intel 4, Intel 3, Intel 20A, and Intel 18A on the market, equivalent to 1.8 nm when compared to TSMC’s node.

In 2025, things become complicated for TSMC: N3X and N2 with a surprise

Indeed, because there are innovations, but not enough. The company mentioned in its symposium that in that year, the N3X will arrive, with a 15% increased frequency and an additional 4% density, but these data, like the previous ones, are based on the original N3, not on the predecessor nodes year by year. The N3X will be only 5% faster than the N3P with the same density, and it may slightly reduce power, but TSMC hinted that it would likely be the same.

That is, in a full year, they can only achieve a 5% increase in speed, why? Because efforts are primarily focused on the original N2. But beware, this is not an HD or HPC version, but a Low Power, or LP, version. In other words, it is a node for low-consumption chips, focused on smartphones, and probably to please Apple after a forgettable 2024, so all data refers to a Cortex-A715 instead of a high-performance node as such.

The data is based on the current N3E that we enjoy in 2023, and they indicate that the N2 will have a +10% or +20% (surprisingly uncertain for just a year and a half away) speed increase, over 15% density, and a +25% or 30% energy consumption reduction.

TSMC claims that the minimum voltage is lower than usual with an outlook for efficiency, but obviously, it will not be enough, and they will not achieve it because this node does not include BSPDN technology as such. However, it will enter the risk production stage by mid-2025, followed by a progressive volume increase.

N2 + BSPDN for the end of 2025

Like Intel, there will be two types of the original N2: the N2 without BSPDN that we just discussed, and the N2 + BSPDN, which will arrive at the end of 2025. It will be more expensive, but it will have the advantages of this technology that we have talked about so much.

Specifically, compared to the standard N2, TSMC announces a 10% density improvement and a +5% or 10% increase in speed, but they left this final range in the air since they are trying to improve performance even more.

The enhancements, according to them, will come through better efficiency and lower crosstalk of power and data signal delivery, indicating that they still need to refine BSPDN and that they are behind Intel by a year, or more.

In 2026, TSMC’s definite take-off: N3A and N2P, N2X

The year when high-performance nodes based on the original N2 will be launched. Late, very late, but hopefully worth it. And we say “hopefully” because few or no details have been provided.

It seems that these are minor improvements like the N3P and N3X that we will see in 2024 and 2025, so we should not expect any miracles. From 2026, we see what we already saw with Intel: Forksheets transistors (a GAA variant focused on density), CFET transistors (second-generation focused on density), 2D transistors, and CNT (carbon nanotubes).

Where will Intel be in that year? Far ahead. Intel 16A and Intel 14A are expected, equivalent to 1.6 nm and 1.4 nm, one focused on high performance and another for the LP sector. Therefore, TSMC faces a real challenge because, besides the nanometric leap, Intel will be ahead with Forksheets transistors if the leaks are correct, putting them between a year and a half and two ahead of TSMC. At least if TSMC’s Roadmap through 2026 is fulfilled and we hope there will be no further delays.

Need Help?

Can't Find What Your Looking For? Just fill out our form and we'll ask our trusted distributors

Contact Cellular Stockpile.

Wholesale Inquiries Only.

Email Cellular Stockpile
Tel: | Whatsapp:

Leave a Reply

Your email address will not be published. Required fields are marked *