If this report about 2nm silicon from Japanese chip manufacturing start-up Rapidus beating TSMC is true, there might just be hope for GPU prices

News
By published

...and Intel 18A isn't really in the running, allegedly.

Rapidus 2HP wafer
(Image credit: Rapidus)

We reported recently on TSMC's relentless march toward 2 nm chip tech, despite the fact that there's barely any 3 nm silicon in our PCs. That's all very nice, but the problem is that it's just TSMC doing 2 nm with little to no competition. At least it was. Reportedly, Japanese chip manufacturing startup Rapidus has released the first information on its own 2 nm node. And it's allegedly better than TSMC's. Oh, and it's supposedly superior to Intel's upcoming 18A node, too.

Needless to say, genuine competition for TSMC would be a huge benefit for the chip industry in terms of competition and pricing. So, what exactly is being claimed for this new Rapidus 2 nm node? Known as Rapidus 2HP, it's said to offer a peak transistor density of 237.31 million transistors per square millimetre.

That compares to 236.17 MTr/mm2 for TSMC's upcoming 2 nm node, branded TSMC N2. In both cases, that's for what's known as high-density (HD) cell libraries, which can only be used for certain types of logic gates. Other elements of a chip, including SRAM memory and analogue circuitry, have different measurements, and so the overall density of a chip will never hit those peak HD figures.

Still, that applies to both Rapidus and TSMC. So if the figures are true, the former seems to have some kind of an edge. To be frank, just being competitive would be great. So, this Rapidus 2HP node would be very interesting even if it's significantly worse for elements like SRAM density.

The tricky bit here is that there are different ways of measuring transistor density, with some estimates putting TSMC N2 at over 300 MTr/mm2. But again, Rapidus doesn't need to be the outright best to be very appealing, especially when you consider that current GPUs from AMD, Nvidia and indeed Intel all use pretty old TSMC N5-derived tech and have yet to even move to one of TSMC's N3-based nodes.

Speaking of Intel, its 18A node is said to achieve around 185 MTr/mm2 density. Well behind TSMC and Rapidus. on paper, then. However, Intel's 18A node is optimised for performance and efficiency. That includes the novel addition of backside power delivery (BSPDN). BSPDN is intended to enhance performance-per-watt, but it also reduces overall density.

We'll get a first look at Intel 18A, supposedly, with the Panther Lake laptop chip at the end of this year. As for the Rapidus 2HP node, it's due to go into volume production in 2027, while TSMC's N2 node is meant to be going into volume production this year.

Earlier rumours suggested Apple's iPhone 17 series, which will likely be announced this month, could be the first devices to get N2 silicon. More recently, however, the consensus is that iPhone 17 will remain on TSMC's N3 node, so it probably won't be until 2026 when we see the first chips made with TSMC N2 technology on the market.

Rapidus IIM-1 foundry

Give it up for IIM-1, Rapidus's first 2 nm foundry. (Image credit: Rapidus)

In that context, Rapidus N2 silicon in 2027 would be timely enough to make a serious impact. If we were guessing, we'd doubt that the likes of AMD or Nvidia would immediately risk their main GPU lineups on Rapidus's unproven manufacturing. But you might see one or two isolated GPUs, maybe a low-end model made available on Rapidus 2HP to test the process out, and if that goes well, broader utilisation could follow.

Still, even if that's true, GPUs on Rapidus silicon are many years away. AMD and Nvidia's next-gen GPUs are expected to be based on TSMC N3 technology and will likely arrive too soon for Rapidus 2HP in any case. So, that probably puts that mooted test GPU in the following generation, with the generation after that the first realistic opportunity for a full lineup of Rapidus-made GPUs. And that, ladies and germs, takes us to 2031 at the earliest. So, not just around the corner, then.

However, if that sounds a little disappointing, do not despair. Word gets around the chip industry very quickly. And if Rapidus's technology is shaping up well, insiders will know all about it and it will very likely influence TSMC's pricing sooner rather than later. The mere threat of direct competition in the near-ish future will make it harder for TSMC to charge the earth. And that could be a very good thing for graphics card prices.

Asus RX 9070 Prime graphics card
Best graphics card 2025

👉Check out our full guide👈

1. Best overall: AMD Radeon RX 9070

2. Best value: AMD Radeon RX 9060 XT 16 GB

3. Best budget: Intel Arc B570

4. Best mid-range: Nvidia GeForce RTX 5070 Ti

5. Best high-end: Nvidia GeForce RTX 5090

TOPICS
Jeremy Laird
Jeremy Laird
Hardware writer

Jeremy has been writing about technology and PCs since the 90nm Netburst era (Google it!) and enjoys nothing more than a serious dissertation on the finer points of monitor input lag and overshoot followed by a forensic examination of advanced lithography. Or maybe he just likes machines that go “ping!” He also has a thing for tennis and cars.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.