TSMC's upcoming 3nm node is critical for PC gaming TSMC's 3nm node is critical for PC gaming, with more powerful graphics cards on the way later next year. The good news is that it is back on track after a bit of a wobble.

TSMC's latest 3nm technology relies on complex multi-pattern extreme ultraviolet (EUV) lithography and was originally scheduled to begin mass production late last year. Ultimately, that did not happen until December, at the end of 2022.

Now, this delay had little direct impact on the PC game. The most obvious casualties were Apple's plans for the iPhone and the new Apple Silicon Mac. iPhone 14 Pro had to stick to 5nm for the new A16 SoC, which was only a very small step away from A15.

The new M2 series Mac chips are a disappointingly modest upgrade over the M1 family, very likely due to the 3nm node delay.

However, it was a bit of a worrisome time for PCs, with increasing reports that TSMC was having problems at 3nm. Will it be a disaster on the scale of Intel's 10nm? Lest we forget, Intel has struggled to get its new 10nm production node to work and has been stuck at 14nm for almost five years.

In TSMC's latest conference call with investors (opens in new tab) (via Anandtech (opens in new tab)), CEO C. C. Wei confirmed that 3nm will enter production and begin contributing significantly to the company's bottom line later this year, launching earlier than the 5nm node coming in 2020 He confirmed that.

"We expect the significant revenue contribution from N3 to begin in 3Q23, with N3 accounting for mid-single digit percentage of total wafer revenue in 2023... We expect N3 revenue in 2023 to be higher than N5 revenue in 2020, which is the fourth year.

However, we PC enthusiasts are not entirely comfortable with relying on 3nm for next-generation GPUs: the 3nm node TSMC is currently scaling, called N3B, is a very specific ultra-high-density technology and could be dedicated to Apple

The 3nm node is a very specific, ultra-high-density technology and could be dedicated to Apple.

Almost certainly, it will not be used by AMD, Nvidia, or Intel in their next-generation CPUs. Instead, TSMC is working on an improved 3nm called N3E, which is on the radar of Nvidia and AMD regarding GPUs. However, N3E is largely unproven, as it will not be in full production until late next year.

The key point of N3E, however, is that it is simpler to manufacture; N3B is an ultra-complex process, a 24-layer multi-pattern EUV process, whereas N3E is a simpler 19-layer single pattern technology. [As a result, N3B achieves 1.7x transistor logic density versus 5nm, while N3E scales only 1.6x. Still, this is a very healthy scaling factor, and yet N3E is claimed to be lower power and higher clocked than N3B

.

Looking further ahead, in December Wei stated that N2, TSMC's follow-on to 3nm technology, is on track for volume production in 2025.

If we follow the normal 2-year cycle for GPUs, Nvidia's RTX 60 series and AMD's RX 9000 family will be available at 2nm in late 2026. After that, everything becomes highly speculative.

Of course, TSMC is not the only game in town when it comes to GPU manufacturing, as Nvidia proved with its RTX 30 series; Nvidia has adopted Samsung's 8nm technology and AMD has stuck with TSMC and its 7nm technology.

Most observers judged Samsung's 8nm to be significantly inferior to TSMC's 7nm, but it was certainly good enough to make things like the RTX 3080 and RTX 3090 cards. Recently, Samsung described the yield of their 3nm process as "perfect" (open in new tab).

Even if this is like awarding an Oscar to your own movie, at least it bodes well for chip production in general. Samsung's 3nm node is not considered as dense or efficient as TSMC's 3nm technology. However, it could still be a viable option.

In any case, as it stands, the outlook for PC gaming performance looks pretty good at least until 2026 and two more generations of graphics cards. This must be good news. On the other hand, it is a bit difficult to say how much the price of graphics cards will fall.