Intel CEO claims 18A node will at least match TSMC's N2 performance and beat it to market

[Daniel Sims]

Shots fired: As semiconductor manufacturers solidify their 3nm processes and intensify the race toward 2nm, TSMC and Intel have recently traded barbs over which company will have the superior node over the next few years. The Taiwan-based manufacturer is confident in its current path, but Intel aims to reassert dominance in the semiconductor industry by jumping to 2nm before anyone else.

Intel CEO Pat Gelsinger has claimed that its upcoming 18A process node (essentially 1.8nm) could outperform TSMC's 2nm chips despite launching a year earlier. The comments contradict recent claims from the Taiwanese competitor. Gelsinger made the remarks in an interview with Barrons.

He isn't sure if one node will significantly outperform the other but spoke optimistically about the company's release window. The outlet framed Intel's race with TSMC against the backdrop of US attempts to secure semiconductor supplies amid tension with China. As the market leader, TSMC provides the 3nm silicon for Apple's iPhone 15 and M3 Mac processors.

The company claimed that an upcoming optimized 3nm node called N3P will attain a power performance area comparable to Intel's 18A. The Taiwan giant expects to bring N3P into mass production in the second half of 2024 – around the same time as 20A (2nm) and 18A.

Additionally, TSMC is confident that its 2nm N2 node, slated for 2025, will outperform N3P and 18A. Following the company's inaugural 3nm process pattern, Apple could get first dibs on N2 and utilize it for the iPhone 17 Pro.

Much of Gelsinger's confidence in 20A and 18A lies in their introduction of the RibbonFET architecture – the company's take on gate-all-around (GAA) transistors and backside power delivery. These technologies will become crucial for companies manufacturing 2nm chips, enabling higher logic densities and clock speeds with reduced power leakage. Meanwhile, TSMC's N3P and other upcoming 3nm nodes will continue utilizing the mature FinFET architecture until it migrates to GAA with N2 a year after Intel.

Intel and TSMC aren't the only companies preparing to build 2nm semiconductors. Samsung also wants to enter 2nm mass production in 2025, while Japanese fabricator Rapidus plans to introduce prototypes by 2025, with mass production beginning in 2027.

Permalink to story.

https://www.techspot.com/news/101314-intel-ceo-claims-18a-node-least-match-tsmc.html

 

[Kashim]

Intel is years behind TSMC, but suddenly they're ahead and going to beat them. A while back he also said that AMD is now in the rear view mirror. Sure, buddy!

I think he means they're going to beat them with the naming scheme not the actual transistor size. You know, like how Intel 4 is actually closer to a 7nm process not 4nm like the name suggests.

 

[redgarl]

No CoWoS, no business...

 

[zamroni111]

Meanwhile, Intel still needs tsmc n5 and n6 to produce most of meteor lake because Intel fake 4 can't

 

[Theinsanegamer]

Actions speak louder then words. Parts of meteor lake are fabbed on TSMC 5 and 6. 14th gen is literally just 13th gen with a 4. Shove 400w into it, who cares about efficiency!

Intel does not inspire confidence.

 

[Kashim]

Meanwhile, Intel still needs tsmc n5 and n6 to produce most of meteor lake because Intel fake 4 can't

This. If your company is so great at what they do, why do you have to outsource your most advanced chips to the competition? I can tell you exactly how to gauge which company has the best/most advanced process; it's the one Apple uses for their top end hardware. Until Intel wins Apple back as a client they got absolutely nothing to brag about.

 

[Atmajaya2020]

As node getting smaller for cpu, the price getting higher, but the performance not getting much better than predecessor..

 

[umeng2002]

The bad thing about Intel being the best in the market is a return to buying a new motherboard every six months. I'm absolutely surprised how AMD has trounced Intel in the platform features space since AM4 and AM5 launched. AMD's I/O was always garbage compared to Intel's before Zen. Now AMD spams PCIe lanes at you like a desperate drug dealer looking for high sales.

 

[wiyosaya]

Uh huh!

 

[Mr Majestyk]

Prat Gelsinger is a funny guy. He honestly believes his BS. Tell us again Prat how AMD is a nothing in your rearview mirror.

 

[FuryX]

Intel probably thought their renown 14nm was close to "1.4" nm and can beat 2nm from tsmc

 

[Kashim]

Intel probably thought their renown 14nm was close to "1.4" nm and can beat 2nm from tsmc

I think you mean their fancy 14nm+++++++++ process. They figured they could keeping coasting with the same tech forever with tiny improvements over the years. If any of you remember Intel was against moving to EUV lithography because it was unproven tech and way too expensive for them. I guess they figured no one else would take the dive and they could just sit on 14nm forever. This is one of the big reasons they're so far behind now.