[36] TSMC Technology Interview: A16 Node, System-on-Wafer, and High-NA

[@TechTechPotato]

Transcript: open.substack.com/pub/morethanmoore/p/q-and-a-with-tsmc-on-next-gen-foundry

[@vohen7362]

Atomic layer deposition

[@azamatbezhan1653]

How do you think, is it possible to make 16nm metal pitch in forksheet design until 2028

[@FuzTheCat]

Moore's Law: "I don't care ... technology scaling ..." What a fantastic answer!

[@j340_official]

TSMC seems to be the opposite of Intel in terms of execution and focus. This is a great interview.

[@plugplagiate1564]

it is a pleasure to listen to one who knows what he is talking about.

[@TheParadoxy]

I really appreciate this content! Even if I don't understand most of the interviews, it gives me a goal for learning more.

[@nextcurve]

"I don't care." LOL!!! Great interview, Ian. Cheers!

[@nextcurve]

Great statement by Dr. Zhang that you can view scaling in terms of 2 dimensions. Chip-level scaling versus massive supercomputing system level scaling.

[@AAA.O]

The most important interview I've watched all year and maybe this entire decade 🔥🔥🔥🔥🔥

[@willberry6434]

Love hearing from TSMC

[@deeg_with_robots]

Very insightful and great questions Ian, thank you

[@Xevos701]

We love Potatoes (And chips) (And cheese too!)

[@Pavlobot5]

I'm curious how far away we are from all off die communication is optical. Like optical PCIe, or even Optical DDR paths. Would probably have to convert the signal back to electrical to go into the slots instead of having an optical connector.

[@gregandark8571]

Everything depends from the mastering of Gallium nitride...

[@cj09beira]

why no mention the sram problem, it would be cool to hear how they see that problem

[@wawaweewa9159]

photonics will seem amazing for low power applications

[@chriskaradimos9394]

great interview

[@whyjay9959]

Their SPR also involves metallization done to both sides instead of just the "front", right?

[@BloodAndDeathRaids]

Hey, I just want to mention that I love your channel, and the stuff you bring is amazing, like this interview. The only problem is that you need to improve the camera quality and the audio so it can be more enjoyable to watch. Good job overall.

[@andrewkinsey8754]

It's literally 4k, how big do you want it?

[@asknight]

@@andrewkinsey8754 it's almost always a mobile user who makes those kind of comments.

[@andrewkinsey8754]

@@asknight Ah that makes sense, as it looks and sounds great on my PC

[@bryce.ferenczi]

​@@andrewkinsey8754 Whenever its zoomed on Dr Zhang, the image is grainy from the less than ideal light conditions. Either a light or better camera would be needed to fix this. Its not about "resolution".

[@spicybaguette7706]

@@andrewkinsey8754 image quality ≠ resolution. The image is a bit grainy and the color balance is a bit off. I think it's recorded with a small camera/phone due to the graininess, which usually comes from a smaller sensor size. Sometimes 1080p footage can look a lot better than 4k due to the way compression works. IMO the audio quality is fine, which is the most important

[@CharlesShopsin]

Great interview! I think the big unanswered question we’re all curious about is, ”Is TSMC working closely with their customers?”

[@nnivo]

10/10. Moore please.

[@jrherita]

I interpreted what he is saying as “Moore’s Law is Dead”, which is interesting. He talked about performance and benefits, but didn’t mention cost. Moore’s Law is funded by cost.

[@Karthig1987]

Cool stuff

[@将軍九八.彁]

Does A16 utilize a four sheet NanoFlex design with a density around 444 MTr/mm2?

[@pjramk]

With the roadmap discussed, not sure how they can attempt to integrate High NA EUV before 2026 or 2027.

[@ps3301]

It is time to increase the wafer size to 18 inch!!

[@Fin1nishingMove]

So far, Intel's schedule seems to be ahead of TSMC when it comes to backside power delivery (called PowerVIA on the Intel side). It's set to debut with 20A, which by my will be an intermediary node intended to polish out the PowerVIA approach. Then, the first "major" node to use it will be 18A.

[@asknight]

Intel has much more important and contemporary issues to handle in a way to ensure they are still a viable company at the time those products would be released to packaging.

[@Fin1nishingMove]

@@asknight Intel's battle is against TSMC, and their main goal by far is to have the leading manufacturing node. Are you perhaps talking about the "degradation" issue? That's nothing but a blip that nobody will remember in a few months... Only KZbin/Reddit would make you think otherwise.

[@SasquatchsCousin33]

Anandtech has a good writeup laying out the different backside strategies. TSMC = most complex, Intel = medium granularity, Samsung = simplest. Its not apples to apples

[@asknight]

@@Fin1nishingMove I think you're severely underestimating the magnitude of the issue, and will continue to do so if you only listen to Intel's public releases because they are basically not saying anything substantive and quickly damaging their reputation. TSMC and their desktop market customer have a golden opportunity here if they can move quickly enough to seize the opportunity. Intel simply has not widened the scope of public notification because either they do not understand the root cause, or are panicking. If something as amateur as oxidation and degradation can eat their lunch in the desktop space, are we supposed to blindly trust them in the even more precise and critical markets?

[@Fin1nishingMove]

@@asknight I already explained why that's not an issue. You are free to believe what you want. Intel's RMA rates on desktop are similar to Ryzen 7000 series. If they were 50% or 100% like some sources claim (sources that are really damaging their reputation by reporting such content), you would be reading headlines in the Wall Street Journal. I see no need to derail this conversation further with the "degradation" topic.

[@tringuyen7519]

Dr. Zhang is the epitome of foundry business. TSMC’s has a plan & execute to the plan. Intel barely has an inkling & delay product launches over & over again.

[@__aceofspades]

You're thinking of Intel back in the 10nm era. Intel's 18A coming later this year, already demo'd is expected by many analysts and insiders to SURPASS TSMC's N3E and put Intel back in node leadership position. Also Intel has bought the majority of High-NA EUV machines while TSMC has sat on the sidelines. TSMC has completely fumbled their lead. Plus TSMC completely botched their Nevada fab rollout, to the point where they told their Chairman Mark Liu he needed to step down due to the failed fab rollout.

[@bingsterc7621]

@tringuyen7519 - Please NEVER bring up Intel in the same breath or sentence with TSMC. It's really DISRESPECTFUL to TSMC, because Intel has evolved itself into a GARBAGE/RUBBISH Corporation. Retail Customers and Business Clients SHOULD IMMEDIATELY STOP DOING BUSINESS WITH Intel. Everyone should altogether STOP SUPPORTING & PURCHASING Intel Products.

[@azamatbezhan1653]

How do you think, is it possible to make 16nm metal pitch in forksheet design until 2028

[@DileepB]

Morris's Law now!

[@crispysilicon]

Yay for breakfast potatoes!

[@EhNothing]

13:30 "....everyone can spell cowos, even the TV reporter can do that..." - are....are TV reporters considered not very smart there?

[@Smile-Linoq]

As a Taiwanese, I can tell you the answer is YES😂 Taiwan’s media quality is poor

[@raykyledecastro6083]

Asian Foundry Tim Cook, cant unsee haha

[@doubledigital_]

When we make chips mulit functional will be a game changer, as everyone knows we cant shrink memory any more than it is .. its why we have to stick it on the top.. but i can see a future where the gates or the cpu its self does multiple things at the same time using the same energy.. and if some random ass nerd can think of this idea im pretty sure these wizzards have thought about it already lol. moores law is not dead if you dont care about it :P

[@links7622]

I'm just a fisherman. I don't understand anything they are saying

[@pharmdiddy5120]

Wait 2nm now?? Witchcraft!

[@asknight]

Idk yet, but I have an itchy suspicion that A16 may indeed be 1.6nm!

[@Wild_Cat]

"muh moore's law!" who cares

[@trixniisama]

Thanks man, you contribute a lot to this discussion

[@photonboy999]

*Moore's Law...* STOP referencing it. There's just too much confusion about what it means. Many people think it refers to a linear rate of transistor density improvements. (as said in video) There's an old Whitepaper that talks about it in a different way. Some people just talk about the relative PROCESSING POWER. You should ONLY bring it up if you immediately define what you mean by it.

[@geoffreywon1706]

👍

[@Kurukx]

Who my Dr ... U are

[@husamabou-shaar9740]

It is disrespectful to not include your guest's name in video name, while all the video is an interview. This is not the case with other videos interviewing Western guests, which makes this uglier.

[@Marc-ww7cc]

What a bizarre comment. It's in the thumbnail and description.

[@husamabou-shaar9740]

@Marc-ww7cc it should be in the video title.

[@husamabou-shaar9740]

​@Marc-ww7cc And it's not in the thumbnail, the gentleman's name is Dr. Zhang, not tsmc, tsmc is the company name... actually, your comment proves my point. So thank you.

[@Marc-ww7cc]

@@husamabou-shaar9740 it's not your channel, you don't get to decide.

[@Marc-ww7cc]

@@husamabou-shaar9740 It is in the thumbnail, "Kevin Zhang: I Don't Care About Moore's Law". I think you're just digging yourself deeper into a hole. And in the description, "Dr. Kevin Zhang, SVP and Deputy Co-COO". You are making a mountain out of a molehill and a total fool of yourself in the process.