Samsung Kicks Advanced-Chipmaking Race into High Gear with Road Map | Hacker News Reader

Samsung Kicks Advanced-Chipmaking Race into High Gear with Road Map

wsj.com

56 pointswallflower18 comments

I’m rooting for both Intel and Samsung. It sounds to me like Intel is ever so slowly getting back on track with Pat there (good interview on the Decoder podcast with him recently, if anyone is interested), but I’d guess we won’t really be able to see if Intel can get their crap together until ~2024-28, considering how long the lead time is for stuff like this. I don’t think Samsung had basically a failed decade like Intel did, but they haven’t been as quick or have as good yields on the last few nodes as TSMC has been, if I understand correctly.

TSMC being a node or 2 ahead has basically let them hold their customers by the balls, forcing them to take price increases and extreme lead times because there’s simply nowhere else to go for the same quality and performance. It would be great if we got to a place where we had 3 foundries on roughly equal footing, especially for national security and global supply chain robustness. But also because competition is good for us mere consumers as well, of course :)

It's great Taiwan is 2 steps ahead. Keeps the Hegemony aligned correctly.

I'm not sure why one or two process nodes ahead is considered such a national security issue. Is it like an encryption breaking ability or something else?

It's because it's not easy to catch up. In some cases, it can actually be impossible. Particularly if issues such as embargoes, sanctions and/or blockades can functionally block access to required equipment or talent. For example, it'd be hard to bring a new fab online when the ships or planes carrying the fabrication equipment get blown up.

Take a look at how Russia attempted and failed to scale up semiconductor fabrication into the 1980s and 1990s. They weren't 2 nodes back. They were WAY further back than that, and even today are still struggling to bring online nationally-developed fabrication technology of which the equivalent has existed elsewhere for 20-30 years.

China's SMIC is going to be bringing on "7nm" soon, despite being unable to purchase EUV equipment due to sales embargo. I can't imagine the effort that was required to make 193nm DUV machines produce chips reliably with a "7nm" feature size. EUV isn't required, but it surely makes things much easier.

SMIC 7nm size is actually 7nm in TSMC terms, or Intel 7 / original Intel 10nm. Which didn't use EUV at all.

Because it lowers demand for the domestic nodes outside of military, making them more expensive and reducing R&D budget.

Also even for non-military uses, if all of that market switched to foreign semis, it’d cause a lot of economic problems if we couldn’t import foreign semis any longer. Absent an embargo or war, that still gives countries that produce the better semis a lot of leverage.

1. Samsung ( Foundry ) has a culture and history of over promise and under deliver.

2. RoadMaps ( at least one like this ) are for investor only.

3. Samsung is also the reason why we have "nm" numbers that dont at least correlate to something. ( 99.9999% of the Internet blame TSMC for it ) And those numbers you see on roadmap means nothing.

4. But credit where credit is due Samsung is very serious about catching up to TSMC. Their total investment numbers dont lie.

IMO There simply isn't enough space for three leading edge Foundry. So either Samsung or Intel will have to let go. And Intel currently has the upper hand in terms of geopolitical and business backing. Which is what I think this article and PR to investor is all about. Samsung is not giving up without a fight. ( And why there are rumours about Samsung buying ARM from Softbank )

> 3. Samsung is also the reason why we have "nm" numbers that dont at least correlate to something. ( 99.9999% of the Internet blame TSMC for it ) And those numbers you see on roadmap means nothing.

I thought the reason for this is that the quoted nm numbers in a modern process node where pretty much for the most optimal kinda gate in the most dense layout (maybe something like SRAM?).

You mean the smallest feature size. That used to be the case until Samsung start playing around with the number a little bit more for marketing, and the whole market force TSMC to follow, and after a few cycles we left with nothing but a name with numbers.

You could probably still find some useless metric that fit the narrative. But that is beside the point.

> You mean the smallest feature size.

That is the word I am looking for.

> That used to be the case until Samsung start playing around with the number a little bit more for marketing, and the whole market force TSMC to follow, and after a few cycles we left with nothing but a name with numbers.

Huh I didn't know that, does anyone publish min feature sizes for modern process nodes?.

I'm so tired of these articles that regurgitate marketing claims as though they were guaranteed or representative of now. You can produce roadmaps all you want - Intel did so for years before finally acknowledging it was not actually achieving them.

It's no different from when some company gets articles saying "X will destroy Y!" when Y is shipping, and X is 6-12+ months out, assuming no schedule slippage.

vapourware is back in fashion!

https://archive.ph/3TEYV

thanks!

It's time for Tim Cook to pick up the phone and start working with Intel and Samsung to fab their chips in the future. We ave to remember that TSMC were never going to leapfrog Intel without the help of Apple's iPhone money. Apple can prop up Intel and Samsung the same way.

> TSMC were never going to leapfrog Intel without the help of Apple's iPhone money

it's not like Intel made it hard in the last decade either

Yeah even Samsung passed them, and they have like a million businesses going on to manage at the same time.

Intel has to work with them. Intel was under the religion of x86 everywhere for way too long.

At this point Apple won't work with Intel, because their M1/M2 is so much better than Intel's chips, why give them process experience and transistor-level knowledge of how the M1 is structured, designed, and performs?

Apple probably has the TSMC execs fully adapted to their culture of secrecy. Samsung and Intel would probably never cave to that properly: Samsung is their main smartphone competitor, and Intel is their competitor in desktop and probably ARM in the future.

Reply on news.ycombinator.com