I really don’t know why people keep taking CCP propaganda at face value. Taiwan is the global leader in chip lithography. It took South Korea a good long while and a hell of a lot of money to get to rough parity. The PRC isn’t going to have as smooth of a time as SK for multiple reasons.
this post was submitted on 05 Jan 2024
221 points (96.6% liked)
Technology
59569 readers
4136 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
It's also worth noting that not only can China not make 5nm but America can't either. It's literally just Taiwan and SK that make sub-7nm chips.
When you say "they can't" do you mean "they haven't constructed a facility and hired people that can do this" or do you mean even if they did those things they would not be able to?
The reason I ask is I have been in several discussions on here where people have insisted it is the second.
There are barriers to suddenly having sub 7nm chip production. The EUV etching laser setup is hundreds of millions of dollars per machine and are made by one country - The Netherlands - who are aligned with American and NATO trade ideals.
The issue also isn't getting one; maintenance and software are required for each machine and are also strictly guarded. It's why places like China and NK don't have advanced chips but have rocket programs. It's gatekeeping their progress.
Taiwan literally bet it's future as a country on advanced technology and it paid off, probably the only one that did. SK has chaebols to soak up the cost and Samsung did exactly that with the help of the government in the 80s to compete with Japanese DRAM. It worked so well Japan stopped making most chips and SK took over.
USSR, India, Germany, America, Japan, Bulgaria, Vietnam, etc. have tried to start advanced chips technology centers but it doesn't succeed for nmone reason or another, typically due to the long time frame and high costs making it unsustainable.
Edit: it's worth noting that China is trying to make the investment and there may be some gains but etching accurate circuit paths at that level of detail (with multiple passes that exponentially explode complexity) without EUV lithography is nigh impossible. Currently they are at 20nm reliably and it puts them around 10 years behind current mainstream computing. (An equivalent is the 22nm chips from 2008.)
Very interesting, thank you.
The facilities cost tens of billions of dollars and the equipment is export restricted. The PRC physically cannot build a facility without spending literally a trillion dollars to develop their own equipment.
Tsmc is building a 5nm facility in Arizona, it will cost $40B.
I get that it is expensive. My question was "if they..." not can or would.
I've had several interactions where people say "even if you moved a facility to China magically and brought all the people and they were willing, it would not work".
Now, I have no love for China but surely that can't be true.
It's not. They don't currently have this kind of technology, but moving a tsmc fab to china with all of it's personnel willingly 'switching sides' would make it possible for them to learn.
With enough time and money, anything is possible.
This is important, it’s expensive and hard but it also just takes a long time to make the foundries.
Intel fab does sub 7nm. Meteor Lake processors main die use Intel 4, and the gpu die uses TSMC 5nm.
Intel’s lithography process branding is intentionally misleading:
- “Intel 7” is a 10nm process
- “Intel 4” is a 7nm process
- “Intel 3” will be (it’s not in product yet) 5nm
This was done because Intel basically missed an entire generation, and AMD and Nvidia (via TSMC and Samsung) basically leapfrogged them. They’re playing catchup now, and this is marketing spin to make their stuff look better on paper by changing the numbers and removing the units so they’re not technically falsely advertising, just misleading.
Beyond the “Intel 3” node, the process names are Intel 20A and Intel 18A, which, despite the lack of hard information at the moment, I’m pretty suspicious are also misleading branding/marketing attempts, because the symbol for “angstrom” is Å, not A.
Its misleading because intel nodes at a given nm is a denser process. Finfet tech nm on its own is already a misleading number. Intels 10nm for example is a denser node than TSMCs 7nm. All companies use a non traditional method of measuring nm (as if they did, how does it make sense that a intel 10nm product have higher density than something that is "7nm") as chip transitior is traditionally defined by the distance between transistor to transistor. And denser = transistors are closer together.
Basically intel is changing its name because its competition more or less did so, because the actual transistor distance in a traditional definition is incorrect, because finfet on its own folds over and the manufacture considers the folding technique to be a transistor when in reality its more complicated then that.
An adjacent comparison would be like treating physical cores and threads the same on a processor the same, or considering AMDs definition of a "core" during buldozer and piledriver to be the same as a traditional core (when in reality it isnt, because although there were 8 integer cores in an "8 core" processor, each pair shared one fpu) so it would challenge the definition. All fabs basically challenged the definition of what "nm" means, and theres no longer a standard to what it really means because they have all abandondoned the traditional naming scheme
Yep. I understand that. It’s still intentionally misleading.
If Intel could match the node size, they’d just call it what it is. They’re almost certainly going to catch up eventually, but this naming crap is 100% a marketing ploy. If you also consider the hilarious and asinine slide deck they put together somewhat recently (and then quickly took down after they were basically laughed out of the room by the tech community), it’s very clear they’re trying to keep up with the Joneses (AMD).
They have HUGE enterprise and consumer marketshare, so they’re clearly not going anywhere… but as someone who does actually understand the physics in play here (EECS), it’s embarrassing to see a company that was a market leader and pioneer for so many years sink to such frankly embarrassing tactics.
Im not necessarily saying what intel is doing is right. Im just saying that its a situation where they are ALL doing it, so everyone is wrong in the first place.
When someone technically says 5nm, its either they all have reached 5nm or none of them (based on how true you are to the definition of a nm process), and chooseing a mixed result means you fell into some companies marketing.
Sort of. The nanometer number is mostly just marketing, and Intels “4nm” is really somewhere between TSMCs 5nm and 7nm as far as density goes. They’re still a ways behind, which is part of the reason their chips are so inefficient comparatively
You are ware that the nm value used for samsungs and tsmcs process is also marketing right?
The whole point of intel changing their name on their process is to allign closer with samsungs/tsmc. Its why intel 4's physical transitior size and depth to next transitor is in the same ballpark as samsungs, despite officially being called something different.
Intels density falls closer to 5nm, and far from 7nm. Intels 10nm was already on its own about the same as tsmcs 7nm. Intel 4 is roughly TSMC 5N.
Thr main thing is intel hasnt caught up to bleeding edge yet (tsmc 2/3nm) and is about a year behind, but they have already passed TSMC 7nm a while back.
I know it's more complicated than just node size, but you're making it sound like intel cpus are roughly the same transistor density as current AMD cpus, so why is amd that much more efficient?
Phoenix is on 4nm, Intel 4 is more comparable to 5nm, so its definitely not what ive been saying remotely when ive explicitly said that Intel 4 is comparable to tsmc N5 node.
If you're thinking about desktops, 0 desktop processors are on the Intel 4 node, so you couldn't even have a point of reference on it. If you want to compare Intels 10nm (renamed version being 7nm) to TSMC 7nm, that would be like comparing Alder Lake/Raptor lake to strictly Zen 3(AMD 5000 desktop cpus).
Do you have a source for that? According to WikiChip Fuse, Intel 4 is comparable to TSMC N3 in density and offers better performance: https://fuse.wikichip.org/news/6720/a-look-at-intel-4-process-technology/4/
On paper, those PPA characteristics positions the company’s new Intel 4 process at performance levels better than TSMC N3 and Samsung 3GAE. On the density front, Intel 4 appears highly competitive against N3 high-performance libraries.
China didn't do shit. Taiwan did.
Think you might need to reread the comment you've replied to, they don't say anything about China doing anything, in fact they make it clear that China can't do it.
Oops
I feel a disruptance in the force, some Huawei are going to fall by themselves from hotel windows last floors.
What a boon it would be if CCP would agree to join Taiwan finally and join this dumb bickering that has lasted since 1949. Not sure how the Chinese people who have been brainwashed to think their system is better would take that, though.