Hacker News Reader

Comment by nomel | Hacker News Reader

You say that like chip fabrication is not the most complicated manufacturing process on earth.

You say that as if it didn't just turn out that China has been producing 7nm chips for at least a year.

Real Question is what yields have they been producing those chips at. Mining chips pay for themselves, so it's not as much of a big deal if your process is bad and only 20% of your wafer is usable. If you want to make a super computer out of chips from a process with 20% yields, then you're spending absurd amounts.

I have a question for you. Do you know that, or why, China stopped submitting new entries to the Top500, while numerous sources are claiming China brought online exascale computers with custom architectures?

Mining chips don't just simply pay for themselves. If it was cheaper to get the chips fabbed at Samsung or TSMC at any other lithography they'd be fabbed there. The company that made the miners isn't even Chinese. That it was fabbed there means that not only was there yield leftover from whatever state project was required, it was also commercially competitive.

This sounds interesting. Source?

I think it's generally accepted that this was a copy paste of the TSMC process. I think we'll have to wait to see if they can advance this, on their own. As the other person pointed out, a good indicator would be yield.

It's a copy past of the TSMC process because they were able to offer far higher higher salaries and attract TSMC talent. Given the reality that they have a lot of that talent and more each year and given the excellent track record of SMIC leadership, they will probably be able to advance on this, limited by the lithography machines they can procure.

If their yield was abysmal, no one would be willing to buy the chips since they aren't selling them at a loss and have plenty of public sector demand. The fact that they were able to sell them commercially and thus competitively means that their yields are okay at least.

> limited by the lithography machines they can procure

I would consider that a profound hurdle to advancement.

Nvidia has already moved to 5nm (released later this year).

It is a hurdle, but as new processes slow it is not insurmountable to raw power.

Only people without experience in semiconductor manufacturing imagine that it is possible to copy and paste a CMOS manufacturing process.

Even with perfect knowledge about how a process is done at TSMC, implementing it in another fab with different equipment requires a very large amount of work done by very competent engineers, to substitute some process steps and tweak the technological parameters for the others.

Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own.

Even within a single company, i.e. when having all the technical documentation and when having continuous access to all the people working there, it is extremely challenging to build a second semiconductor plant that will be able to make the same chips as a previous plant, with the same manufacturing process.

It took Intel many years of huge efforts until they have perfected a methodology to "copy exactly", to be able to build additional fabs for expanding the production of an already existing processor.

> very large amount of work done by very competent engineers,

Of course, and this is partly why they were successful: https://www.bloomberg.com/news/articles/2022-07-27/synopsys-...

> Someone who succeeds to reproduce a stolen manufacturing process is very likely able to further develop such a process on their own.

And, this is why we'll have to wait and see if their current, successful, strategy will result in future, independent, advances.

Yes, but once NVIDIA cannot sell to them, all the millions of dollars spent on NVIDIA products will now go to a domestic Chinese alternative. They will catch up fast.

They do have indigenous CPUs and GPUs.

Seems rather naive to assume that it for some reason would remain out of reach for a totalitarian state with a 1B+ population. And especially considering that there appear to be a few Chinese names on pretty much every major research paper released these days.

> remain out of reach

That's not how tech works. They will master and improve on 7nm, and probably future nodes. But the "reach" isn't 7nm, it's the continuous advancement. Rumors put Nvidia and AMD on 5nm, and Apple on 3nm.

As is, they've shown they can get something very close to the TSMC process working [1], partly by buying talent [2]. We'll have to see how that translates to long term improvements. But, they definitely can't rely on that strategy, long term.

1. https://www.theregister.com/2022/08/01/column_7nm_chips_chin...

2. https://www.bloomberg.com/news/articles/2022-07-27/synopsys-...

I think it's more nuanced. Left to develop it on their own, they might not succeed with a catch up effort in 5 years. Or 10. But in 20 years probably. 30 years certainly.

exactly - i think making commercially viable versions available for import into the country do more to disrupt their indigenous industry than banning the tech!

I don't think so. It's a little too convenient for the West to think this way, so there's a duty of skepticism. I think you can look at military stuff (fighter jets, spacecrafts, aircraft carriers), or civilian stuff like EVs, 5G network equipment, and see a clear ability to develop competitive and innovative indigenous technology when the sustained political will is there. It's bootstrapped off of imported platforms for sure but goes beyond what most other countries are able to produce.

> fighter jets

Eh, this, and probably others in your list (Nortel breach with 5G is a candidate), weren't really independently developed [1]:

> When the Soviets refused to part with their Su-33 design secrets, China purchased an Su-33 prototype aircraft from Ukraine, dubbed the T-10K-3, and quickly set about reverse engineering it.

> The J-20, ... America’s F-22 Raptor. Plans ... were stolen by a Chinese national named Su Bin, who was sentenced to 46 months in federal prison for his crime.

> Lockheed Martin’s F-35 Joint Strike Fighter was also compromised by Su Bin, leading to China’s J-31 program

I think it will take some time to see if they can become independent. Or, maybe independence doesn't matter! Copy and improve may end up being a successful strategy.

1. https://www.popularmechanics.com/military/aviation/g23303922...

They didn't start from scratch, for sure. But it has been copy and improve, not just let's make cheap copies lagging behind the original.

Everything in competition is about time and effort. No one cares if they figure it out 20 years down the road. It literally won't matter at that point, because the US will have much better by then. The world may be a different place, etc. It's all about the now.

I say "catch up" as in, covering the gap with where the US is by then. China has caught up in other fields already.

>The world may be a different place, etc. It's all about the now.

And that new world could be dominated by a new technology. Truth is that yes, with enough resources they can catch up technologically as many nations have in the past. That's no big surprise really, being able to produce ones own chips is so useful for defense you'd expect China and India to have that on a priority list somewhere. It wouldn't give them any advantage over other world powers

Reply on news.ycombinator.com