Live data from Hacker News

How China built its ‘Manhattan Project’ to rival the West in AI chips

japantimes.co.jp

91–100 of 645 posts

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#91
post #55

Earlier quoted context omitted.

They are nowhere close to beat ASML. This isn't a moat ASML can keep for long though. There can be alternatove technologies to achieve the same goal. So far only China has that incentive. The real problem is process scaling is slowing down. How many more generations of lithography machines will ASML design? Probably not many. This means there will be no edge left in 5 or 10 years, as eventually brute force will work…

> So far only China has that incentive. The US is close to having that incentive, if the rift between the US and Europe keeps widening. The Netherlands has one lever, but damn it's a long one.

ASML develops and ships their machines at the pleasure of Uncle Sam because the USA licensed them the tech and remains a crucial part of the supply chain intentionally. It's not a lever. It's a partnership that is mutually beneficial and neither side can really ruin the other without damaging themselves.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#92

With Nvidia scaling down their consumer GPU production [0] I wonder if we will see consumer GPUs shipping from China in the future. Western companies seem to be abandoning the consumer/prosumer market which will have bad implications for hobbyists and aspiring professionals down the line. [0] https://www.pcmag.com/news/nvidia-might-cut-rtx-50-gpu-suppl...

You don't need CUDA for gaming but software is still just as big of a moat. Gaming GPU drivers are complex and have tons of game-specific patches. With their new Radeon/RDNA architecture it took AMD years to overcome their reputation for having shitty drivers on their consumer GPUs (and that reputation was indeed deserved early on). And I bet if you go read GPU discussion online today you'll still find people who avo…

Its not really just that AMD drivers are not that great (they are not) but they have been stable for a long time.

Its that nvidia relentlessly works with game developers to make sure their graphics tricks work with nvidia drivers. Its so obvious you miss it. Look in the nvidia driver updates they always list games that have fixes, performance ect. AMD never (used?) to do this they just gave you the drivers and expected developers to make their game work with it. The same strategy that MS used for their OS back in the 90's.

Thats at least how things got where they are now.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#93

A "Manhattan Project" would be building some shocking new technology that didn't previously exist. If they're cobbling together old parts, it sounds more like something you'd to to keep things running in case a conflict erupts: > The availability of parts from older ASML machines on secondary markets has allowed China to build a domestic prototype

A “manhattan project” can just mean a massive secret scientific war project? Seems apt.

and a critical nation scale ambition

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#94
post #35

Earlier quoted context omitted.

> scaling down their consumer GPU production [0] >> Due to Memory Shortages I don't think Nvidia wants to give up on consumer. They're a gateway into the overall AI ecosystem. Having feet planted there also make sure they can play the local game when that begins to blow up. Nvidia wants a robotics play, too. This is a pragmatic choice. And most of the money is in commercial.

> I don't think Nvidia wants to give up on consumer... This is a pragmatic choice. You mean, NV is after the money with a heavy heart and a sad tear or two over the abandoned consumers, like "We love you so much but sorry, we must go pragmatic on you"? > And most of the money is in commercial. This is a serious systemic failure and it's even wilder that it's accepted without question.

That's not the right light to view this chess move.

If Nvidia had infinite supply and infinite resources, they would absolutely continue doing consumer. There are constraints that prevent them from doing so at the typical volumes.

Giving up on consumer also means giving up on a gateway to more CUDA ecosystem users.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#95

Earlier quoted context omitted.

You don't need CUDA for gaming but software is still just as big of a moat. Gaming GPU drivers are complex and have tons of game-specific patches. With their new Radeon/RDNA architecture it took AMD years to overcome their reputation for having shitty drivers on their consumer GPUs (and that reputation was indeed deserved early on). And I bet if you go read GPU discussion online today you'll still find people who avo…

Its not really just that AMD drivers are not that great (they are not) but they have been stable for a long time. Its that nvidia relentlessly works with game developers to make sure their graphics tricks work with nvidia drivers. Its so obvious you miss it. Look in the nvidia driver updates they always list games that have fixes, performance ect. AMD never (used?) to do this they just gave you the drivers and expect…

The whole “improve a game’s performance on the driver side” thing: does AMD simply not do that at all? Or just far less?

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#96

Earlier quoted context omitted.

You don't need CUDA for gaming but software is still just as big of a moat. Gaming GPU drivers are complex and have tons of game-specific patches. With their new Radeon/RDNA architecture it took AMD years to overcome their reputation for having shitty drivers on their consumer GPUs (and that reputation was indeed deserved early on). And I bet if you go read GPU discussion online today you'll still find people who avo…

Cuda is 20 years old and it shows. Time for a new language that fixes the 20 years of rough edges. The Guy (Lattner) who made LLVM is working on this: https://www.modular.com/mojo Good podcast on him: https://newsletter.pragmaticengineer.com/p/from-swift-to-moj...

mojo been in the works for 3+ years now.... not sure the language survives beyond the vc funding modular has.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#97
post #55

I'd argue ASML's moat isn't the machine itself but the ecosystem: Carl Zeiss optics, decades of supplier relationships, institutional knowledge. This is clearly a significant achievement, but does anyone with semiconductor experience have a sense of how far "generates EUV light" is from "production-ready tool"?

They are nowhere close to beat ASML. This isn't a moat ASML can keep for long though. There can be alternatove technologies to achieve the same goal. So far only China has that incentive. The real problem is process scaling is slowing down. How many more generations of lithography machines will ASML design? Probably not many. This means there will be no edge left in 5 or 10 years, as eventually brute force will work…

> even if they use older machines and run longer exposure times

How do longer exposure times and older machines enable 2nm process nodes?

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#98

A "Manhattan Project" would be building some shocking new technology that didn't previously exist. If they're cobbling together old parts, it sounds more like something you'd to to keep things running in case a conflict erupts: > The availability of parts from older ASML machines on secondary markets has allowed China to build a domestic prototype

They are acquiring parts to reverse engineer them and build their own

> setting a goal of producing working chips on the prototype by 2028

They might be, but if they plan on getting a factory running in 3 years, they're presumably planning on using what they purchased.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#99

With Nvidia scaling down their consumer GPU production [0] I wonder if we will see consumer GPUs shipping from China in the future. Western companies seem to be abandoning the consumer/prosumer market which will have bad implications for hobbyists and aspiring professionals down the line. [0] https://www.pcmag.com/news/nvidia-might-cut-rtx-50-gpu-suppl...

Yeah PRC probably going to dump 1T into indigenize semi efforts by end of decade, but IMO good chance they're going to treat strategic semi as commodity utility business than make NASDAQ lines go up model. When western semi has 1st tier suppliers taking 30% margin, asml taking 50%, tsmc taking 50%, nvidia taking lol 70%, there's alot of fat on the pyramid to trim. PRC doing cost plus 10-20% will basically be able to…

Your analysis is out of date I think. This has already happened. Poor NXP just got their asses handed to them by the PRC. The fab they have in Italy looks nice but PRC has many of those.

Re: How China built its ‘Manhattan Project’ to rival the West in AI chips

#100
post #82

Earlier quoted context omitted.

Exactly. That's where this is heading, and the West-- as usual-- is pursuing quarterly profits and forgetting to look up.

Is anyone here calling legislators about it to inform them of this? Does anyone here have leverage to affect strategy?

this is one of those 'elections matter' cases. There's no strategy. Americans made it clear they want a country run by real estate crooks, crypto bros, gambling advocates and bizarre entertainment personalities. A country sized Las Vegas maybe and the beauty is the people get always what they deserve.
Post reply on HN