News China doubles US research output on next-gen chips — chipmaking export bans are fueling a research wave

[Admin]

China is far outpacing US-based studies on next-gen chip technologies, with PRC-based authors making up 50% of the top-cited microchip research essays since 2018.

China doubles US research output on next-gen chips — chipmaking export bans are fueling a research wave : Read more

 

[2Be_or_Not2Be]

China increases research spending, and the US pulls back funding for science research & even local chipmakers. So who's playing checkers while the other is playing chess?

 

[mitch074]

China increases research spending, and the US pulls back funding for science research & even local chipmakers. So who's playing checkers while the other is playing chess?

Actually, the USA are playing chess expecting China to play checkers... But China is playing mahjong.

 

[JRStern]

Neuromorphic will be the thang, just as soon as anybody figures it out.
But, don't get too excited about citations, if China writes all the papers and they all cite each other, well, there it is..
Lots of published papers say nothing, or are just wrong.
US commercial companies have almost stopped R&D publications over the last 15-20 years.
Even stopped patenting anything important because patents reveal technology.

 

[rluker5]

Neuromorphic will be the thang, just as soon as anybody figures it out.
But, don't get too excited about citations, if China writes all the papers and they all cite each other, well, there it is..
Lots of published papers say nothing, or are just wrong.
US commercial companies have almost stopped R&D publications over the last 15-20 years.
Even stopped patenting anything important because patents reveal technology.

I read a while back that Intel already has neomorphic computing running.

 

[gg83]

LMao! This is a very funny headline.

 

[gg83]

Neuromorphic will be the thang, just as soon as anybody figures it out.
But, don't get too excited about citations, if China writes all the papers and they all cite each other, well, there it is..
Lots of published papers say nothing, or are just wrong.
US commercial companies have almost stopped R&D publications over the last 15-20 years.
Even stopped patenting anything important because patents reveal technology.

Nailed it. I couldn't say it as well as you did.

 

[DS426]

China increases research spending, and the US pulls back funding for science research & even local chipmakers. So who's playing checkers while the other is playing chess?

China's economy is going to be getting a lot less business from the U.S., so they'll have to pull that back at some point (or won't be able to increase spend as much as they otherwise would). Pulling back on regulations can spur investment and innovation as well.
Gotta love the hyper-focused cause-and-effect talking points. 😉

 

[smogfactory]

The US was warned that all their restrictions on China with regards to AI and chips would do is turbo charge chinese investment and development. Shock/Horror that is exactly what has happened.

 

[JRStern]

I read a while back that Intel already has neomorphic computing running.

There have been neuromorphic projects for fifty years, probably longer.
It's probably some flavor of in-memory computing.
Has to be mapped to existing problems and programs, like LLM.
Even then it might turn out to be done better with software on more generic hardware, where generic includes something like a B100/B200, but not exactly like.
Something like Intel processors with CXL, and with FPGA, that they spent all that money on eight years ago, might be a winner.
Or intelligent storage drives, relatively small ones used in parallel.

etc.

 

[Pierce2623]

There have been neuromorphic projects for fifty years, probably longer.
It's probably some flavor of in-memory computing.
Has to be mapped to existing problems and programs, like LLM.
Even then it might turn out to be done better with software on more generic hardware, where generic includes something like a B100/B200, but not exactly like.
Something like Intel processors with CXL, and with FPGA, that they spent all that money on eight years ago, might be a winner.
Or intelligent storage drives, relatively small ones used in parallel.

etc.

Turns out AMD seemingly did much better than Intel at gobbling up an effective FPGA player to pad out their lineup.

 

[TerryLaze]

Turns out AMD seemingly did much better than Intel at gobbling up an effective FPGA player to pad out their lineup.

It's not exactly gobbling up if you have to give them a big chunk of your company in shares...

 

[Mindstab Thrull]

China has like, over 4x the population of the USA. I'm not surprised more research is being done there. My question is, if you scale down to, idk, per capita, are they doing more and publishing more, or is it much closer?

 

[TerryLaze]

China has like, over 4x the population of the USA. I'm not surprised more research is being done there. My question is, if you scale down to, idk, per capita, are they doing more and publishing more, or is it much closer?

Don't forget that they have to do all the research again that the other countries already did years ago since nobody will give away their IP.

 

[mitch074]

Don't forget that they have to do all the research again that the other countries already did years ago since nobody will give away their IP.

Not exactly - as a lot of research is published, they can make an educated guess as to what's a dead-end, what's obliterated with IP protection, and what's interesting.
This is how BYD became a leader in LFP batteries : it was discovered elsewhere, a bit improved in Taiwan, deemed unusable in automotive industry and thus stuck in niche sectors... They managed to make them work, built a few cars around that, and now they outsell Tesla.

 

[TerryLaze]

Not exactly - as a lot of research is published, they can make an educated guess as to what's a dead-end, what's obliterated with IP protection, and what's interesting.
This is how BYD became a leader in LFP batteries : it was discovered elsewhere, a bit improved in Taiwan, deemed unusable in automotive industry and thus stuck in niche sectors... They managed to make them work, built a few cars around that, and now they outsell Tesla.

Sure, they still have to do a lot of duplicate research was my point, I was wrong in saying all the research, it's just a lot.

 

[subspruce]

Actually, the USA are playing chess expecting China to play checkers... But China is playing mahjong.

Actually they're playing Xiangqi, but that's beside the point.

 

[JRStern]

Turns out AMD seemingly did much better than Intel at gobbling up an effective FPGA player to pad out their lineup.

At the time Microsoft thought FPGAs were going to be the hot ticket for AI, which certainly propelled Intel into not only buying Altera but overpaying for it with a sloppy, poorly timed deal. Six months later nobody was doing AI with FPGAs. CXL was begun by Intel to support FPGA's - and whatever other coprocessors might come along. But now CXL is superseded by HBM, which is a debatable move. We'll just have to wait and see what "neuromorphic" ever ends up meaning.

 

[mitch074]

Actually they're playing Xiangqi, but that's beside the point.

Eh... Not exactly. Mahjong can accept more than 2 players, and there are many Chinese.
Also, no one can prevent a Chinese person from playing mahjong - they tried in Singapore, it's the only thing they couldn't regulate.

Sure, they still have to do a lot of duplicate research was my point, I was wrong in saying all the research, it's just a lot.

True. However, even if you buy / copy someone else's process, you still have to do a lot of re-engineering to really master it for yourself. It's a bit akin to the give a fish/teach fishing parable - more cost up front, much more profit later on.