News Russia says it's on track to manufacture 28nm chips in its own fabs by 2030 — the tech first debuted 15 years ago

[Admin]

Russia aims to begin local mass production of Elbrus-based CPUs using domestic 28nm-class process technology by 2030, despite sanctions, tool limitations, and talent shortages.

Russia says it's on track to manufacture 28nm chips in its own fabs by 2030 — the tech first debuted 15 years ago : Read more

 

[setx]

Elbrus is dead. RISC-V will bury it pretty fast.

They had their chance when there was a need for "domestic alternative" to closed x86 and ARM, but now it makes no sense to train people and optimize software for it.

 

[usertests]

It's bad for any bleeding edge chips, but 28nm is a long lived node.
https://omdia.tech.informa.com/om01...conductor-applications-in-the-next-five-years

Elbrus is dead. RISC-V will bury it pretty fast.

This 28nm process technology will purportedly enable MCST, the developer of SPARC-based Elbrus processors, to build CPUs with performance that meets the expectations of Russian companies.

If they can make their 8-core or whatever Elbrus CPU on 28nm, and it becomes mandated as a standard for secure computers in government, banks, etc., then it can have a purpose. Outside of Russia, there will be little interest since any Intel N100 or Rockchip or whatever will demolish it.

There's also a possibility that sanctions get lifted and Russia can reliably purchase TSMC wafers again.

 

[setx]

If they can make their 8-core or whatever Elbrus CPU on 28nm, and it becomes mandated as a standard for secure computers in government, banks, etc., then it can have a purpose.

It can't. Article correctly says that companies tried to use it but faced lack of support. And support won't appear out of thin air for something as specific as Elbrus even if government mandates its use.

 

[Siana Gearz]

Just like Transmeta Crusoe and Efficeon, Elbrus is made to emulate x86 chips in software, and just like those, they don't need a license. It was demonstrated to successfully execute Linux and Windows 2000 or XP (allegedly by now also 7).

Prior tests have shown it to be hiiiiideously slow though even when commission manufactured on a newer node in Taiwan.

I think it's an operation like any other in russia: pulling in government grants on the basis of need, and then later nothing works and nobody is at fault, because a chunk of money had to be diverted on bribes so quality of acquired tools and materials wasn't up to spec, and a bunch of people who signed up for the project have departed for excusable reasons, this is how it always goes. I don't think they'll have working 28n and I don't think they'll make anything good.

The same lab also has prior experience in SPARC compatible clone design. Another lab has experience producing a MIPS R5000-like design, but since it hasn't changed much in nearly 20 years, who knows whether they have the expertise to retool to RISC-V.

 

[uplink-svk]

Russia is currently 3 decades behind, harr harr, whilst us in the rest of Europe, we didn't even begin to race and most likely even won't any time soon. Everything worth mentioning is made either in US , Taiwan or soon to be "ex"-copy-paste China. What a time to be alive 😕

 

[aldaia]

Russia is currently 3 decades behind, harr harr, whilst us in the rest of Europe, we didn't even begin to race and most likely even won't any time soon. Everything worth mentioning is made either in US , Taiwan or soon to be "ex"-copy-paste China. What a time to be alive 😕

Ever heard about companies like Infineon? They have had 28 nm chips in production for several years.
Ever heard about a company named ASML? All companies producing leading-edge chips use their big (and expensive) machines.
Ever heard about IMEC? Their research is fundamental for companies like Intel, TSMC, Samsung, Micron.

Europe mostly lacks big leading edge companies. We have the technology but lack an industrial base to make use of that technology

 

[Alex/AT]

Would probably manage before 2300.
/s/