Chinese RISC-V Project Teases 2025 Debut of Freely Licensed Advanced Chip Design (theregister.com) 83
China's Xiangshan project aims to deliver a high-performance RISC-V processor by 2025. If it succeeds, it could be "enormously significant" for three reasons, writes The Register's Simon Sharwood. It would elevate RISC-V from low-end silicon to datacenter-level capabilities, leverage the open-source Mulan PSL-2.0 license to disrupt proprietary chip models like Arm and Intel, and reduce China's dependence on foreign technology, mitigating the impact of international sanctions on advanced processors. From the report: The prospect of a 2025 debut appeared on Sunday in a post to Chinese social media service Weibo, penned by Yungang Bao of the Institute of Computing Technology at the Chinese Academy of Sciences. The academy has created a project called Xiangshan that aims to use the permissively licensed RISC-V ISA to create a high-performance chip, with the Scala source code to the designs openly available.
Bao is a leader of the project, and has described the team's ambition to create a company that does for RISC-V what Red Hat did for Linux -- although he said that before Red Hat changed the way it made the source code of RHEL available to the public. The Xiangshan project has previously aspired to six-monthly releases, though it appears its latest design to be taped out was a second-gen chip named Nanhu that emerged in late 2023. That silicon ran at 2GHz and was built on a 14nm process node. The project has since worked on a third-gen design, named Kunminghu, and published the image [here] depicting an overview of its non-trivial micro-architecture.
Bao is a leader of the project, and has described the team's ambition to create a company that does for RISC-V what Red Hat did for Linux -- although he said that before Red Hat changed the way it made the source code of RHEL available to the public. The Xiangshan project has previously aspired to six-monthly releases, though it appears its latest design to be taped out was a second-gen chip named Nanhu that emerged in late 2023. That silicon ran at 2GHz and was built on a 14nm process node. The project has since worked on a third-gen design, named Kunminghu, and published the image [here] depicting an overview of its non-trivial micro-architecture.
Huh? (Score:3)
Alibaba have previously open sourced their XuanTie riscv cores.
This isn't the earth shattering news you're looking for.
Re: Huh? (Score:1)
Perhaps a comparison would be useful?
It seems the key might be in the superlatives?
Re: (Score:2)
Perhaps a comparison would be useful?
A comparison of both RISC-Vs to a high-end ARM like Apple's M4 would be very interesting.
RISC-V in the data center would be revolutionary and benefit China. China should work toward a RISC-V ecosystem to rival ARM.
Re: (Score:3)
A comparison of both RISC-Vs to a high-end ARM like Apple's M4 would be very interesting.
I'm curious if Apple will release the M5 or not due to the Daystrom failure stigma. Or perhaps they will just opt to name it the MI5 and throw in some surveillance and spying technology.
Re: (Score:2)
Re: (Score:2)
Except that the XiangShan V3 "Kunminhu" is expected to offer three-four times the performance of a XuanTie T-Head C910, at the same clock speed.
That is comparable to cores from Apple, AMD, Intel and Qualcomm.
Re: (Score:2)
My understanding is that those are more application specific SoCs with various cloud features like AI accelerators on board. These are more general CPUs.
The licence is very good too. Haven't looked at the Alibaba one but this is comprehensive.
RISC-V has been getting there... (Score:2, Flamebait)
If one watches what new boards come from MILK-V, and how fast their SBCs gain features (even with antiquated SoC drivers and antediluvian Linux kernels), this is not a surprise. RISC-V doesn't have the patent issues that ARM does, and is a tried and true ISA, where it has been used for keyboard controllers, and a lot of MCUs for a long time because they are relatively cheap, and with the architecture being able to go from 16 bits to 128 bits, it can be adapted to be a server-grade CPU.
I am hoping China can
far from there... (Score:1)
they have all been experiments
the ISA is fine but there is a HUGE job actually scaling blocks like a branch predictor
this is hard mathematics that uses traces... that relies a lot on previous experiments going wrong i.e. the size of the experiments and the number
everything small is nothing like a complex system and previously would have been solved with specialised logic now the proven way is a MCU thats a Huge difference than a state of the art CPU
often all the news is established companies using the ISA t
Re:far from there... (Score:4, Funny)
You: Alex, I'll take punctuation, for $200.
Alex (in a booming voice from the ether): No way, You. Either learn to use it from the get-go or stop trying to weasel it out of me.
You don't say... (Score:4)
Re: (Score:1)
Who would have thought overly aggressive sanctions would result in domestic research and development? RISC-V won't have to beat existing architectures on performance, it just needs to be comparable enough in terms of energy efficiency.
If energy efficiency were the primary metric in data centers, we would probably be here talking about the world’s data centers slam full of Raspberry Pi Beowulf clusters running off solar panels.
Performance does matter. Why do you think AMD and Intel are still rival competitors.
Re: You don't say... (Score:3)
Re: (Score:2)
Raspberry Pi's aren't really very power efficient versus computing output.
Good point -- which would change it to something like "If energy efficiency were the primary metric in data centers, we would probably be here talking about the worldâ(TM)s data centers slam full of Mac Mini Beowulf clusters running off solar panels." That would highlight that capital costs are another big driver, and that even as good as Rosetta 2 is, familiarity with the ISA is another big factor.
Re: (Score:2)
Re: (Score:2)
RPi is for teaching software and hardware (Score:2)
They are functional general purpose and lower cost designs mainly for purpose of teaching Computer Science.
And Computer Engineering, don't forget the hardware side. It's an inexpensive single board Linux but it's also embedded Linux with tons of GPIO, I2C, SPI, etc for hardware. And it can power some of these attached devices to a degree.
Solar is supplementing coal in China (Score:2)
Raspberry Pi's aren't really very power efficient versus computing output. China has plenty of solar energy,...
Yet they are burning coal as fast as they can dig it up and import it. Coal usage is still on an upwards trend. Solar is supplementing coal, not displacing it.
"Share of coal in emissions 79% of total CO2 emissions from fuel combustion, 2022
Share of coal in electricity 61.7% of total electricity generation, 2022"
https://www.iea.org/countries/... [iea.org]
"China accounted for 95% of the world’s new coal power construction activity in 2023"
https://www.carbonbrief.org/ch... [carbonbrief.org]
"China’s coal-fired po
Re: (Score:2)
Re: (Score:2)
Not true: https://www.carbonbrief.org/an... [carbonbrief.org]
Sorry, it is true. You are not using an absolute metric as I am, you are using a percentage. Coal and solar use are both increasing. Therefore supplementing not displacing. The percentage metric is decreasing because solar is increasing faster than coal. But coal is still increasing.
Re: (Score:2)
If renewables increased 74TWh and the total used increased only 49TWh.
Those two numbers do not represent what you claim. They were measuring different things. so no, one being larger was meaningless.
Re: (Score:2)
If renewables increased 74TWh and the total used increased only 49TWh.
Those two numbers do not represent what you claim. They were measuring different things. so no, one being larger was meaningless.
They were measuring the exact same thing. TWh for electricity... But you still need to find excuses why 74 isn't bigger than 49. Good luck to you in your search. I eagerly await your findings.
You wrote: "China’s electricity demand in May 2024 grew by 49TWh from a year earlier. At the same time, generation from clean energy sources grew by a record 78TWh"
Total demand for electrical generation is not total coal use.
Clean sources for electricity generation is not total coal use. Its not even coal use for electricity.
You are not messing the same thing my sources are.
Re: (Score:2)
It's all there in the link...
China’s electricity demand in May 2024 grew by 49TWh from a year earlier. At the same time, generation from clean energy sources grew by a record 78TWh, (including a record rise from solar of 41TWh)
Other sources went down, since 78 is bigger than 49 in this reality...
Thank you for proving the two numbers do not have the relationship you claimed elsewhere. Demand increasing doesn't tell us how the mix of supply is changing. And non-clean sources covers more than coal. So no, you have not shown anything about coal itself.
Coal itself is only a part of non-clean energy sources. The data from multiple sources shows coal up, it didn't tell us what other polluting energy sources have done. Perhaps they went down, being displaced by coal, Imports of coal are increasing and
Re: (Score:2)
So coal is both "the biggest % of total energy"/quote> I said coal was the biggest % of the pollution.
Re: (Score:2)
.
So coal is both "the biggest % of total energy"
I said coal was the biggest % of the pollution.
Re: (Score:2)
With clean energy expanding by more than the rise in electricity demand, fossil fuel output was forced into retreat
You misrepresent all fossil fuel numbers as coal.
You misrepresent fuel for electrical generation for all fuel usage.
Coal went down 16 TWh
That is only electricity generation, not all coal use.
My numbers represent total usage of coal, not just electricity.
Re: (Score:2)
It's also a huge export opportunity. Billions of people are going to get new electricity grids this century, and they are going to copy the Chinese model of being well distributed and interconnected.
Europe could have done it, but we were too slow.
Re: (Score:3)
RISC-V is going to dominate the microcontroller industry. Chips cost literally nothing.
50 of them for $1.34 https://www.aliexpress.us/item... [aliexpress.us]
Re: (Score:2)
Chips cost literally nothing.
Stop using "literally" to mean "figuratively" (the opposite of literally).
Literally [wikipedia.org]
50 of them for $1.34
$1.34 is not "nothing".
Re: (Score:2)
You might want to read that definition again. Merriam-Webster is commenting on usage, which I assume is what that big "VIRTUALLY" means. They are explicitly saying that the word is often used in a way that makes the statement false.
RISC-V free on Raspberry Pi Pico ARM MCU (Score:2)
RISC-V is going to dominate the microcontroller industry. Chips cost literally nothing.
50 of them for $1.34 ...
They are free with the Raspberry Pi Pico microcontroller. The Pico 2 has been recently released, like the original it is ARM based, dual core. However there was a surprise addition, they added dual core RISC-V to the microcontroller, No price increase. Now it only run one dual core or the other, it selects ARM or RISC-V based on the the code that is flashed to it.
I'm really hoping they do the same with the full Raspberry Pi SBC.
Sanctions were proper, that's why Biden kept them (Score:4, Insightful)
Who would have thought overly aggressive sanctions would result in domestic research and development?.
Trump's sanctions against China were retaliatory, not protectionist. In other words they were the proper use of sanctions again an unfair trading partner. Not unfair in terms of dollars or amount of goods/services, but unfair in terms of openness, unfair in terms of predatory behavior.
That is why Biden kept them, an extremely rare point of agreement for a man prone to undue anything Trump had done. They were appropriate, and sill are.
Re: (Score:2)
Re: (Score:2)
Sure, my only point is that they aren't strategically effective and in turn end up making China a stronger adversary rather than a stronger economic collaberator.
I don't think so. Advanced chips through legacy chips are on the CCP's list of strategic technologies where the "plan" calls for domestic design and manufacturing.
Re: (Score:2)
Even if that was true, what has the practical outcome been? Have they corrected this unfair advantage, or have they just given China an even bigger advantage by making US products less attractive?
In a couple of years people will be complaining that China unfairly gives away its IP. Open source will be declared communist, which is hilarious when you think about it.
Well, good (Score:2)
Things have become far too stagnant anyways and Risk-V could change that. That it is China that makes these advances is no surprise. The West essentially forced them to do it.
Re: (Score:1)
China would be doing it regardless of Western sanctions. Have you not been listening to Xi and the Parasites (a new rock group) from the CCP over the years?
Re: (Score:1)
I have a working mind. You do not seem to.
Domestic design and production, CCP plan (Score:2)
I have a working mind. You do not seem to.
Actually he is correct. CPUs and MCUs are on the strategic technology list, things that are supposed to become domestically designed and manufactured according to CCP planning. It's a smart thing to do. Which is why the US and EU are starting to think/move about the design and sourcing of critical, and legacy, chips.
Re: (Score:1)
I think people in the west do not listen/read to Xi.
What is CCP btw?
Re: (Score:2)
An informal abbreviation for Communist Party of China, primarily used in the US. Surely you know that.
Re: (Score:2)
yep, the CPC initials are derived from the traditional format of "Communist Party" + [country name]. It's kind of similar to the naming scheme used to identify branch offices.
So, "Communist Party, China branch
Re: (Score:1)
It (CPC) is actually, the official name ...
You do not need to "derive party names" from anything.
The question is if the name has an "of" in the name, then the Country comes at the end (or behind the "of" at least) - basic English grammar.
So the Vietnamese party, is Communist Party of Vietnam. Because they named themselves like this.
But the Laos one is: Lao People's Revolutionary Party (LPRP). Again, because they named themselves like this.
And then there is the Cambodian People's Party - CPP.
In general it ma
Re: (Score:3)
Indeed - and now it leaves the West with a difficult decision to make. That is, if this new chip is good, should the West use it? Should the West get the benefit of it? If you're worried about 5G equipment, you should surely be worried about a CPU, so is it safe for the West to use? Is there a way to validate it?
Had the West created this innovation, we'd all be using it with abandon. We'd have maintained a certain level of 'control', and if anyone was planting anything nefarious in it, it would have been us
Re: (Score:1)
What do you mean, "decision"? Of course if that design is good, it will be used. All that 5G and "network equipment" panic is artificial and based on lies. The thing is, both 5G equipment and network equipment can and surely have been carefully examined for backdoors. The result of that is that US equipment (Cisco) is likely backdoored, thinly camouflaged as an incredible series of security-bugs and that major Chinese equipment that gets exported is not backdoored. Or we would have heard about it. Instead,
This is a neat upside of sanctions ... (Score:4, Insightful)
... vis-a-vis China. It forces them to take alternative routes and makes FOSS hardware a viable option. Hence big chinese money investing in it.
What would be totally hilarious if this actually pushes China ahead in the mid- and long-term, because they are forced to optimize along different metrics rather ever smaller scale of lithography or other cutting-edge and sanctioned stuff. After all, if you stay above and around the 20nm scale, you can produce comparatively cheap these days and don't need high end production hardware. In the end China might have cheap, manifold and robust premium-grade open source chipsets that roll up the market from below when the world runs out of some high-end microdevice that is critical to many industries.
I'm hoping for truly FOSS hardware and if China helps us get there I don't really have a problem with that.
Re: (Score:1)
What would be totally hilarious if this actually pushes China ahead in the mid- and long-term, because they are forced to optimize along different metrics rather ever smaller scale of lithography or other cutting-edge and sanctioned stuff. After all, if you stay above and around the 20nm scale, you can produce comparatively cheap these days and don't need high end production hardware.
Yes, but what you do wind up needing is a larger power budget to run the hardware, which means more nuclear and/or coal plants in China.
oh no I said something bad about China (Score:2)
Fifty cent moderation party to the rescue to fight facts!
Re:This is a neat upside of sanctions ... (Score:4, Insightful)
Yes, but Biden's recent sanctions have named Tencent and Huawai as Chinese Military entities so Linux Foundation (currently US) has to ban almost all the Chinese contributors to linux.
Linus went on little celebratory racist tear when they banned Russians but now the Russian and Chinese linux developers are likely to fork an EastLinux kernel and probably the RISC-V work will land there predominantly.
I think LF should move to Switzerland but for now we're going to have a ton of chaos to contend with due to these sanctions.
It's unlikely Trump will reverse hostility to China but maybe DoD can convince him how destroying linux is bad for NatSec.
Sanctions were retaliatory not protectionist (Score:3)
It's unlikely Trump will reverse hostility to China but maybe DoD can convince him how destroying linux is bad for NatSec.
The hostility is bilateral, predatory even from China. Those Trump sanctions, that even Biden kept - amazing considering how he hates anything Trump had done, are retaliatory in nature not protectionist. Unfair trade, in the openness sense not dollar amount sense, and predatory behavior based. A proper use of sanctions. Trumps position on trade is basically reciprocal. If you are open we are open, if you are closed to a degree we are closed to a degree, ... you chose, we reciprocate.
Re: (Score:2)
Biden's tariffs were different from Trump's.
Wrong.
"Biden will keep Trump's China tariffs, and add new ones on electric vehicles"
https://www.npr.org/2024/05/10... [npr.org]
Tariffs are protectionist by definition.
"Protectionist tariffs" and "Retaliatory tariffs" distinguish between an attack and a defense. When attacked, defense is warranted.
They are there to protect your country/workers/economy etc. Why twist yourself into a knot claiming otherwise? Seems a bit pointless.
No knots, attack vs self defense is a simple and different concepts. Even though both may use the same methods.
Sure that's why he put tariffs on steel from the entire world. All those countries were tariffing the US and it was retaliation...
Steel was being sold at a loss as part of predatory trade practices.
Re: (Score:2)
Americans on both sides no longer think free trade is in the country's best interest and are turning more protectionist... Trump is gong to protect America from Europe, Canada and Mexico next. But it's "retaliatory" and "not protectionist" because "reasons"
Trump does use the threat of tariffs as a negotiating tool. But with respect to what actually get done, Trump's trade perspective is reciprocal. Open both ways, or barriers both ways, but not open in one way with barriers in the other.
Re: (Score:2)
Steel was being sold at a loss as part of predatory trade practices.
LOL. Trump said himself he did it to protect American jobs and that it was vital for defense.
Pprotect against predatory and unfair trade practices. Trump's trade perspective is reciprocal. Open both ways, or barriers both ways, but not open in one way with barriers in the other.
Re: (Score:1)
RE: "It's unlikely Trump will reverse hostility to China but maybe DoD can convince him how destroying linux is bad for NatSec"
Maybe Musk can translate the situation for Trump, and reduce tech-related trade sanctions, but I expect Chinese FOSS HW is a done deal.
Not that much of a problem IMHO. (Score:2)
I'm pretty sure if there is a trusty Kerneldev in mainland Russia they will find a way to merge his code, and if it's only his Californian buddy doing the merges for him. At this level, sanctions are basically a non-issue.
Aside from that, most Russian devs and IT experts are expats by now. I spent New Year's with a whole bunch of them in northern Germany. Me and my buddy were a minority, everyone was speaking Russian.
They're from all over the place: Berlin, Cyprus, Sofia, Sicily, the Baltic, Turkey ... Just
Re: (Score:2)
China is also all-in on the industry version of the IoT. That's all about remote control of industrial plant and equipment, sensors, actuators, robotics etc. To scale that they need tons of cheap silicon, but not necessarily high density silicon.
It's also why their GPS-equivalent has 10cm2 resolution in civilian mode plus capacity for high density data packet transmissions - ideal for control of roaming robotic equipment.
Freely licensed... (Score:2)
Re: (Score:1)
Then copy paste it into google translate.
Or look for a Chinese site to translate it.
Google is actually absurdly awful in translating. Only Facebook is worth!
Re: (Score:2)
Maybe 5 minutes?
Re: (Score:1)
Unlikely. Especially if one of the target languages uses Hanzi/Kanji.
Or - in general - if an Asian language is involved.
Of course, you could "prompt" that to an "AI" (LLM) and try your luck. Would be interesting to know your results.
Defining "Open Source CPU" wrong (Score:2)