Yeah, I'd agree with that. Expanding on that I'd say it's that they never ended up in a place where they had real lock in for the ISA, combined with there being better ISA implementations in each of the gate count niches. For any combo of gate count and licensing, Cortex M and R, SH, Xtensa, or PowerPC all made more sense for new designs depending.
Classic five stage RISCs are basically a commodity these days.
> “A license for all of mainland China has already been sold, lock, stock and barrel.” This is precisely what the U.S. government was trying to avoid when it intervened in the previous acquisition of MIPS Technologies.
Why does the U?S? Government care? The MIPS is a mature design that has suffered from neglect for many years and while its useful I can’t imagine how it could be any sort of strategic asset.
Sure, I believe I remember an avionics design win but can’t find the reference right now. Those design wins wouldn’t be affected by any such purchases (chips are fabbed by a third party and any procurement contract wouldn’t permit any mask change — quite the opposite).
But any new design would pick a new part anyway, and MIPS could simply be dropped from the candidate list — except there would be no reason to: this deal is China-specific and any US supplier would be buying a license from imagination — or from a merchant chip supplier.
Not likely to be many. They were never available in a rad hard form. Mil spec micros tend to be old platforms at this point. The most sophisticated was an old PowerPC and there is a VHDL SPARC clone that is popular for ASIC and FPGA implementations.
Do the people in charge of this in the US government have any clue? Quite possibly they don't. They may not understand the market reality that MIPS is a shadow of its former self. And they may think that CPU architectures are some kind of black magic that China is incapable of inventing on their own.
Even if the lower level staff have a good understanding of reality, that could easily be lost in the transmission by the time it filters up to the senior level
While China is very far behind in terms of chip manufacturing technology, the US government has essentially showed China that if they don't have it themselves, they'll be sanctioned at whenever and whatever.
Taiwan is distinct from the PRC - they have their own government, currency, and most importantly, a totally different relationship with the west that led to the rise of companies like TSMC.
Either your comment is about Taiwan having a legitimate claim to the name "China," or you're claiming that Taiwan is rightfully a part of the PRC.
The first one is intentionally awful communication, and the second one is a whole other can of worms.
> Technically, most countries are not allowed to recognize the ROC (Tawain) as an independent country but as a Rogue province.
If we are talking about international law, governments are free to recognise or refuse to recognise whichever foreign states or governments they want.
PRC tries its best to bribe and threaten other countries into not recognising ROC, and mostly succeeds. But that's realpolitik, not international law.
Recognition of a state does not imply recognition of all its territorial claims. (For example, the US recognises Australia, but the US doesn't recognise Australia's territorial claims in Antarctica.) So a state could choose to recognise both PRC and ROC simultaneously, while not recognising PRC's claims to Taiwan nor ROC's claims to the mainland. There is nothing in international law which prevents such a policy. The problem is political not legal.
ROC doesn't even seriously intend their territorial claims on the mainland any more. The only reason they don't drop them is that PRC would interpret that as an "independence move" and threatens to invade in that case.
The political reality is any country announcing recognition of ROC, even if it insists it still maintains its recognition of PRC, is going to have diplomatic relations withdrawn by PRC, and PRC will also try to retaliate in trade/etc.
> The one China policy was accepted when the UN removed the ROC in favor of PRC.
The UN's policies are not binding on its member states. The UN Secretariat has to follow the will of the majority of its member states as expressed in General Assembly resolutions. But each UN member state is free to choose a different position if that's the decision of their government.
We can all choose to do anything... but, there are consequences that come with those choices.
You chose to spread propaganda rather than concede that I'm technically right. This entire argument is pointless. TSMC pays Chinese Tax through their subsidiary in Shanghai. How can you claim they aren't Chinese? Do you even know what the C stands for in all your acronyms?
You do seem to be an expert in 'propaganda'. North America is the same as South America, because they both have America in the name, right? And I pay taxes on my Costa Rican subsidiary in Costa Rica...so Costa Rica owns the United States, right?
Do tell us about those 'consequences' your threatening. Gonna send us to one of your concentration camps out West for 'reeducation'?
> While China is very far behind in terms of chip manufacturing technology
SMIC, the leading fab in mainland China, is now doing 14nm. That makes them about six-seven years behind the leading edge. However, 14nm is good enough for very many applications. A lot of the most critical applications for national security, such as military embedded systems, are several generations behind that.
Anyway, design of CPU instruction sets and microarchitectures is at a different level of the stack from semiconductor fabrication – how advanced you are in one does not directly correlate to how advanced you are in the other. CPU design is about getting the maximum performance and features out of the budget provided by the fabrication process, power and thermal constraints, etc. How good the best fabrication process you have available to you is doesn't really say anything about how good you are as a CPU designer
This is why I feel that the “China steals all its tech” argument to be facile US propaganda. Simply having the tech secrets is a small slice of the pie when building a business.
If those people were right, little weak Taiwan is somehow able to withstand the force of Chinese industrial spy agencies to learn its secrets. That’s idiotic. China has vastly more resources and could easily get the secrets if it wanted. The fact is, building great processor manufacturing is extremely hard and knowledge has to be developed within to overcome future hurdles.
> stealing information always gives you a big head start if you're starting from scratch.
That's like saying cheating in school makes you smarter and more capable. That might sometimes be true, for example if it helps you reallocate resources more efficiently, but its highly context dependent.
There exists a biased, Western narrative that casts Chinese and East Asians in general as incapable of originality or novelty; that all they do is copy the hard work of Westerners. This is less true of Japan or Taiwan today, but it was very much true of Japan throughout the 1970s and of Taiwan throughout the 1990s.
If we stand back and take an objective look at how industrial and scientific "copying" actually works in practice throughout the world, including among the leading industrial nations, I think it would show the China narrative to be exaggerated. Moreover, the narrative, whether true or not, is dangerous. It causes us to underestimate their prowess, and to the extent we assume that by disengaging and preventing copycat behavior we'll stave off a rising China, we're in for a huge surprise. Not only will China continue to rise, but in setting up an Us vs. Them world we'll have accomplished part of their Divide & Conquer strategy that China will inevitably (and, indeed, already has begun to) employ. And the risks will be greater as without strong bilateral ties--including ties that appeal to the self-interests (e.g. money) of powerful domestic organizations--resort to military aggression becomes a more prominent and easier option for both sides.[1]
[1] Why did the U.S. invade (a 2nd time) and then conquer Iraq, and is constantly on the verge of war with Iran? Because sanctions don't just cut off a country from the U.S., they cut off the U.S. from that country, which means there's no domestic political pressure to avoid war beyond the enlightened beliefs of a diminishing minority. That's why Saudi Arabia constantly gets a free pass--because of the strong U.S. business interests in Saudi Arabia, not merely because of oil, per se, of which there exists plenty of alternatives around the world. Similarly, it's also why Saudi Arabia has always been much more circumspect with their support for terrorism, which is something worthwhile as Saudi Arabia might otherwise provide much greater support for terrorist organizations if they weren't so concerned with protecting their American relationships.
> This is why I feel that the “China steals all its tech” argument to be facile US propaganda.
So very wrong. China created the Thousand Talents Plan, with more than 7,000 overseas participants currently, to spy globally. Most of the recent US university and NASA arrests are related to that.
Arbitrary sanctions against Chinese companies (e.g. the war against Huawei) and seizures of Chinese assets (e.g. TikTok) are driven by a sizable faction in Washington DC that feels entitled to rule the planet uncontested. They would like to unwind the past 75 years of Chinese economic development, and push most of the Chinese population back into poverty, to make sure that an independent global power can never develop.
Chinese state-sponsored actors and corporate espionage have also been stealing technology from US companies so it's not like their hands are clean here.
Why is a Chinese citizen allowed to come to the US to buy property or start a company - but if I want to do the same thing I am forced by Chinese law to give half of it away to a Chinese "partner"?
China wants to have it both ways: concessions for 3rd world developing economies yet to be treated as a world power with its own technology and innovations. China wants the benefits of trade, yet manipulates the hell out of its currency.
The current US administration is certainly hamfisted and inept, no argument there. But even a broken clock is right twice a day. People are right to be upset about some aspects of China's policies. That doesn't mean punishing Chinese companies is a good solution, but the US is also under no obligation to just sit here and let China do as it pleases while talking out of both sides of its mouth.
The same activities, lack of copyright, harboring fugitives etc were how the US got its start, to the consternation of the then dominant superpower. From Lowell’s theft of the spinning mills to Dickens’ (and Tolkien‘s) complaints about lack of copyright, the US didn’t care until it was on top.
Not to defend the British whose IP was being pilfered, they’d done the same in their turn.
> are driven by a sizable faction in Washington DC that feels entitled to rule the planet uncontested. They would like to unwind the past 75 years of Chinese economic development, and push most of the Chinese population back into poverty, to make sure that an independent global power can never develop.
Cite for that? Especially that bit about the goal of pushing the Chinese population back into poverty.
My impression is the situation would be quite different if China was not run by authoritarian communists.
> They would like to unwind the past 75 years of Chinese economic development...
The entire reason that development existed in the first place is that the US allowed it despite completely unfair market access and more than a few provocations. Maybe if the CCP equalized market access relations would be better - but that's incompatible with their iron grip over society.
MIPS has excellent cores in the middle range (The "I class"). They can't compete at the high end with ARM, as high end need to be optimized for the latest processes in close cooperation with the leading fabs (TSMC and Samsung) and this is very labor intensive and so costly. So in the CPU IP world, it's ARM area. They're not the best for the very small cores either.
But in the middle, like the I-7200 or I-6500, they have very good cores and related IPs (like a coherent LLC). From last time I looked, still much more mature than anything from RISC V suppliers for example. If the I-7200 used the RISC V ISA (with extensions), it would likely be a reference in its class. Instead MIPS went with a new compact ISA that nobody cares about, too bad. Of course, the RISC V suppliers will mature over time and will at some point offer competitive or better solutions. But right now, the quality of some MIPS IPs shouldn't be underestimated. It would be too bad if this were wasted, but unfortunately it could happen. Hopefully the IP and people behind it will be acquired and the know-how will end up in coming RISC V chips. One can hope...
Presumably (as someone who doesn't really know shit about this) it's fully possible to make a MIPS chip that speaks Arm instructions? Instructions arrive and are decoded to uops, right? So is the instruction set itself software defined?
MIPS hasn't been neglected, though it's clearly not one of the new hotness and hasn't seen the attention that ARM has. Look at what companies like Octeon have produced.
If someone stepped up with nation-state levels of investment, I think you could see some very interesting outcomes. That's the concern.
The patents expired a long time ago, and MIPS has always occupied the ultra-low-end in things like routers and phones. RISC-V may take its place in the future, due to how similar it is.
SGI used MIPS to make big servers (and supercomputer clusters) back in the day - although I only ever got to play with an Origin 200[0] and the little Indy that was its console. More cost-effective than Sun hardware at the time (at least in the education market).
I have a first edition Hennessy and Patterson textbook on the mips architecture from my early 90s college days which I used to write a mips emulator. They BOTH worked on it as original innovators, Hennessy for Mips in specific, Patterson for being a main innovator of RISC architecture to begin with. The pair won the 2017 Turing award for their respective contributions to RISC.
54 comments
[ 3.2 ms ] story [ 129 ms ] threadThis all started at the time when ARM was ready to sell their cores on USB sticks to noname Chinese companies.
Classic five stage RISCs are basically a commodity these days.
Why does the U?S? Government care? The MIPS is a mature design that has suffered from neglect for many years and while its useful I can’t imagine how it could be any sort of strategic asset.
What am I missing?
But any new design would pick a new part anyway, and MIPS could simply be dropped from the candidate list — except there would be no reason to: this deal is China-specific and any US supplier would be buying a license from imagination — or from a merchant chip supplier.
Even if the lower level staff have a good understanding of reality, that could easily be lost in the transmission by the time it filters up to the senior level
Either your comment is about Taiwan having a legitimate claim to the name "China," or you're claiming that Taiwan is rightfully a part of the PRC.
The first one is intentionally awful communication, and the second one is a whole other can of worms.
The one China policy was accepted when the UN removed the ROC in favor of PRC.
If we are talking about international law, governments are free to recognise or refuse to recognise whichever foreign states or governments they want.
PRC tries its best to bribe and threaten other countries into not recognising ROC, and mostly succeeds. But that's realpolitik, not international law.
Recognition of a state does not imply recognition of all its territorial claims. (For example, the US recognises Australia, but the US doesn't recognise Australia's territorial claims in Antarctica.) So a state could choose to recognise both PRC and ROC simultaneously, while not recognising PRC's claims to Taiwan nor ROC's claims to the mainland. There is nothing in international law which prevents such a policy. The problem is political not legal.
ROC doesn't even seriously intend their territorial claims on the mainland any more. The only reason they don't drop them is that PRC would interpret that as an "independence move" and threatens to invade in that case.
The political reality is any country announcing recognition of ROC, even if it insists it still maintains its recognition of PRC, is going to have diplomatic relations withdrawn by PRC, and PRC will also try to retaliate in trade/etc.
> The one China policy was accepted when the UN removed the ROC in favor of PRC.
The UN's policies are not binding on its member states. The UN Secretariat has to follow the will of the majority of its member states as expressed in General Assembly resolutions. But each UN member state is free to choose a different position if that's the decision of their government.
You chose to spread propaganda rather than concede that I'm technically right. This entire argument is pointless. TSMC pays Chinese Tax through their subsidiary in Shanghai. How can you claim they aren't Chinese? Do you even know what the C stands for in all your acronyms?
Do tell us about those 'consequences' your threatening. Gonna send us to one of your concentration camps out West for 'reeducation'?
Just...sad.
You don't need reeducation... you need a first dose of education.
Just...dumb.
What does being a propaganda shill on HN pay these day?
Taiwan is officially the Republic Of China. CHINA. You are beyond dumb.
SMIC, the leading fab in mainland China, is now doing 14nm. That makes them about six-seven years behind the leading edge. However, 14nm is good enough for very many applications. A lot of the most critical applications for national security, such as military embedded systems, are several generations behind that.
Anyway, design of CPU instruction sets and microarchitectures is at a different level of the stack from semiconductor fabrication – how advanced you are in one does not directly correlate to how advanced you are in the other. CPU design is about getting the maximum performance and features out of the budget provided by the fabrication process, power and thermal constraints, etc. How good the best fabrication process you have available to you is doesn't really say anything about how good you are as a CPU designer
If those people were right, little weak Taiwan is somehow able to withstand the force of Chinese industrial spy agencies to learn its secrets. That’s idiotic. China has vastly more resources and could easily get the secrets if it wanted. The fact is, building great processor manufacturing is extremely hard and knowledge has to be developed within to overcome future hurdles.
That's like saying cheating in school makes you smarter and more capable. That might sometimes be true, for example if it helps you reallocate resources more efficiently, but its highly context dependent.
There exists a biased, Western narrative that casts Chinese and East Asians in general as incapable of originality or novelty; that all they do is copy the hard work of Westerners. This is less true of Japan or Taiwan today, but it was very much true of Japan throughout the 1970s and of Taiwan throughout the 1990s.
If we stand back and take an objective look at how industrial and scientific "copying" actually works in practice throughout the world, including among the leading industrial nations, I think it would show the China narrative to be exaggerated. Moreover, the narrative, whether true or not, is dangerous. It causes us to underestimate their prowess, and to the extent we assume that by disengaging and preventing copycat behavior we'll stave off a rising China, we're in for a huge surprise. Not only will China continue to rise, but in setting up an Us vs. Them world we'll have accomplished part of their Divide & Conquer strategy that China will inevitably (and, indeed, already has begun to) employ. And the risks will be greater as without strong bilateral ties--including ties that appeal to the self-interests (e.g. money) of powerful domestic organizations--resort to military aggression becomes a more prominent and easier option for both sides.[1]
[1] Why did the U.S. invade (a 2nd time) and then conquer Iraq, and is constantly on the verge of war with Iran? Because sanctions don't just cut off a country from the U.S., they cut off the U.S. from that country, which means there's no domestic political pressure to avoid war beyond the enlightened beliefs of a diminishing minority. That's why Saudi Arabia constantly gets a free pass--because of the strong U.S. business interests in Saudi Arabia, not merely because of oil, per se, of which there exists plenty of alternatives around the world. Similarly, it's also why Saudi Arabia has always been much more circumspect with their support for terrorism, which is something worthwhile as Saudi Arabia might otherwise provide much greater support for terrorist organizations if they weren't so concerned with protecting their American relationships.
So very wrong. China created the Thousand Talents Plan, with more than 7,000 overseas participants currently, to spy globally. Most of the recent US university and NASA arrests are related to that.
https://en.wikipedia.org/wiki/Thousand_Talents_Plan
Mainland Chinese people and the CCP don't believe their current system incentivizes workers to invent, hence the above program.
And the obligatory useful idiot link for you:
https://en.wikipedia.org/wiki/Useful_idiot
Why is a Chinese citizen allowed to come to the US to buy property or start a company - but if I want to do the same thing I am forced by Chinese law to give half of it away to a Chinese "partner"?
China wants to have it both ways: concessions for 3rd world developing economies yet to be treated as a world power with its own technology and innovations. China wants the benefits of trade, yet manipulates the hell out of its currency.
The current US administration is certainly hamfisted and inept, no argument there. But even a broken clock is right twice a day. People are right to be upset about some aspects of China's policies. That doesn't mean punishing Chinese companies is a good solution, but the US is also under no obligation to just sit here and let China do as it pleases while talking out of both sides of its mouth.
Not to defend the British whose IP was being pilfered, they’d done the same in their turn.
It was ever thus.
Cite for that? Especially that bit about the goal of pushing the Chinese population back into poverty.
My impression is the situation would be quite different if China was not run by authoritarian communists.
The entire reason that development existed in the first place is that the US allowed it despite completely unfair market access and more than a few provocations. Maybe if the CCP equalized market access relations would be better - but that's incompatible with their iron grip over society.
But in the middle, like the I-7200 or I-6500, they have very good cores and related IPs (like a coherent LLC). From last time I looked, still much more mature than anything from RISC V suppliers for example. If the I-7200 used the RISC V ISA (with extensions), it would likely be a reference in its class. Instead MIPS went with a new compact ISA that nobody cares about, too bad. Of course, the RISC V suppliers will mature over time and will at some point offer competitive or better solutions. But right now, the quality of some MIPS IPs shouldn't be underestimated. It would be too bad if this were wasted, but unfortunately it could happen. Hopefully the IP and people behind it will be acquired and the know-how will end up in coming RISC V chips. One can hope...
If someone stepped up with nation-state levels of investment, I think you could see some very interesting outcomes. That's the concern.
https://en.wikipedia.org/wiki/SGI_Tezro
SGI used MIPS to make big servers (and supercomputer clusters) back in the day - although I only ever got to play with an Origin 200[0] and the little Indy that was its console. More cost-effective than Sun hardware at the time (at least in the education market).
[0]: https://en.wikipedia.org/wiki/SGI_Origin_200
> It’s been a long, strange trip for MIPS. What started out as a university research project (under Dr. David Patterson, no less)
while https://graphics.reuters.com/USA-CHINA/TECH/yzdvxxdlnvx/USA-... has the more correct:
> MIPS was developed in 1981 by Stanford professor John Hennessy and his students