Interesting to me because this is very likely to drive Linux support and usage in Mainland China!
Windows is unlikely to ever support LoongArch.
Had a very enlightening discussion with the authors as well, kudos to them for an amazing article!
What’s the game theory-ish take on this? Isn’t cutting off China’s access to chips counter productive in the long run, because of course they’re going to throw the country behind being self sufficient in this regard.
Is the process where cutting off an opponent’s access to a resource causing them to become more powerful a thing?
It is said that post WW2, one big stabilizing factor of the world is that we globalized and created interdependence on the 6 continent supply chain. It was in part due to technology gains, cheap energy and resources to power that tech and the fear of creating enemies that could now potentially fight back with nukes. A truly existential threat. If we all lean on each other to some degree, it works out better for most folks.
And now we see that this is rapidly degrading over the last decade.
I think the belief was that if the west cutoff China, they would be left in the dark. Instead they have more than enough knowledge on how to make these things that it became a driving force to active decouple from the west.
Provided the other economic/demographic/political issues don't become a major problem to China, this will definitely be a large blunder.
> I am fairly convinced now that in about ten years time, the tech war will be seen as a huge strategic blunder.
without tech war, China could progress much faster in emerging tech(AI?). But since there is no counterfactual data, there always will be people who use this to push narrative.
The blunder would be guaranteed if China spends a small fraction of the time and effort thinking out of box to look into other computing paradigms ... and relaxing some of the unnecessary limitation on individual rights, private properties, etc.
People will say that because everyone has an agenda, the point is to not stop China from building chips. You are naive if you think China wasn't already putting everything they can into building chips at home, did you really think China was okay buying chips from Taiwan in the first place?
The point is to slow them down and just make it harder to compete today, the US is pouring more money into chip research and manufacturing than the entire rest of the world combined, and nobody thinks the US is going to fall behind anytime soon. Not to mention the 5 other massive strategic disadvantages China has just purely on geography alone (i.e. energy policy).
TL;DR Who cares if China is self sufficient on tech in 10 years, they were gonna be self sufficient in 15 years anyways, and it's not like China is going to be loading up on Nvidia's stack forever.
If there was going to be a time to do it, it would be now. Any later would just not matter as they would be ramping up their own production anyways.
1. Stopping technology transfer worked exceedingly well for the west weakening the Soviet Union during the Cold War.
2. A distinct lack of (non-violent) alternatives for the West preventing China becoming the world's leading technological superpower (and hence also strongest military).
I doubt this will be successful in the long run, because China is not burdened by the Soviet Union's extremely inefficient way of organising its economy. Not to mention that China is the worlds biggest market.
The powers that be seem to actively be betting that China will collapse. Some geopoliticists predicted it would happen 10 years ago, but the arguments for it are somewhat compelling:
- China is facing a massive demographic bomb from one child policy
- China is a financial house of cards. Some of the stuff I've read about is 3 billion apartments built as investments/savings which are actually worthless, local/state debt and shell games that is frighteningly leveraged, and the general cheap/empty consumption/construction that state driven funding has produced
- China is turning sharply authoritarian, possibly headed to Stalinist levels of purges and cult of personality with Xi.
- China can feed its population and gets its oil due to free trade and shipping, quite extended supply lines, which are going to get a lot more chaotic in the future. If/when China attempts to invade Taiwan, the US will impose a naval blockade on its shipping (Zeihan claims a "couple destroyers in the Indian Ocean" ... probably a BIT more involved than that) ... China will starve and its economy will stop functioning.
I think this is part of a multipronged trade war to attempt to collapse the Chinese Communist party. If you want to talk about "powers that be", consider what is on the table here: the Putin regime will likely collapse from the failures in Ukraine, and Russia (who is facing a demographic bomb almost as bad as China's) is expending its last functional generation of youth on human waves in Ukraine.
So the two major nuclear powers might collapse entirely, and the only real economic rival to the US.
This is merely a demonstration of China's general competence level in CPU design and fabrication. Loongson is poorly resourced and has only niche applications and sales. Huawei and Alibaba are at a completely different level of resource availability in comparison.
It is a thing but it depends on the opponent's will, orgsnizational prowess, resources and human capital. The latter includes talent pool and scientific knowledge. Not all countries can pull it off. Most countries don't have the scale necessary, many don't have the organizational prowess.
Are you calling for the government to pick a winner? The Chinese word for this fierce if at times chaotic competition is "juan". It worked for them in EV and PV. The outcome remains to be seen in chips and commercial space launches. But even their mostly (ex-)students-run open source Xiangshan RiscV project https://github.com/OpenXiangShan/XiangShan shows a remarkable level of sophistication.
It might be about buying time - if our side develops some dominating technology before they manage to reach parity in chips, then they won't have time to catch up.
The other question is, is it morally right? For what legitimate reason do we have to cut them off of chips?
The literal only reason I see is jealousy. Power tripping.
The whole weapons thing is bullshit. You don't need the top of the line ic to power military equipment. This is really just petty jealousy and economic hording by the USA.
From my Chinese perspective, the time to ponder this question has long passed. It's not useful.
In a street interview in China (Xinjiang, even) about sanctions, an old man gave an interesting answer. He did not express resentment for the sanctions. Instead, he said: let them sanction, in the end we can't rely on others for our own prosperity and development.
He is right. There is no point in Chinese resenting the sanctions. There is no point in foreigners criticizing the sanctions, we all know it's no use. The only thing that matters now is for Chinese to focus on work. Focus on our own R&D. No need to say much, just get things done.
I wish I could find it again, but I remember reading about how Russians are using an old style chip-screen. I don't remember the name and can't find it. It's kinda like a FPGA but it's not reprogrammable and can be done cheaply on small batches. Really neat tech.
I think the idea is to starve the Chinese on AI advances, but there isn't much point spending a fortune training such models when you can have an insider leak it to you for much less money.
In 10 years time, China may come up with a production of 10-7nm. It may or may not be profitable to mass produce. By then, they will be 12 years into their 25 years Great Depression. Their real estate and stock market will still be in the toilet, and their demographics will go down 200-300m. They will have 30% of their population over 65. It would be iffy to say at that point whether the Chinese government has enough money, besides addressing their enormous debt, to keep pouring into a losing enterprise like this.
The western technology companies will be somewhere way way farther along/bigger by then - look at spacex flight this week for comparison at how fast western technologies move.
It's because of last year tensions around Taiwan and TSMC. Reliance on this company hit an all time high and tensions showed that it was strategically risky. So both side have been pushing local foundries projects via grants and other incentives.
How much do these CPUs cost? There'll be international buyers if they can beat Intel and AMD in terms of price to performance ratio, even if the chips are slower and consume more power. For general home and business computing Zen 1 performance is more than adequate especially since these are designed for Linux.
I guess it's not easy to even get your hands on it. Likely pretty low volume.
What kills it for the home and small business use is that it's MIPS. Just not worth saving 50 or so dollars of AMD / Intel premium for this trouble. As I understand, this CPU doesn't have an iGPU which makes it an even worse deal.
2 minor gripes!
1. Not MIPS, LoongISA, there are working Binary translators already, and they're commobly used in government office work!
2. They do have an iGPU, refer to Geekerwan's video!
Otherwise agree though, they're unlikely to go international at this point!
It seems Linux is going to win the desktop wars in China.
Windows is clearly never going to be ported to this oddball MIPS architecture so their stack will all be dependence on open source but sadly they will not return the favor by releasing their source or maybe my perception of China open source is biased.
That's not quite true and your bias is showing. The company behind Loongarch is reasonably friendly to FOSS and most of their LA64 changes (to Linux, llvm, binutils, gcc, etc) are upstreamed in a timely fashion. As for the chipset, kernel drivers are all open source although not all have been accepted upstream. I have personally been maintaining the out-of-tree Linux drm driver for their onboard gpu chip and it's available on github https://github.com/cl91/linux/tree/gsgpu-devel. If you have a barebone 3A6000+7A2000 setup, as far as I know the only closed source component needed is their userspace mesa driver for the onboard gpu.
+1 while I don't have any idea this deep in hardware engineering, in normal appdev Chinese open source contributions are well known and there's a whole class of up and coming apps which are great, just not internationalized well enough for others to use yet
The CPU does supposedly have instruction set extensions for faster execution of binary code translated from x86, ARM and RISC-V.
They think that it would be sufficient for running Windows/x86 apps at near-native speed in QEMU.
No ISA spec has been published. A few research articles are available from Chinese universities but most are only available in Chinese.
>or maybe my perception of China open source is biased.
FWIW The Chinese government seems to be trying to encourage open source, they even run their own equivalent of github called gitee. https://gitee.com/explore
See excellent article. The 3A6000 is Loongson’s best and newest and performs between the level of Zen (GloFo 14-12nm) and Zen 2 (TSMC 7nm) but limited to four cores.
A lot depends on the cost to produce that chip, whether Loongson can scale below 14-12nm they’re using currently (EUV without ASML?), and what development resources Loongson can deploy for their next microarchitecture revisions, which aren’t discussed here.
> The 3A6000 is Loongson’s best and newest and performs between the level of Zen (GloFo 14-12nm) and Zen 2 (TSMC 7nm) but limited to four cores.
Which is fine. For most applications, it's the order of magnitude of performance that matters. Not +10..30% or say, Zen 3 vs. Zen 2 (although gamers & AMD/Intel execs will argue otherwise).
Machines with >10y old cpu's are still perfectly useable for (most) everyday tasks, as long as the software support is there. Especially when coupled with enough RAM, decent gpu, SSD etc.
On the way from "pointless museum piece" to "latest & greatest", this cpu is 9/10s there.
They are certainly accelerating. I think the consensus is that it is within fabrication there is a "chokepoint".
In terms of design, China has world class companies. However in terms of fabrication and especially semiconductor manufacturing equipment, they are still somewhat behind.
Also x86 is very sticky - it is going to take them a decade to get rid of that.
> In terms of design, China has world class companies.
And not just companies. This is currently the world's top open-source RISC-V development https://github.com/OpenXiangShan coming from the Chinese Academy of Science.
(Whether China will remain investing in RISC-V, given that the US government has started to pressure US RISC-V development to limit their involvement with China is another question.)
Interestingly, they've got binary translators for both ARM and x86 which were demoed in Geekerwan's video. Apparently, they're apparently donating a compilation farm for a Debian repository.
Definetly interesting to see how Lithography will develop independently in China, and if it will ever pass the EUV chokepoint.
The chip uses LoongSon's own "LoongArch" ISA. They moved away from MIPS supposedly because of fears of licensing issues.
It is very much like the last revision of MIPS but modernised a bit: No more branch delay slot. No HI/LO registers. The arithmetic ops that trapped on overflow are also gone.
The article missed out on context. The chip runs LoongSon's own "LoongArch" ISA which they had developed out of MIPS'. Chips and Cheeses test programs are written in that ISA. There are other articles on the site that tell more about it.
LoongSon does however supposedly have a instruction set extension for running binary-translated x86 code faster (and ARM and RISC-V), but they have not published much about it.
Interesting to me because this is very likely to drive Linux support and usage in Mainland China! Windows is unlikely to ever support LoongArch. Had a very enlightening discussion with the authors as well, kudos to them for an amazing article!
What’s the game theory-ish take on this? Isn’t cutting off China’s access to chips counter productive in the long run, because of course they’re going to throw the country behind being self sufficient in this regard.
Is the process where cutting off an opponent’s access to a resource causing them to become more powerful a thing?
I am fairly convinced now that in about ten years time, the tech war will be seen as a huge strategic blunder.
It is said that post WW2, one big stabilizing factor of the world is that we globalized and created interdependence on the 6 continent supply chain. It was in part due to technology gains, cheap energy and resources to power that tech and the fear of creating enemies that could now potentially fight back with nukes. A truly existential threat. If we all lean on each other to some degree, it works out better for most folks.
And now we see that this is rapidly degrading over the last decade.
I think the belief was that if the west cutoff China, they would be left in the dark. Instead they have more than enough knowledge on how to make these things that it became a driving force to active decouple from the west.
Provided the other economic/demographic/political issues don't become a major problem to China, this will definitely be a large blunder.
1 reply →
> I am fairly convinced now that in about ten years time, the tech war will be seen as a huge strategic blunder.
without tech war, China could progress much faster in emerging tech(AI?). But since there is no counterfactual data, there always will be people who use this to push narrative.
The blunder would be guaranteed if China spends a small fraction of the time and effort thinking out of box to look into other computing paradigms ... and relaxing some of the unnecessary limitation on individual rights, private properties, etc.
In ten years time, China will be older and broker and with even less international friends. It would be a great strategic triumph.
5 replies →
People will say that because everyone has an agenda, the point is to not stop China from building chips. You are naive if you think China wasn't already putting everything they can into building chips at home, did you really think China was okay buying chips from Taiwan in the first place?
The point is to slow them down and just make it harder to compete today, the US is pouring more money into chip research and manufacturing than the entire rest of the world combined, and nobody thinks the US is going to fall behind anytime soon. Not to mention the 5 other massive strategic disadvantages China has just purely on geography alone (i.e. energy policy).
TL;DR Who cares if China is self sufficient on tech in 10 years, they were gonna be self sufficient in 15 years anyways, and it's not like China is going to be loading up on Nvidia's stack forever.
If there was going to be a time to do it, it would be now. Any later would just not matter as they would be ramping up their own production anyways.
1 reply →
By its proponents? It's already apparent to many. The current trade war started already over half a decade ago under Trump.
3 replies →
Powers that be are betting that the Chinese won't be able to catch up. These folks might be swallowing the "Chinese aren't creative" too long
An alternative explanation might involve both of:
1. Stopping technology transfer worked exceedingly well for the west weakening the Soviet Union during the Cold War.
2. A distinct lack of (non-violent) alternatives for the West preventing China becoming the world's leading technological superpower (and hence also strongest military).
I doubt this will be successful in the long run, because China is not burdened by the Soviet Union's extremely inefficient way of organising its economy. Not to mention that China is the worlds biggest market.
25 replies →
The powers that be seem to actively be betting that China will collapse. Some geopoliticists predicted it would happen 10 years ago, but the arguments for it are somewhat compelling:
- China is facing a massive demographic bomb from one child policy
- China is a financial house of cards. Some of the stuff I've read about is 3 billion apartments built as investments/savings which are actually worthless, local/state debt and shell games that is frighteningly leveraged, and the general cheap/empty consumption/construction that state driven funding has produced
- China is turning sharply authoritarian, possibly headed to Stalinist levels of purges and cult of personality with Xi.
- China can feed its population and gets its oil due to free trade and shipping, quite extended supply lines, which are going to get a lot more chaotic in the future. If/when China attempts to invade Taiwan, the US will impose a naval blockade on its shipping (Zeihan claims a "couple destroyers in the Indian Ocean" ... probably a BIT more involved than that) ... China will starve and its economy will stop functioning.
I think this is part of a multipronged trade war to attempt to collapse the Chinese Communist party. If you want to talk about "powers that be", consider what is on the table here: the Putin regime will likely collapse from the failures in Ukraine, and Russia (who is facing a demographic bomb almost as bad as China's) is expending its last functional generation of youth on human waves in Ukraine.
So the two major nuclear powers might collapse entirely, and the only real economic rival to the US.
8 replies →
This is merely a demonstration of China's general competence level in CPU design and fabrication. Loongson is poorly resourced and has only niche applications and sales. Huawei and Alibaba are at a completely different level of resource availability in comparison.
It is a thing but it depends on the opponent's will, orgsnizational prowess, resources and human capital. The latter includes talent pool and scientific knowledge. Not all countries can pull it off. Most countries don't have the scale necessary, many don't have the organizational prowess.
Resources spent re-inventing wheels is resources not spent gaining a lead.
Are you calling for the government to pick a winner? The Chinese word for this fierce if at times chaotic competition is "juan". It worked for them in EV and PV. The outcome remains to be seen in chips and commercial space launches. But even their mostly (ex-)students-run open source Xiangshan RiscV project https://github.com/OpenXiangShan/XiangShan shows a remarkable level of sophistication.
It might be about buying time - if our side develops some dominating technology before they manage to reach parity in chips, then they won't have time to catch up.
It's the opposite of "embrace, extend, extinguish".
The other question is, is it morally right? For what legitimate reason do we have to cut them off of chips?
The literal only reason I see is jealousy. Power tripping.
The whole weapons thing is bullshit. You don't need the top of the line ic to power military equipment. This is really just petty jealousy and economic hording by the USA.
From my Chinese perspective, the time to ponder this question has long passed. It's not useful.
In a street interview in China (Xinjiang, even) about sanctions, an old man gave an interesting answer. He did not express resentment for the sanctions. Instead, he said: let them sanction, in the end we can't rely on others for our own prosperity and development.
He is right. There is no point in Chinese resenting the sanctions. There is no point in foreigners criticizing the sanctions, we all know it's no use. The only thing that matters now is for Chinese to focus on work. Focus on our own R&D. No need to say much, just get things done.
6 replies →
I wish I could find it again, but I remember reading about how Russians are using an old style chip-screen. I don't remember the name and can't find it. It's kinda like a FPGA but it's not reprogrammable and can be done cheaply on small batches. Really neat tech.
I think the idea is to starve the Chinese on AI advances, but there isn't much point spending a fortune training such models when you can have an insider leak it to you for much less money.
7 replies →
In 10 years time, China may come up with a production of 10-7nm. It may or may not be profitable to mass produce. By then, they will be 12 years into their 25 years Great Depression. Their real estate and stock market will still be in the toilet, and their demographics will go down 200-300m. They will have 30% of their population over 65. It would be iffy to say at that point whether the Chinese government has enough money, besides addressing their enormous debt, to keep pouring into a losing enterprise like this.
The western technology companies will be somewhere way way farther along/bigger by then - look at spacex flight this week for comparison at how fast western technologies move.
SMIC is doing volume production at 7nm now.
7 replies →
Can anyone with more hardware insight tell me if the brunt of this story is actually true, and how related to this chip AMD's actions were?
> How a Big U.S. Chip Maker Gave China the ‘Keys to the Kingdom’
> Advanced Micro Devices revived its fortunes through the deal, and sparked a national-security battle
https://www.wsj.com/articles/u-s-tried-to-stop-china-acquiri...
https://archive.vn/V7cZq
Different company! Loongson runs their own MIPS based ISA which is self-developed.
IIRC that company is called Zhaoxin, which has had mild success getting ~Zen 2 performance domestically.
I'm not to sure why WSJ is citing that as a 'national security issue', everything related to China seems to be called that now.
Yeah it's American vs Chinese empire stuff at this point. New cold war and all that.
1 reply →
It's because of last year tensions around Taiwan and TSMC. Reliance on this company hit an all time high and tensions showed that it was strategically risky. So both side have been pushing local foundries projects via grants and other incentives.
How much do these CPUs cost? There'll be international buyers if they can beat Intel and AMD in terms of price to performance ratio, even if the chips are slower and consume more power. For general home and business computing Zen 1 performance is more than adequate especially since these are designed for Linux.
I guess it's not easy to even get your hands on it. Likely pretty low volume.
What kills it for the home and small business use is that it's MIPS. Just not worth saving 50 or so dollars of AMD / Intel premium for this trouble. As I understand, this CPU doesn't have an iGPU which makes it an even worse deal.
I have been trying to get my hands on one of these (in Australia) for about 2 months now but no luck. Tempted to give Ali Express a try on this.
https://www.aliexpress.com/item/1005006181125200.html
2 minor gripes! 1. Not MIPS, LoongISA, there are working Binary translators already, and they're commobly used in government office work! 2. They do have an iGPU, refer to Geekerwan's video!
Otherwise agree though, they're unlikely to go international at this point!
It seems Linux is going to win the desktop wars in China.
Windows is clearly never going to be ported to this oddball MIPS architecture so their stack will all be dependence on open source but sadly they will not return the favor by releasing their source or maybe my perception of China open source is biased.
> not return the favor by releasing their source
That's not quite true and your bias is showing. The company behind Loongarch is reasonably friendly to FOSS and most of their LA64 changes (to Linux, llvm, binutils, gcc, etc) are upstreamed in a timely fashion. As for the chipset, kernel drivers are all open source although not all have been accepted upstream. I have personally been maintaining the out-of-tree Linux drm driver for their onboard gpu chip and it's available on github https://github.com/cl91/linux/tree/gsgpu-devel. If you have a barebone 3A6000+7A2000 setup, as far as I know the only closed source component needed is their userspace mesa driver for the onboard gpu.
+1 while I don't have any idea this deep in hardware engineering, in normal appdev Chinese open source contributions are well known and there's a whole class of up and coming apps which are great, just not internationalized well enough for others to use yet
The company behind Loongson has been very FOSS friendly. There was a 2009 Google Summer of Code project porting Debian (or was it Gentoo?) to Loongson CPU. https://opensource.googleblog.com/2009/05/2009-google-summer...
The CPU does supposedly have instruction set extensions for faster execution of binary code translated from x86, ARM and RISC-V. They think that it would be sufficient for running Windows/x86 apps at near-native speed in QEMU.
No ISA spec has been published. A few research articles are available from Chinese universities but most are only available in Chinese.
>or maybe my perception of China open source is biased.
FWIW The Chinese government seems to be trying to encourage open source, they even run their own equivalent of github called gitee. https://gitee.com/explore
Its a perfect attack vector for Chinese hackers.
At this rate, China is gonna be able to catch up with Intel/AMD soon?
See excellent article. The 3A6000 is Loongson’s best and newest and performs between the level of Zen (GloFo 14-12nm) and Zen 2 (TSMC 7nm) but limited to four cores.
A lot depends on the cost to produce that chip, whether Loongson can scale below 14-12nm they’re using currently (EUV without ASML?), and what development resources Loongson can deploy for their next microarchitecture revisions, which aren’t discussed here.
> The 3A6000 is Loongson’s best and newest and performs between the level of Zen (GloFo 14-12nm) and Zen 2 (TSMC 7nm) but limited to four cores.
Which is fine. For most applications, it's the order of magnitude of performance that matters. Not +10..30% or say, Zen 3 vs. Zen 2 (although gamers & AMD/Intel execs will argue otherwise).
Machines with >10y old cpu's are still perfectly useable for (most) everyday tasks, as long as the software support is there. Especially when coupled with enough RAM, decent gpu, SSD etc.
On the way from "pointless museum piece" to "latest & greatest", this cpu is 9/10s there.
4 replies →
They are certainly accelerating. I think the consensus is that it is within fabrication there is a "chokepoint".
In terms of design, China has world class companies. However in terms of fabrication and especially semiconductor manufacturing equipment, they are still somewhat behind.
Also x86 is very sticky - it is going to take them a decade to get rid of that.
> In terms of design, China has world class companies.
And not just companies. This is currently the world's top open-source RISC-V development https://github.com/OpenXiangShan coming from the Chinese Academy of Science.
(Whether China will remain investing in RISC-V, given that the US government has started to pressure US RISC-V development to limit their involvement with China is another question.)
Interestingly, they've got binary translators for both ARM and x86 which were demoed in Geekerwan's video. Apparently, they're apparently donating a compilation farm for a Debian repository. Definetly interesting to see how Lithography will develop independently in China, and if it will ever pass the EUV chokepoint.
A decade or two isn't a long time, politically.
used to be MIPS based, is it still the case, could not find the info at that page.
The chip uses LoongSon's own "LoongArch" ISA. They moved away from MIPS supposedly because of fears of licensing issues.
It is very much like the last revision of MIPS but modernised a bit: No more branch delay slot. No HI/LO registers. The arithmetic ops that trapped on overflow are also gone.
There are published docs only for the scalar ISA subset: https://loongson.github.io/LoongArch-Documentation/README-EN...
Chips and Cheese has an analysis of the "LSX" and "LASX" SIMD sets. LSX is very much like MIPS's MSA. LASX is extended to 256 bits: https://chipsandcheese.com/2023/02/26/loongsons-lsx-and-lasx...
The ISA manual is here: https://loongson.github.io/LoongArch-Documentation/LoongArch...
MIPS ISA? Is this a problem in practice in 2024?
Not really. As long as there are compilers.
They just use GCC and LLVM.
Are there any comparable CPUs made in Europe?
x86_64? Seriously?
I would understand if that would be for legacy support, like x86_64 decoded to risc-v.
The article missed out on context. The chip runs LoongSon's own "LoongArch" ISA which they had developed out of MIPS'. Chips and Cheeses test programs are written in that ISA. There are other articles on the site that tell more about it.
LoongSon does however supposedly have a instruction set extension for running binary-translated x86 code faster (and ARM and RISC-V), but they have not published much about it.
Since RISC-V has also its roots in mips, maybe loongson should move to RISC-V for good with a x86/arm hardware translater.
All that to run binary distributed x86 apps...
13 replies →