Since US tech embargo on China is still in effect and US plans to squeeze Chinese tech companies by restricting their access to computer chips. China's reaction to this was to make one of their goals to have a home-grown alternative for Intel, AMD, TSMC and the like to decrease dependence on the US. So I was wondering what kind of progress have they been making lately and what are their prospects of ever having a viable alternative to Intel and AMD?
There is all this talk about SMIC, Loongson and Zhaoxin and that in early 2020 Zhaoxin allegedly published a chip comparable to 2017 level intel and AMD tech and they are planning to have parity with Intel in a few years. Then there are Some sources are saying that China is failing horribly at acquiring manufacturing tech for more advanced processors. And I really can't make any real sense out of it.
So when will I be able to buy a computer that uses a hardware that dunks on burger made tech and sends my personal data to Beijing instead of Washington, if ever?
>>10249>what about OpenPOWERanybody know anything more about this ?
>>10254>How hard will it be for China to acquire the expertise, tech and build a manufacturing base for more advanced chips to catch up and keep up with the rest of the world?The important bottleneck in chip making is the lithography machines. The tech leader is ASML, that's a company in the Netherlands that is blocked by the US from selling to China.
(For the next paragraph smaller number = better)
China has made it's own lithography machines company and that one is called SMEE, they have machines in place for 28nm mass production at the moment, and they are working on getting mass production roll out of 14nm next year. The chip maker in Taiwan TSMC who is allowed to have ASML lithography gear is currently doing 7nm mass production and they are working on getting 3nm in to production at some time in the future.
China is behind 3 generations for the high end chips, but over100nm chips are still used in many applications. For chips that don't need that much performance they are already set. The small numbers in nm are most critical for portable devices because it means less battery drain and heat.
From the synthetic spec numbers it seems that China is catching up, how the realworld processing performance is stacking up, i don't know.
The US primary strategic goal of chip asfixiation has failed, china can self supply their strategic needs for chips. The secondary objective is commercial, they are trying to make Chinese products less competitive. That has had some effect but China is still able to source high end chips. And the US can't push too hard because China could cut the US off from TSMC. and the US is only now beginning to build out their own chip self sufficiency with new chip factories.
I guess that we should root for every side that is building out chip production because that means the chip supply gets better. This post was typed on a computer that's using 32nm chips (if i'm not mistaken)
>>10249>what about OpenPOWERYou can buy these already, it's a evolution of the type of processors that used to be in Macintosh computers.
For example a Power9 based blackbird desktop computer from raptor is the closest thing to a fully open hardware that is actually usable as a powerful PC. You have to go through a lot of hoops to set it up correctly but you also do get a lot of security benefits. Compared to RISC V this is a lot more mature and has more processing horsepower.
If you have the money and necessary technical inclination for the set up process, it comes highly recommended.
>>10337> Is there any way we can verify that the manufacturer actually made the chips according to the published plans?No.
> Could they still insert backdoors that we can't find?Yes.
>>10365That sounds interesting, do you have more resources to learn about it?
>>10337What about a microscope?
>>10908China can veto the acquisition, and even if it goes through, American export regulations only applies to products that have been developed in the USA. Unless Nvidia plans to move the full company to the USA, it is unlikely to change the current situation.
Arm continued to work with Huawei despite Trump fucking with them, so I don't think your concerns are well founded.
>>10973Arm more processing / power
X86 more processing / money
Although that distinction is slowly shrinking
What is putting Arm on the future map is that RiscV is very modular, processors can be tailored for their intended workload very easily and that will come with massive speed increases.
X86 isn't going to die anytime soon it's probably has got at least 20 years left.
>>11008In my opinion people are not looking at processor backdoors rationally, instead of trying to get a "clean untainted chip" people should be trying to repurpose the backdoor functionality. It's all just silicon transistors, why not try to make use of it? It probably got some unique features that are very useful for a number of applications.
>>11035It swings both ways, x86 has legacy baggage, but it also has a lot of software that is optimized for it.
Maybe going with powerpc would have been better, i don't know.
Of course if RiskV sort of develops into a gnu linux equivalent for hardware, it would take out all the hurdles for optimizing software and hardware in tandem, and we'd get amazingly efficient, fast and stable computing with a fraction of present day effort. But it will be very difficult and time consuming to get the ball rolling on this.
>>11627xiaomi phone
many laptops are chinese. i have a stinkpad with linooks, it just werqs dawg
>>11632you can use VPN on an android
>vidyai wouldn't do that on a laptop tbh
i havethe cheapest thinkpad and it's fine for everything besides vidya
How China Made An Exascale Supercomputer Out Of Old 14 Nanometer Tech
>If you need any proof that it doesn’t take the most advanced chip manufacturing processes to create an exascale-class supercomputer, you need look no further than the Sunway “OceanLight” system housed at the National Supercomputing Center in Wuxi, China.
>Before this slew of papers were announced with details on the new Sunway many-core processor, we did take a stab at figuring out how the National Research Center of Parallel Computer Engineering and Technology (known as NRCPC) might build an exascale system, scaling up from the SW26010 processor used in the Sunway “TaihuLight” machine that took the world by storm back in June 2016.
>The 260-core SW26010 processor was etched by Chinese foundry Semiconductor Manufacturing International Corporation using 28 nanometer processes – not exactly cutting edge. And the SW26010-Pro processor, etched using 14 nanometer processes, is not on an advanced node, but China is perfectly happy to burn a lot of coal to power and cool the OceanLight kicker system based on it. (Also known as the Sunway exascale system or the New Generation Sunway supercomputer.)
>If the 160 cabinet scale is the maximum for OceanLight, then China could best the performance of the 1.5 exaflops “Frontier” supercomputer being tuned up at Oak Ridge National Laboratories today and also extend beyond the peak theoretical performance of the 2 exaflops “Aurora” supercomputer coming to Argonne National Laboratory later this year
>The bottom line is that NRCPC, working with SMIC, has had an exascale machine in the field for a year already. (There are two, in fact.) Can the United States say that right now? No it can’t. The United States is counting on its exascale machines to be more energy efficient – Frontier and El Capitan for sure, we shall see with Aurora – but we have no idea how computationally efficient any of these future machines really are.
So all know, the current 28nm Chinese risc processor using Sunway TaihuLight relased in 2016 is 4th on the TOP500 supercomputer list.
Ptychographic X-ray laminography: No trade secret or hardware trojan can hide
https://spectrum.ieee.org/chip-x-rayEvery cell transmitter, phone and computer must be inspected. Silicon Trojans must perish.
>>15971no more nsa agents watching my futa hentai through the backdoors
westerners in shambles
>>18528This is a garbage article that preys on the ignorance of people who know very little about the subject. The whole article can be basically summarized as "China will build their own machines, with blackjack and hookers", well sure, nobody was stopping them from doing so. Saying that China will build its own tools because the government will foot the bill is hardly a new idea, countries have tried and failed before.
Much of China's semiconductor manufacturing advancements have come from technological transfers and plain old stealing, including the 7nm design, but copying is hardly a good way to stay in the semiconductor race, especially once you encounter large technological barriers like in this case for example the lack of EUV machines. Nobody has claimed 7nm (N7) couldn't be manufactured with a DUV machine, but it's hardly economically viable, achieving it means little if you can't manufacture it at scale to be cost-competitive.
And for all the copium I've yet to see China actually invest in their own indigenous manufacturing advancements to compete at the leading edge, it's either stunts to grab headlines but in practice it goes nowhere or focus on much less sophisticated manufacturing which they can actually do at scale, which is also still needed and useful, but hardly anything groundbreaking like their fans constantly claim.
>>18601>people who know very little about the subject.like yourself? they have been shipping 7nm chips at least since 2021 but didn't publicize it so the west only found out in 2022
the bloomberg article here
>>15971 quotes a techinsights report from 2022 where they basically say that they found 7nm chips in a chinese bitcoin miner they were analyzing. minerva (the company that designs and sells that SoC) had been selling this particular miner since mid 2021. smic process is scalable and viable
>build its own tools because the government will foot the bill is hardly a new idea, countries have triedand succeeded, take the united states for example: the department of energy funded the EUV LLC, a consortium of three american companies that developed all the important parts for EUV lithography. the most important company of those three is cymer, a firm from texas that manufactures the EUV light sources necessary for the EUV process and that ASML was basically forced to buy
so although ASML is dutch in name, all the important R&D comes from the US, in part funded by venture capital, in part funded by the american state. and I'm sure that you would find even more state money if you were to dig into the "venture capital" part. this might surprise you, but is a common practice for the government to finance research and then give away the results to the private sector (but not the broad public) basically for free
tl;dr
>subsidies for me, not for thee >>15971this is both an old article (2022) and a nothingburger.
this is the limit of the DUV technology, the fact that china managed to do this is impressive but without newer machines they cant get beyond 7nmn. Meanwhile TSMC, Intel, etc. will be getting into the angstrom era.
https://www.pcgamer.com/intel-puts-tsmc-on-notice-with-step-towards-angstrom-era-chips/see this article from 2 weeks ago.
>Intel says it has completed development of its upcoming 20A and 18A chip production processes. The first chips built on the first of these new nodes—the 20A node—will be made in the first half of next year.20 angstrom is 2nm. I.E. non chinese companies are already using EUV lithography to make 2 and 1.8 nm nodes.
>>18528>Previous reports have indicated that Loongson's 3A6000 processor will allegedly provide performance that is on par with AMD's Ryzen 5000 CPUs and Intel's 11th-Gen Core CPUs, which both debuted in 2020.yes and later this year intel will be making chips that are far faster and smaller than what they did before.
>>18901You realize they're not reliable, easy to make, affordable or usable in good yields?
Its nice that we have people learning the ways, but it won't replace how many decades from this we are now.
>>10378would have to be some kind of special X-Ray microscope or regular electron microscope, both of which are likely to damage whatever they scan, and then still it'd only be practical to check small parts at a time, unless some kind of algorithmic solution is used.
>>15893oop, guess I'm too late
Open Source tastes best when forced anyways
>>15893holy BASED
when will we have home-bootstrappable low tech CPUs to cascade in like 5 cycles of re-etching upwards into formally verified FOSS CPUs? 50 years?
>>21476This, coupled with what seems to be Huawei's superior android app compiler than can increase performance by more than 30%, Huawei is in a great position.
https://www.xda-developers.com/huawei-ark-compiler-android-app-performance/Not sure how common this Ark compiler is and it doesn't necessarily work with google play store downloaded apps, unfortunately. Other than that, I might get a huawei as my next phone. I got the cheapest compatible pixel phone and installed grapheneOS. It's not too bad but I do wonder how effective it is.
RISC laptop* incoming for $300. The end of history is over anons.
https://www.cnx-software.com/2024/04/30/muse-book-laptop-spacemit-k1-octa-core-risc-v-ai-processor-16gb-ram/*with decent specs, screen and connectivity
>>25345They buried the lede on this device. It has a RVA22 CPU with RVV (Vector) 1.0.
Finally, we're seeing devices with the important batch of extensions ratified December 2021.
https://news.ycombinator.com/item?id=40314731 >>26015>Or even better the "Fourth Industrial Revolution" is actually the second "Starwars", and coming "Third AI Winter".Think now this mostly can't be the case. If cost of automating services like tellers or cashiers, and logistics industries (of course logistics came to mind even before) drops to the point where these things could be reasonably automated it would have a significant impact. The odd thing to me is that these could largely already be automated by simple web-apps or button panels and scanners, or in the case of logistics by train yards. Is natural language processing, or sophisticated image recognition really necessary for such automation - or more compellingly, would they be sufficient for such a realization? Capital simply systematically undervalues labor and so such an automation has only been partial; and it may be that reducing the cost of implementing these services would be such that it could cause radical shifts in industry. In some industries, like logistics, this is nearly a guarantee
maybe starting 2027 it looks like.
Is StarFive (
https://www.starfivetech.com/en/site/company) really Chinese or no? If they count then might they already have a cool Chinese RISC-V:
https://milkv.io/mars#buy (Shenzhen MilkV Technology Co., Ltd (Milk-V)) can take a SATA SSD with an adapter which was always an issue with older SBC.
Unique IPs: 49