> The development of the first Loongson chip was started in 2001. The aim of the Godson project was to develop "high performance general-purpose microprocessors in China", and to become technologically self-sufficient as part of the Made in China 2025 plan.<p>-Wikipedia<p>Right on time
<i>Loongson</i> (the company) have been around since 2001.<p><i>Loongarch</i> (the ISA that debian is now supporting) has only been around since 2021. Previously Loongson used MIPS and another ISA known as LoongISA.
I thought they pivoted to RISC-V<p><a href="https://en.wikipedia.org/wiki/RISC-V#Strategic_background" rel="nofollow">https://en.wikipedia.org/wiki/RISC-V#Strategic_background</a>
Well, that's what you get when you can do nation-state planning with an event horizon measured in decades, practically infinite cash and with the ruthlessness to engage in corporate espionage plus the total elimination of political campaigning (because you're the CCP, the only gang in town).<p>In contrast, Western corporate execs have an event horizon between "next quarterly reports" and "vesting/bonus period", additionally in any sufficiently large organizations you will have fiefdom fights that are counterproductive to the company at large, and Western politicians can only think to "next important election", so basically a few months as well because there is <i>always</i> an election going on in some state or local division.<p>We have gotten <i>societally incompatible</i> with not just coming up with new ideas, but with maintaining what we already have. It's like in 1984 - we can't even think of such timeframes any more.<p>And no, I'm not calling for a dictatorship. I'm rather calling for dissolution of the stock market for speculation, and for consolidating elections to once every four years so that politicians have time to cool down from the constant campaign hype pressure.
> for consolidating elections to <i>once every four years</i> so that politicians have time to cool down from the constant campaign hype pressure.<p>You talk about "western" politicians, but this point feels like it only really applies to the US with its ridiculous two-year cycle.
A 4-years cycle is also nowhere near enough for any long term project. This is the cost of democracy where you have to give people short term wins in order to win an election.<p>Infrastructure projects can take decades. The country’s budget is a 0 sum game. Developing something expensive will lead to cuts or shortages somewhere else and those people suffering will make you pay in the next election because they won’t care about the greater good from 20 years from now.<p>Your opponents will exploit any perceived failure and use it against you. And they will equally take your victories and sell them as their own if your long term project happens to bear fruit on their term. So everyone focuses on what looks good inside a term of a few years to sell in the next election.
> You talk about "western" politicians, but this point feels like it only really applies to the US with its ridiculous two-year cycle.<p>I'm German, it's even worse here because we have 16 states plus the EU elections and that means you have about three to five distinct elections in any given year. And with the exception of the small states no one gives a fuck about (sorry Saarland), each and every single of the state elections is seen as a sign for federal politics.
And a general election process that lasts from January to November of every fourth year, and an executive hobbled until the January after that because, goodness knows why, you get to remain in power for weeks after someone else has taken the mandate to govern.
> and an executive hobbled until the January after that because, goodness knows why, you get to remain in power for weeks after someone else has taken the mandate to govern.<p>Heh in Germany we have taken our sweet time as well. 2017 for example? That stretched into 2018, half a year [1].<p>The key difference is, while the outgoing executive holds their position until the new government coalition is formed, there is a strong tradition in a) the outgoing government to not abuse their positions to force their successors' hands, b) for everyone (including the far-left, excluding the far-right) to cooperate with each other to keep the country running and c) for cooperation across the aisles in general, even when there is a stable government. And there is no problem when the government/parliament cannot pass a budget either. Governmental authorities and public services keep on running assuming the last year's budget.<p>[1] <a href="https://www.mehr-demokratie.de/nachrichten/einzelansicht/die-laengste-regierungsbildung-in-der-deutschen-geschichte-haben-wir-fuer-sowas-ueberhaupt-zeit" rel="nofollow">https://www.mehr-demokratie.de/nachrichten/einzelansicht/die...</a>
> goodness knows why,<p>It takes time for the horses to bring the electors to Washington.
> In contrast, Western corporate execs have an event horizon between "next quarterly reports" and "vesting/bonus period",<p>The big tech companies ran without profit for a long time. People worried a lot about that but they did it anyway.
High interest rates in general are a strong deterrent from investing with that kind of time horizon. The risk free rate of return has skyrocketed, and investors no longer have to turn to risky gambles to make a return. The same thing happened in Canada when housing was propped up by the government in ‘08, leading to real estate sucking all of the money out of the economy. The US is in this position now because of how allergic all politicians have become to using taxes for the purpose of cooling down the economy. The only feasible campaigns are full inflationary spending, all the time, and the only difference is who the money goes to. The stupid electorate who vote for candy for dinner instead of eating vegetables are equally to blame.
> I'm rather calling for dissolution of the stock market for speculation<p>this would do so much untold good, i hope i live to see it
Didn't Richard Stallman use a Lemote Loongson computer before?<p><a href="https://stallman.org/stallman-computing.html" rel="nofollow">https://stallman.org/stallman-computing.html</a>
Yes, and that machine was exceptionally good; by far the best laptop i ever owned: mate screen, good keyboard, small and lightweight for the time, and exceptionally open and well supported for Linux despite being a mips arch. I would gladly get myself another one if I could.
That had a MIPS CPU
comes a long way.. they have some pc models sell in china but i guess only IT dev people would give it a try. china is pushing the state-owned companies and civil servant to use linux (some linux distro build by china company and replace windows and all America product) and the china-build CPU, but LooongArch also seems is not the #1 choice. I hope they can chooose LoongArch and built some debian based OS to use. This would be a 100 millon user market..<p>Also seems russia is interested to do some stuff based on LoongArch<p>just found one on JD: 14inch, LoongArch 3A6000, 16G mem, 512G storage, 4G GPU storage, sold for 6499RMB around 920USD
For the curious:<p><a href="https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html" rel="nofollow">https://loongson.github.io/LoongArch-Documentation/LoongArch...</a><p><a href="https://docs.kernel.org/arch/loongarch/introduction.html" rel="nofollow">https://docs.kernel.org/arch/loongarch/introduction.html</a>
I briefly owned one of the first Loongson netbook laptops, the Lemote I believe it was nicknamed. I've always been very much into open hardware, so I picked one up as soon as I could.<p>And I believe Debian was the only distro I was able to run on it back then too. Also NetBSD.<p>I ended up giving it to someone at the local hacker space. It's fun to try but it's not a daily driver.<p>Didn't Stallman use one as daily driver for a while?
That's a long way and congrats to the Longson team!<p>My college friend participated in the Google Summer of Code 2009, migrating openSUSE to MIPS. The CPU they used was an earlier version of Longson (forgot which one).
You surely meant to write "a loong way", right?
I actually bootstrapped openSUSE for loongarch64 last year:<p><a href="https://hackweek.opensuse.org/projects/bootstrap-opensuse-on-loongarch" rel="nofollow">https://hackweek.opensuse.org/projects/bootstrap-opensuse-on...</a>
Are Loongson powered machines available in the west? Are they only available from Chinese manufacturers?
You can just buy them off AliExpress, I've got myself a 3A6000 board from there - it's an interesting machine, feels much like a (UEFI) PC and is sure faster than any readily available ARM SBC, about the level of Zen1 I'd say.
Huh. Is this arch supported by LLVM/Rust?<p>It was announced recently (<a href="https://lwn.net/Articles/1044496/" rel="nofollow">https://lwn.net/Articles/1044496/</a>) that Apt, the Debian package manager, would require Rust by May 2026.<p>It does look like LoongArch is a supported target of Rust at Tier 2: <a href="https://doc.rust-lang.org/beta/rustc/platform-support/loongarch-linux.html" rel="nofollow">https://doc.rust-lang.org/beta/rustc/platform-support/loonga...</a><p>In fact, Rust's targets page is much bigger than I remembered! Good work
> Huh. Is this arch supported by LLVM/Rust?<p>Yes.<p>> It was announced recently (<a href="https://lwn.net/Articles/1044496/" rel="nofollow">https://lwn.net/Articles/1044496/</a>) that Apt, the Debian package manager, would require Rust by May 2026.<p>No, that was just the wish expressed by one of the APT maintainers. No actual decision has been made yet.
Hasn't RISC-V kinda taken its place by now?
Are there RISC-V chips with comparable performance for desktops and laptops on the market?
I thought so, yes: <a href="https://www.theregister.com/2025/03/05/china_alibaba_risc_v_c930/" rel="nofollow">https://www.theregister.com/2025/03/05/china_alibaba_risc_v_...</a><p>Not sure how they compare exactly. And of course it never hurts to have two players in the game.<p>I just meant I've heard a lot more about stuff going on in the RISC-V field in China. A lot of it is focused on the embedded stuff yes but not only there.
Are C930 CPUs actually available?<p>For the Loongson 3A6000, there is an independent performance review: <a href="https://chipsandcheese.com/p/loongson-3a6000-a-star-among-chinese-cpus" rel="nofollow">https://chipsandcheese.com/p/loongson-3a6000-a-star-among-ch...</a><p>I couldn't find something comparable for C930 CPUs.
Debian's RISC-V machines are rather slow.<p>It takes several days to build the gcc-15 package in riscv64 but just a few hours on loong64.
Yup. Their best builders run JH7110 (based on the in-order SiFive U74 cpu, RVA20 + partial B) and these are slow, somewhere between rpi3 and rpi4.<p>This will improve in 2026, with the first chips integrating RVA23 microarchitectures, such as the Tenstorrent Atlantis SoC and development board, with Ascalon, announced for 2026Q2.
Wow man. Back in the day the Godson processors were supposed to be these MIPS chips from China that ran Linux. I wanted one just for the sheer curiosity of it all but couldn’t get one here in the US.<p>I wonder if there is a way to get them from Taiwan / Korea. I can’t go to mainland China.
There seems to be a community-run site about this:<p><a href="https://loongfans.cn/en/pages/intro.html#i-m-sold-where-can-i-buy-one" rel="nofollow">https://loongfans.cn/en/pages/intro.html#i-m-sold-where-can-...</a>
This site links to <a href="https://areweloongyet.com/en/" rel="nofollow">https://areweloongyet.com/en/</a><p>> Even though the Loongson Corporation has yet to release the remaining volumes of the LoongArch manual, thanks to a wealth of public information such as publicly available QEMU and Linux changes, "undisclosed" information such as instruction encoding and behaviors is, in effect, already public. The absence of manuals can no longer hinder optimization efforts by The People.<p>If the architecture is still undocumented, I would not consider Debian’s move to be wise.
You can buy LoongArch hardware on AliExpress, for example.
Always thought they’d ship me garbage. Not sure why. I’ve bought lots of stuff there. But I don’t know how to tell it’s not some rebadged Via chip before it arrives. Just the game, I suppose. To Aliexpress!
Buy from listings with many sales and good reviews.<p>Look for reviews with real images and real seeming phrases in many languages, not 10 accounts all posting the same phrase with no pictures.<p>Buy from stores with a name, preferably who have established a "brand" for themselves across many products. UGreen are a great example of this for USB gadgets.<p>Don't buy from stores named Shop195772040, these will take your money and disappear or ship fakes. Don't buy suspiciously cheap items with no sales, these will do the same.
Experiences can vary from seller to seller, but I've owned a few oddball motherboards from china that shouldn't exist. (old server cpu sockets shoved onto a Micro ATX Board) Any time I've had issues getting a refund took maybe a week or two at most. Although the last issue I had was a few VRM components exploding and throwing shrapnel all over inside the case. So buy at your own risk on some things.
I don't think LoongArch is the same thing with MIPS according to Wikipedia: <a href="https://en.wikipedia.org/wiki/Loongson" rel="nofollow">https://en.wikipedia.org/wiki/Loongson</a>
MIPS ist dead, so not much gain in keeping compatibility with it anymore.
> I don't think LoongArch is the same thing with MIPS according to Wikipedia: <a href="https://en.wikipedia.org/wiki/Loongson" rel="nofollow">https://en.wikipedia.org/wiki/Loongson</a><p>Where did I make that claim? I never said that.
Once I used an old MIPS debian in qemu to test big endian. I thought since LoongArch has its root in MIPS it may be big endian. It turns out "LoongArch bit designations are always little-endian".
I want one of these chips