China may prove Arm wrong about RISC-V's role in the datacenter

Arm might not think RISC-V is a threat to its newfound foothold in the datacenter, but growing pressure on Chinese chipmaking could ultimately change that, Forrester Research analyst Glenn O'Donnell tells The Register. Over the past few years the US has piled on export bans and trade restrictions on Chinese chipmakers in an …

[Jason Bloomberg]

A shot in the head is worth two in the feet

I have always said that America's attempts to isolate China were misguided.

It is always better to keep essential supplies going and threaten to reduce supply to steer a recipient in a desired direction. It's easier, cheaper, and less disruptive for them to comply than resist. You can keep them on that leash.

As soon as it's "none for you" they have no option but to find an alternative, and that will usually have them looking for a sovereign home-grown solution so they are never placed in that position again.

At that point you've lost any leverage you had over them, and may have even gifted them the upper-hand.

[martinusher]

Re: A shot in the head is worth two in the feet

This has been pointed out by some quite eminent economists. Political interference in the supply chain means that the build/buy balance gets messed up so what once didn't make sense from a business perspective now becomes an imperative. The result is a double hit to the US semiconductor industry -- not only do we lose a ton of business in the short term but we have encouraged the creation of formidable competitors which threatens our markets in the longer term.

Its not just semiconductors. Boeing, our de-facto civil aircraft monopoly, has done quite well using political pressure to hold back potential competitors in the small aircraft segment. They'll continue to do this right up to the point where COMAC eats their global lunch. (We'll obviously do what we can to slow the Chinese down using sanctions and export restrictions, they'll respond, we'll drag our feet with certifications and stuff, but eventually the dam will break.)

(....and don't even mention Huawei.....their 'crime' being to nab the lead in 5G technology and so threaten the monopoly that we had in 4G)

The only way to stay on top is to run faster.

[J.G.Harston]

Re: A shot in the head is worth two in the feet

Remember what happened when the US banned exports of oil to Japan in the 1930s.

[StrangerHereMyself]

Re: A shot in the head is worth two in the feet

As I stated earlier the U.S. will soon restrict the usage of RISC-V to hobble its progress, since its availability will allow China and Russia to keep up with Western CPU technologies.

And by restricting usage think a ban on using it in U.S. government hardware, including military hardware. That alone would cripple RISC-V.

[3arn0wl]

Re: A shot in the head is worth two in the feet

I still can't see this prediction happening, even if the US gain control of ARM (as seems likely), and continues its masochistic war of market ideology.

The outcome would be the polarization of ISAs, with the US (and a few of its allies) using ARM / X86 and the rest of the world going RISC-V (for digital sovereignty). That just diminishes US markets... Talk about cutting off your nose to spite your face.

And the idea that nowhere else has the knowledge / skills to produce the goods is simply wrong - RISC-V is node-agnostic : and this continual struggle to get things below 7nm is arbitrary.

[StrangerHereMyself]

Re: A shot in the head is worth two in the feet

The U.S. control of ARM comes through its markets. Commercial companies like ARM will have to abide by U.S. laws just to keep access to its markets.

[3arn0wl]

Re: A shot in the head is worth two in the feet

The tech trade dispute already weaponized the ARM ISA... and look how that's acting as a catalyst on RISC-V. If you're thinking that the US might take things further... that would only encourage China (and other countries) to adopt RISC-V as soon as it's feasible to do so. (ROMA laptops produced by the million rather than a few hundred.)

And having inadvertently hastened the development and uptake of RISC-V, are you really suggesting that the US government would then prevent their companies from exploiting it? That's really not likely, because it would mean a huge loss of market share to US tech.

[Anonymous Coward]

Re: A shot in the head is worth two in the feet

Never underestimate the sheer stupidity of some of the members of any government, not just the US. It is well within the realm of possibility for another loose cannon like Drumpt to win an election and go overboard with asinine political moves that are popular with ignorant people who know nothing about reality.

One thing you can count on: the US will bleed the RISC-V market for every dime they can suck out of it, the same as they do all others.

[Anonymous Coward]

Re: A shot in the head is worth two in the feet

"Commercial companies like ARM will have to abide by U.S. laws. just to keep access to its markets."

FTFY

[BOFH in Training]

Re: A shot in the head is worth two in the feet

RISC-V is opensource (at least the core parts). Regardless US bans it or not, it's already out there and available for pretty much every and any one who wants to use it. Or modify it.

Only control US has is in the advanced fabs. Does not matter if you have the best / most powerful designs if you can't get it fabbed.

But thats probably at most 10-20 years away from being irrelevant - I expect China to have it's own fabs at similar nodes by that time. They will either buy/steal/invent the tech as needed. After all Chinese government has allocated at least 100B bucks on that industry, not to mention the amounts being invested by the private sector there.

[lkcl]

Re: A shot in the head is worth two in the feet

> RISC-V is opensource

no - it isn't: it's Trademarked. there exist open source *implementations* of that Trademark: none of those implementations grant You (the Trademark Licensee) the right to use the word "RISC-V" in connection with that commercial product.

> Or modify it.

as long as you remove the word "RISC-V" if you do not seek a Trademark License in your commercial product (or do not care about Trademark Law) ... yes.

> I expect China to have it's own fabs at similar nodes by that time.

they'll have low-cost tablets and smartphones and IPTVs as soon as the software (android) is ready, because 28nm is the "low-cost sweet-spot" and you don't need more than 1.5 ghz for a *low-power* low-cost (battery-operated, mobile) processor.

but there is something really important here that everyone is missing, best explained here:

https://news.ycombinator.com/item?id=24459314

there's a glacially-slow trainwreck coming, unfortunately, as it takes time for companies to appreciate that RISC-V - the "simple" RISC ISA - has been drastically *over*-simplified. it will be perfect for embedded uses (eating ARM's Embedded market), but for high-performance - particularly where software interoperability begins to matter - it's going to be a cluster****.

[hoola]

Re: A shot in the head is worth two in the feet

The Chinese think long-term and always have done. That is why they are so successful and are such a huge manufacturing base.

They develop, try and repeat until they get it right. The West has helped hugely by pushing so much manufacturing out to China, all in the quest for cheaper prices.

China will become independent of the West's perceived stranglehold on technology IP or stands and jus make their own.

The losers will be the West, again.

[StrangerHereMyself]

Re: A shot in the head is worth two in the feet

Uhmmm, no. If China loses access to ASML's EUV and DUV machines they'll quickly fall behind, perennially stuck in the early 1990's.

[BOFH in Training]

Re: A shot in the head is worth two in the feet

What makes you think China is not working on creating it's own version of ASML?

Buy/Steal/Invent is China's way of doing things, in that I order, I suspect.

They are throwing alot of money on doing this anyway. Especially since China government tends to have very long term views on things, instead of the wall street current quarter outlook or next election outlook of the politicians.

Am not a fan of the Chinese government, but you do have to acknowledge that they have a better way of planning for the future they want, compared to most of the world.

[StrangerHereMyself]

Re: A shot in the head is worth two in the feet

I know they tried. But failed.

[orphic]

Re: A shot in the head is worth two in the feet

If China loses access to DUV and EUV remains inaccessible? Firstly, China already has an installed base of DUV machines. This is like a household losing access to buying cars when it already owns two. That scenario won't affect such a household until their current cars are obsolescent. Like Cuba with its 1950s cars or Iran with its US-purchased F14s, it would be forced to maintain those machines beyond obsolescence if necessary. So the idea of a decision knocking China back to the 1990s is flawed.

The second idea implicit in your argument is that China cannot develop DUV and EUV. This is misleading as China has a competitive ArFi DUV machine close to mass production. It could easily replace foreign DUV machines within a 5-year period. With regards to EUV, the idea of its machine complexity is both overstated and misleading. ASML's EUV machine tin-droplet LPP light source represents a triumph for non-radioactive EUV light source technology. It does not as popular media like to portray it, represent the only method of generating light source EUV.

Others have generated EUV through a synchrotron approach, which has the added advantage of not being affected by US-controlled patent hurdles. Russian, Chinese, and Japanese researchers have invested in pursuing this approach which not only provides a purer stronger light source but can micro-fabricate at lower nodes than ASML's product without messy tin splattering.

[Ken G]

Re: A shot in the head is worth two in the feet

Yes. I was reading yesterday that China had licensed production of Russian Sukhoi jets, then decided to make their own instead. Russia cut off their supply of essential parts (mostly the engines) and China had to go elsewhere for those, namely Motor Sich in Zaporizhzhia, Ukraine.

[Tom 7]

Re: A shot in the head is worth two in the feet

The thing is I would bet China has one advantage over the west - it's not driven by the too-late market. If China decides it wants to make chips from sand upwards it can divert the resources to get that to happen. I would bet its got lawyers working on sidestepping IP problems as we speak.

[Proton_badger]

ARM

Didn't ARM only say that RISC-V was not a threat in the data center in the short-medium term? Even they acknowledge the long term threat, only question is whether it'll happen quicker than they think.

[StrangerHereMyself]

Re: ARM

ARM is merely proclaiming wishful thinking. RISC-V will quickly transition into the datacenter and desktops / laptops, if only in China and India.

[Ken G]

Re: ARM

Agreed, especially since not everyone needs "high" performance, just a good performance/cost ratio. If it runs at half the speed of Intel but costs a quarter as much and takes out some supply chain risk, why wouldn't a lot of governments especially go for it?

I can see a two-tier architecture where government cloud uses RISC-V and commercial ARM/Intel. We also need to see where the cheap mobile phone providers go.

[Charlie Clark]

Poor article

One of the most attractive things about RISC-V over Softbank-owned Arm is the relatively low cost of building chips based on the tech, especially for highly commoditized use cases like embedded processors

The type of architecture has very little to do with the cost of actually making the chips. Maybe he means designing them? But RISC chips (in the form of MIPS, et al.) have been in embedded controllers for decades. The licence fees for ARMs own microcontrollers mean that they're still competitive in this area, though every wants to move up the value chain because these chips are all, er, as cheap as chips.

The article then goes on to say that both CPUs are possible but there's currently no indication that there's anything remotely as powerful as ARM or x86 out there or on the horizon.

Powerful RISC-V designs are possible but China is still a long way from having the skills to design them. But, perhaps more importantly, it is even further away from being able to build the machines that can make them. However, I also agree that the sanctions applied to China will probably be counter productive over time. There we implemented largely on the basis of a whim to try and please crowds. But, until America can have Christmas without Chinese made tat, they're never really going to make an impact.®

[P. Lee]

Re: Poor article

It depends what "competitive" means.

Maybe they don't need an epyc server. Maybe they just need a good-enough netbook or phone cpu. While it might be nice to have an iphone 14, the major productivity gains from having a mobile phone are not dependent on Apple's latest CPUs. If you want to have a computer so that your business staff can do their jobs, a desktop i3 may well be plenty to get everything done - a gen12 i7 is not normally needed. Economic progress rarely needs top end tech.

I'm sure India hasn't missed what the West has done to China and Russia. I'm sure they'd also be up for some tech independence.

[Charlie Clark]

Re: Poor article

The only difference between modern phone CPUs and server ones is scale. RISC-V is a long way from being ready for either. Network or disk controllers are completely different beasts.

[StrangerHereMyself]

Re: Poor article

There are billions of low-powered embedded CPU's being used every year. Most of these are ARM based but in the near future I see these all transitioning to RISC-V.

Surely that will hit ARM amidships if not sink it outright?

[Charlie Clark]

Re: Poor article

Such transitions are ongoing in any industry. As a result ARM produces more sophisticated microcontrollers that are still low power and these will continue to find new markets. More importantly, seeing as ARM doesn't make the chips itself, it also doesn't have to worry about inventory if a particular design becomes unpopular.

Manufacturers on the other hand have to weigh up the licence fees against the costs of switching architectures. They don't switch often because licence fees make up a very small part of a very small part.

[3arn0wl]

It's more about control

Well... the licence fees (and royalties) ain't nothing, but that just gets passed on to the consumer. The bigger issue is control. In 2016 Nvidia announced that they had designed some microcontrollers with RISC-V because they could do things with it that they weren't allowed to do with ARM.

And rumors are flying around at the moment about notable phone OEMs looking at using RISC-V for co-processors too - presumably because they've seen that it can do some things better.

[StrangerHereMyself]

Re: Poor article

Eventually those high-end microcontrollers will be implemented with RISC-V and ARM will slide into oblivion.

There's no technical reason why high-performance MCU's can't be done with RISC-V. It's mostly a matter of manufacturing node size.

[Bruce Hoult]

Re: Poor article

>The article then goes on to say that both CPUs are possible but there's currently no indication that there's anything remotely as powerful as ARM or x86 out there or on the horizon.

That's just not correct.

It is well known in the industry that RISC-V startup Rivos, for example, has been founded by some of the same people who founded PA Semi, which as bought by Apple in 2008, and created Apple's A-series chips and the M1/M2. A number of other engineers who worked on the M1 have also gone to Rivos. Apple is suing them for taking "too much" information with them.

Interestingly, Apple is not asking for an injunction to stop Rivos from working, or even fixed damages, but for a ROYALTY on sales. That means Apple expects them to be successful. They have the people, they have the funding.

The way these things go with Rivos being founded in May 2021, an M1-class (maybe better) RISC-V will emerge from them probably around 2024-2025.

There are others too, but that's the slam-dunk example.

[Tom 7]

Re: Poor article

"China is still a long way from having the skills to design them". Is it? How do you know this? How much would it take to employ the right people with those skills? The only way the west can stop people going there to work is to become a state like them.

[3arn0wl]

Is there any more news on Esperanto's ET-SoC-1 AI inferencing chip? Last I heard was that Samsung were testing it, and were quite impressed with it.

[Anonymous Coward]

"If China got really serious about this, we could have a truly viable third architecture to contend with."

It reads like two arguments are conflated which are not necessarily so closely related.

For RISC-V, I would not look at China alone, as there are significant examples from universities (reported by El Reg partners)

https://www.nextplatform.com/2022/06/09/strong-showing-for-first-experimental-risc-v-supercomputer/

to start-ups like Ventana Microsystems, to MIPS https://www.hpcwire.com/off-the-wire/mips-pivots-to-risc-v-with-performance-and-scalability/

that indicate that high performance RISC-V is very possible.

Not to mention Intel; they look like they can't wait to put a spoke with a RISC-V die on it in their ARM-based competitors wheels.

When it comes to China, I would take things like YiTian as more serious indications that Chinese designers are indeed smart enough to create high-performance chips. They might not have EUV, but they sure have the brains; some recent numbers seem to support this assumption:

https://www.servethehome.com/arm-based-alibaba-cloud-t-head-yitian-710-crushes-specrate2017_int_base/

[boatsman]

uhm... relativity !!

I read your quoted article a bit..

https://www.servethehome.com/arm-based-alibaba-cloud-t-head-yitian-710-crushes-specrate2017_int_base/

"...... an ASUS AMD EPYC 7773X (“Milan-X”) CPU with 64 cores has published results of 440 or 6.875/ core, but with half as many cores and older generation DDR4 (albeit with a larger L3 cache.) ....."

301 vs 440 *but* half the cores / half the energy or worse.

actually 150.5 vs 440

chip architecture is a piece of paper, that might be briliant.

actually building a chip that implements the architecture, is a veeeery difficult job, and china / india / etc

do not have the brains walking around to do that

*with*

the necessary ecosystem / tools AND experience.

there is no 5 day course for "experience"

There is enough smart people in these countries, without a doubt. but experience / tools / etc is not for sale.

And last but not least: ASML does not sell the Ultra violet machinery to China. fullstop.

[Anonymous Coward]

Re: uhm... relativity !!

Your points are surely valid, ecosystem, ASML and all that.

As far as experience goes, however, I tend to be less radical.

Especially considering that since decades a significant number of talented Chinese individuals study and do research abroad before setting up R&D labs in China and/or have been trained in China by companies who wanted to take advantage of the low wages.

This could be a fun pub bet: meet in 5 years and see who owns a pint or two to whom :)

[dshan]

Another Processor Transition

10-15 years from now Apple will be undergoing yet another processor transition — this time from ARM to RISC-V. No way they'll keep paying ARM licence fees once they are confident they can get the performance they need from a free processor architecture.

[CheesyTheClown]

Nope... not a risk... we just don't use them... ARM that is

We're currently planning 5 exa-scale computers in the next 5 years. We had been working with ARM as we're developing our own CPUs. Then after extensive research and money spent and of course the Trump embargos on China, we (a European collaboration) moved on to RISC-V. Trump and Biden taught us that no one is immune to being crippled by friends... and unlike China, we just don't have the resources to bounce back.

So, we've more or less dumped ARM beyond some POC related things. So we'll build systems with a few dozen or maybe even thousand CPU cores, but then we'll scrap them and replace them with a RISC-V ISA.

Here's the real reason why ARM offers nothing in the datacenter over RISC-V.

There are so many varieties of ARM that compilers just aren't optimized for them. It's not like Intel or AMD where compilers are written to optimize for each generation of chip. For example, one generation of compiler can optimize code for CPU cores organized in a ring bus, the next generation will be organized in a mesh. But with ARM, you never know what you're going to get and the compilers don't optimize for a generation or even a given core. As a result, ARM processors are kinda useless when you need performance. Consider that code optimized for a NVidia processor probably isn't optimized for Qualcomm and certainly code written for Apple M2 isn't optimized at all for NVidia. You end up using compilers which target the minimum common denominator which means no support for many instructions or architectures.

RISC-V will be worse. As most ARM processors actually are mostly designed by one company. On the other hand, RISC-V processors will be designed by pretty much anyone. They'll have nothing in common and their instruction sets are highly customizable. As a result, we will very likely become dependent on JIT compilers which on Exascale machines are actually optimal as they can optimize code for the platform as they run based on their current states.

Already, I can't wait to get my Alibaba RISC-V laptop to start optimizing code.

[Anonymous Coward]

I think Arm's CEO is blowing smoke about RISC-V not being a "challenger." Instructions sets don't hold processors back too much; it is a matter of someone putting in the time and energy to develop better and more efficient cores. I foresee a massive drop in Arm's and Intel's and AMD's revenues for processors in 5-10 years.

[Lordrobot]

RISC-V... it's no match for out lead sleds

"We really don't see [Toyota] RISC-V as a significant competitor to [GM] us in the [Auto Industry] datacenter space right now, or in the near future," said Chris Bergey, SVP and GM of Arm's Infrastructure business, characterizing the rival ISA as being best suited to niche applications. "Nobody is ever going to drive a Toyota more than 30 miles from home." [GM Board Members enjoy a hardy smug laugh 1971]