Hacker News new | past | comments | ask | show | jobs | submit login
A history of ARM, part 2: Everything starts to come together (arstechnica.com)
104 points by dinosor 6 days ago | hide | past | favorite | 34 comments

The "the emerging standard" picture resonates with me. It seems like the other manufacturers were fine making their chips with their own weird architectures, but ARM did everything to promote itself as a standard, having which comes with a lot of supply chain advantages to its customers.

Fun fact buried in the article, Acorn renamed itself to Element14 and are today of the biggest Raspberry Pi distributors (edit: incorrect, see below):

> Acorn sold its stock and used the money to refinance and restructure the company. In 1999, the company renamed itself “Element 14” and changed its focus to developing telecommunications products.


So they're still selling ARMs!

Different company, the ex-Acorn one was bought by Broadcom in 2000 and disappeared, Farnell renamed itself Farnell Element14 circa 2010 (although it looks like they're just back to being Farnell again now).

Yes, the Element14 that is selling the Raspberry Pi is a subsidiary of Premier Farnell, which is now owned by Avnet.

Farnell is the giant electronics behemoth that sells just about every electronic component on the planet. Farnell will also sell you just about every ARM based CPU that is freely available on the market, whether you want 1, or 1 million of them.

Ah, thanks for clarifying! I wasn't expecting such a name to have been used by two different companies.

It's a reference to silicon, the element with atomic number 14. Makes sense that it would be used by more than one company.

So sad to see ARM being run into the ground by greedy capitalists. The thing that did it was the stock flotation in the late '90's, the root of all evil.

After that it was merely a matter of time before greedy shareholders wanted the company sold to the highest bidder.

Luckily we have RISC-V to fall back upon.

Fascinating story. I do find it odd that the main designer is referred to throughout as Sophie Wilson although when designing the original ARM chip she/he was actually Roger Wilson :-/

Why? Deadnaming trans people is considered rude and invalidating. Additionally, I believe many (most?) trans people would consider themselves as always having been the same gender, even before they came out and started presenting as such.

Ok, thanks for the clarification.

The ending is kinda sad. The curtain closes, and RISC-V has replaced it.

But it is for the best.

I find this obsession with RISC-V on HN fascinating. It’s not even close to challenging ARM or x86 outside of very low power/embedded applications. That might change at some poi t but I think it’s very far from certain.

RISC-V has a very strong value proposition that will carry it forward beyond what ARM or x86 can offer.

The only place you can buy ARM IP from is ARM but you can buy RISC-V IP from anyone. That means it's one company versus many, potentially a whole industry. If you buy ARM processor IP you are locked in to buying processor IP from one company. With RISC-V you can change vendors and have several compete for your business at any point.

ARM is already showing signs of being shut out of the future of the market. For instance Si-Five's X280 processor has found a rapidly expanding niche as a "companion processor" to custom AI hardware, something that isn't practical with ARM's narrow offerings or licensing model. See https://youtu.be/RrVRMFjYti0?t=1732

I think it’s the appeal of FOSS regardless of uptake or viability. RISC-V is definitely taking off in a lot of areas but the religiousness around it seems to be the idea of “FOSS alternative is inherently better and any shortcomings are either irrelevant or a matter of time+effort because FOSS”

See also the discussions of Vulkan taking over the world vs other APIs even though very very few things use Vulkan.

Or the perennial “why doesn’t everyone use Linux”/year of Linux in the desktop.

Don’t get me wrong. I use Linux and Vulkan, and I appreciate FOSS. However I think there’s a bit of zealotry around making FOSS products the one true one in their domains, which often leads to the myth leading the actual product.

There’s also a bit of the “free market solves all issues” with RISC-V where there’s an expectation that it’ll cause many designs to spring up and race to a cheaper bottom.

In the not to distant future your commercial operating system will police everything you do on your computer, checking if you're not talking to underage girls too much and whether you're looking at kiddie porn.

It's this kind of crap that has me moving to Linux Mint. I see my computer as my property with which I can do whatever I see fit. Even if it's illegal.

Open source hardware and software are the only things preventing these Orwellian schemes.

Advocating for Linux as a means to watch child porn and groom children is a hot take I didn’t think I’d be hearing, even from Stallman at his worst.

I get your general point of complete privacy+control but, man, that’s definitely NOT the argument I think one should be making.

I'm not advocating anything. Talking to minors is not an offense and I find it despicable that any Social Network would mark that as "suspicious."

And my point is that I want to be able to use my computer for whatever I see fit without some company peeking over my shoulder to judge me.

I'd say that SiFive has already proven you wrong with their cores which compare favorably to ARM9.

IMHO RISC-V is already up to par with all but the most exotic ARM cores and will soon surpass it.

Well same (well more tbh) could be said about PowerPC back in the early 2000s when comparing it to x86. Look how that turned out. I mean it might happen but I don't see how is this in anyway a foregone conclusion.

> wrong with their cores which compare favorably to ARM9

ARM9 is a 32bit core, the last design of which was released back in 2006. Are you saying RISC-V is competitive with ARM chips from 15 years ago?

IMHO RISC-V seems to be in a spot (in relation to ARM & x86) where ARM was back in ~2000 compared to Intel. Very promising and superior in certain case, but still a very long way to become competitive on the higher end. Well it took 20 years for that to change, why would RISC-V do any better?

They're definitely confusing ARM9 with ARMv9 (the latest one with aarch64).

SiFive IS competing (and doing quite well) with current ARM cores.

So the fastest SiFive chip seems to be the P600? Are there any benchmarks where it’s compared with similarly priced ARM chips? I couldn’t find any

> obsession with RISC-V on HN

I don't know, because even though HN mostly lives in a bubble, RISC-V is exploding everywhere it might be relevant.

Yes, it's definitely a legitimate competitor to ARM on the lower end, embedded and various specialized applications.

> RISC-V is exploding everywhere

Because it's open source. Designing your own ARM cores was and is prohibitively expensive and virtually impossible for x86. That might be a huge advantage that will allow RISC-V to supplant ARM but it's not obvious that this will happen.

It's only a matter of time before RISC-V takes over the market completely, even high-end stuff.

Currently the biggest gains are to be had from small embedded processors where the ARM licensing fee makes up a hefty percentage of the total cost. It's here where you see RISC-V making inroads very quickly.

BTW this will also eat away at ARM's bottom line because the company is mostly focused on the embedded market.

> It's only a matter of time before RISC-V takes over the market completely, even high-end stuff.

Why? I suppose at some point there were people who were saying the same about Itanium, Power, MIPS and other. Some of them were quite successful in some areas. RISC-V will probably do great as an architecture for 'small embedded processors' sure. How does that mean they will outcompete ARM? It took decades for ARM to be competitive with Intel for server PCs and servers. Why would it be any easier for RISC-V? Sure opensource is cool, but nobody will be giving out designs of high performance RISC-V cores for free...

> BTW this will also eat away at ARM's bottom line because the company is mostly focused on the embedded market.

Increased competition is great, maybe ARM will have to adjust their business model to compete. Maybe their lose the low-end to RISC-V. But what incentives would Apple or Qualcomm would have to switch to RISC-V? The cost to develop high performance cores would outweigh any licensing costs and I really can't think of any reasons besides that.

Because the lack of licensing will foster innovation!!

Look at where RISC-V is today. It's already competitive with most ARM products and I'm convinced someone somewhere is already concocting a super-duper RISC-V core that blows the doors off Intel and ARM latest offerings.

Qualcomm and Apple will merely want to pump up their profits even higher and ARM is an obstacle, both in licensing costs and terms. With RISC-V they can do whatever they want and they're not beholden to the whims of some other company.

Both are already designing their own cores so it really makes no difference to them, except that with RISC-V they can innovate even more.

> Look at where RISC-V is today. It's already competitive with most ARM products

Not at the high end. AFAIK it's still very far.

> Qualcomm and Apple will merely want to pump up their profits even higher and ARM is an obstacle, both in licensing costs and terms

Well Apple probably can do whatever they want with ARM, so I highly doubt that's a good reason to switch.

> Because the lack of licensing will foster innovation!!

I mean maybe. TBH I'm not too certain though. It might be the opposite. I find it hard to imagine any competitive high end RISC-V cores will be significantly cheaper to license. Also since it would be very expensive to develop them and there is very little money in licensing tech to others (ARM is not doing that great financially) companies which can build competitive cores might have zero incentive to even make them available to third parties. e.g. why would Apple, Qualcomm, Samsung give away their tech to their competitors? Now at least everyone is on a mostly level playing field and can license same designs for ARM, why would that be the case with ARM?

You would have multiple companies licensing high-end RISC-V cores. And competition will almost always translate into lower prices.

It wouldn't surprise me that you could even get a "source code" license for the core and modify it to your needs. AFAIK this isn't possible with ARM (you can either roll your own from scratch or use a off-the-shelf ARM design but not modify it).

> You would have multiple companies licensing high-end RISC-V cores

Perhaps. But why would they license their cores to their competitors instead of just making them themselves. Licensing does not seem like a great business mode in this industry. Has anyone based ARM been successful at it? And even ARM itself is not doing that great financially not even close to Nvidia, AMD, Intel, Qualcomm etc.

The industry is full of "the next big things" that now only exist on archive.org.

> The curtain closes, and RISC-V has replaced it.

For Intel? That would even be a slightly less absurd statement than for ARM.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact