Apple brought their chipset in house, using ARM is a base, to control their supply chain, cut costs, and allow for a RISC based architecture (which honestly just makes sense now these days). The biggest hinderance to RISC-V adoption will be if ARM remains customizable, cheap (from an IP perspective), and has widespread adoption. Honestly, I'm okay if RISC-V remains somewhat niche, but provides the necessary market pressure to keep non-open competitors honest. There's a meta-benefit to open standards as well.
ARM is already attacking one of their largest licensees, and RISC-V is already in use at Apple. As Apple’s hardware accelerated x86 translation proves, they can proficiently transition to any hardware platform. The ISA really isn’t that important anymore, and I expect Apple to do what makes the most sense regarding price/performance.
Well switching to a new architecture is still probably extremely expensive which would permanently shift price/performance in favor of ARM unless RISC actually provides something unique.
Freedom from licensing etc. shouldn't be a problem for Apple unless maybe if they want to sell their chips to 3rd parties which doesn't seem too likely.
My suspicion is they'll fork the arm architecture at some point (either via a legal agreement with arm or just taking the deploy the lawyers approach and doing it unilaterally). Apple like their walled gardens and closed eco-systems so there's no clear reason for them to switch to RISC-V whilst they're very happy with their arm eco-system.
I worked at a semiconductor company that had an ARM architectural license and we implemented our own ARM CPU. We still had to pass ARM compliance tests.
I heard from some of the people on the CPU team that Apple did not have to pass those tests because they had an exemption from ARM. Apple was one of the original ARM investors when it was spun off from Acorn in 1990. I think that agreement allows them to do things with ARM like adding custom things that other ARM licensees aren't allowed to do.
You are making it sound like a burden, but I do not understand why it would be. Surely you are interested in safety checks giving you more confidence that you are positioning yourself to take advantage of the Arm ecosystem (instead of releasing something that is not supportable by compilers and OS experts)?.. Or am I misunderstanding the nature of the compliance tests?
I think the compliance tests were making sure that your add instruction or memory load instruction followed the spec.
You mention ARM ecosystem and that is precisely the point. Apple controls their ecosystems. How do you write apps for the iPhone? You use Apple's development environment with Apple's compiler. If Apple decided NOT to implement an instruction for some reason they could simply make their compiler never output that instruction.
I worked on chips that ran embedded applications with no ability for ordinary users to change the software. What is the value of meeting an external ARM controlled spec for that?
I also worked on chips that only ran Android and nothing else. If you are also the company porting Android and writing all drivers for your own platform then people may argue whether it is worth only being 99% compatible.
Later I worked on chips where people may run Android, Windows Mobile, or plain Linux on this chip using GCC, Clang, Microsoft's compiler, or whatever. For that you definitely wanted to comply with specs.
In a long time horizon, it's probably in Apple's interest to migrate away from ARM just so that they have one less company to negotiate with or receive restrictions from
Apple helped to design Arm v8 they probably have huge influence on the direction of travel of the ISA. Plus they have huge investment in Arm based software and probably get access to all of Arm’s IP (eg big.LITTLE).
They will migrate when there is a major performance advantage (as in the past).
I've had the same thought and took it even further. The cloud companies are in the business of managing capital, namely datacenters. They are not in the business of designing µarchs; to the extent that they do (ie. Graviton) it is to lower costs. For this reason I belive it is economically favorable for them to jointly create and fund an open µarch. Note thay they are already doing this through a middleman known as the Intel/AMD monopoly. Also note that a similar thing has happened already with Linux. Collectively throwing a few hundred million at the problem (or buying sifive, for example) could save money in the long run. Although I can't be certain about that.
I hope someone pitches this idea to these companies. I think some people at Google may be thinking in this direction being that they are paying for tape out of open hardware at no cost to the developers and are investing in tooling.
That's what Alibaba is already doing too. They bought T-Head and have been releasing open source RISC-V cores that have been taped out in a move to "commoditize your complement", thus bringing total costs down by making their vendors' markets more competitive.
> Why if Arm is correct does that add a risk that it's the bad actor?
Because nobody wants to fight over the licensing of a product they're shackled to. If you could buy a RISC-V board with software support similar to a Raspberry Pi, ARM's goose would be cooked. Every enthusiast would ditch ARM in a heartbeat for a more open ISA, and ARM licensees would see it as an opportunity to finally wiggle free of ARM's insane license restrictions. All we need is the software support, which should be pretty forthcoming since most projects have already been optimized for RISC.
ARM could unseat x86 because both ISAs were encumbered with licenses at the time. Now, ARM is competing with much less restrictive architectures, and all it would take is a FOSS RISC instruction set to ruin their value prop.
You're assuming high end RISC cores will be available for free and or under much more favorable licensing than ARM cores. Which seems unlikely, why would someone sell their cores to a competitor for less than a company whose only business is designing them?
Same reason why people would contribute to an open source compiler and toolchain and distribute for less than the cost of the old school paid compiler vendors like Borland. These contributors aren't really in the business of selling compilers and simply have strategic reasons to drop the floor of that market as much as possible. The same applies to RISC-V cores, with probably the most prominent example being Alibaba/T-Head and their open source cores.
Sound more like another "this is the year of Linux"..
Also hardware is pretty different than software. Hardware design (and obviously manufacturing) requires much more significant investment.
> These contributors aren't really in the business of selling compilers and simply
That's the thing. Apple was/is not in the business of selling compilers, nor are the people/companies who contributed to gcc and most other open-source projects. They have basically nothing to lose and a lot to gain from contributing to open-source software.
It's not obvious to me CPU design could work that way. For starters everyone who could develop advanced RISC-V cores is unlikely to give them anyway to their competitors since they would be in the business of selling CPUs. i.e. do you really think Qualcomm, Apple etc. would make their designs free? Why?
> Alibaba/T-Head and their open source cores
Because at this point they have more to gain than by keeping them proprietary. RISC-V is not yet overall competitive with ARM. So it's not like these RISC-V cores could be commercialized that successfully. Keeping them open probably makes further development faster.
> Because at this point they have more to gain than by keeping them proprietary. RISC-V is not yet overall competitive with ARM. So it's not like these RISC-V cores could be commercialized that successfully. Keeping them open probably makes further development faster.
This open source core is about equivalent to a Cortex-X1. https://github.com/OpenXiangShan/XiangShan A collab between Alibaba, Tencent, and the Chinese Academy of Sciences.
So you believe that firms should abide by contracts but that firms shouldn’t try to enforce the terms of those contracts.
I find it odd how Qualcomm is portrayed as though they were innocently minding their own business when they suddenly got sued. They bought Nuvia knowing what was in the two sets of contracts.
Also Qualcomm had and still has the ability to design their own Arm cores.
> So you believe that firms should abide by contracts but that firms shouldn’t try to enforce the terms of those contracts.
No one is saying that they should or shouldn't (or at least I haven't seen anyone saying this).
But, when you exercise a right, there are social consequences. Freedom of speech is a right (in the US at least), but picketing funerals will still get you dis-invited from a lot of parties. Suing big customers (unless you're obviously in the right when viewed from the outside) will make at least some people more nervous to do business with you.
I'm not talking about Qualcomm, really. I'm talking about companies like Apple, who really only have a cursory attachment to ARM as an ISA. Then there's the hundreds of smaller manufacturers who have zero attachment to ARM and would much rather build hardware on their own terms. Those are ARM's moneymakers, and those are the companies that frankly have the most to gain from using RISC-V.
If Qualcomm is a relevant topic regarding ARM's success, then they've arguably already failed.
Does Apple pay anything meaningful to ARM? They're a founding member with (I believe still) a significant stake. I find it doubtful they didn't secure themselves a perpetual license when they founded the company, and that seems to be what the Internet believes to be true.
I'm not so sure it's the expensive large chips, made in relatively small quantities, that make ARM the most money, do you have a source? I'd have guessed they actually make more on the billions of small ARM cores that ship every year that end up by multiples in pretty much every device with a battery or power cord. And these, I think, are at the biggest risk of leaving ARM. RISC-V development is mature enough at this end of the market that it's relatively easy for users to transition, there are multiple competitive cores on the market, and there's no concern here about backwards compatibility because these are embedded systems where there's usually not an expectation of having to run user code at all. It will be much harder for the likes of Qualcomm where there's a huge ecosystem built around their ARM processors - but as a share of cost-per-processor, they probably stand to gain the most. Qualcomm is a founding member of the RISC-V foundation after all.
> They're a founding member with (I believe still) a significant stake
No, ARM is 100% owned by SoftBank. Apple sold their shares a while ago.
Overall ARM doesen't really make that much money, especially compared to some of their clients. It's not even obvious to me if would make sense financially even for Qualcomm to design their own RISC-V cores compared to licensing from ARM.
I mean they could and did design their own cores but they are still using ARM designed ones for their top-end chips.
> smaller manufacturers who have zero attachment to ARM and would much rather build hardware on their own terms
Or they believe that designing their own cores would be cost prohibitive and prefer licensing them from ARM.
It's also not obvious to me than licensing conditions for RISC-V could be more favorable than what ARM is offering. Why would you anyone even license competitive cores their competitors if they can make more money selling them themselves?
Completely agree with this and with your other comments in this thread.
There simply isn't that much money in designing cores. The money is in selling SoCs or devices. Arm at least has made increasingly high quality cores available at reasonable prices to all comers. A future where Arm's business is made unviable is not necessarily better for consumers.
You can license RISC-V cores from Si-Five today but they if reports are correct they were in discussions to sell to Intel. If that were to happen who knows what would happen to their offering. There will be others of course but as you say it's not obvious that what they offer will be any better that what Arm offers today.
And for the foreseeable future Arm is immune to takeover by any of its deep-pocketed customers.
I'm very happy that RISC-V exists from a number of perspectives but there needs to be a realistic assessment of its potential impact.
You’re right but the point is that Intel was very interested. If you’re relying on cores from firms that are attractive to SoC / device makers then there is no guarantee those cores will continue to be available post acquisition.
It’s not obvious who is in the right, and in my armchair opinion it looks like ARM has a slightly weaker argument. IANAL, but all of this makes ARM appear more litigious than egregiously wronged.
Apple is in a unique position with a perpetual ARM license, so they don't feel any particular competitive pressure from RISC-V. In fact they're adopting it for microcontrollers and such.
