> Having spent years in the open source technology movement, I can attest to the hunger for community-driven solutions," Art Swift, president of Wave’s MIPS IP business, said in a statement. "However, until now, there has been a lack of open source access to true industry-standard, patent-protected, and silicon-proven RISC architectures."
Softbank's investment theses are interesting. They represent a kind of vision I haven't found an equivalent for in the western VC culture. But this article feels a little soft...
> Yet Mr Muller’s “drawings” are anything but simple. They are computer code which give Arm’s customers a blueprint for the construction of microprocessors, information-processing machines so complex that firms are happy for Arm to shoulder the burden of their fundamental design. Those clients—consumer-hardware giants such as Apple
A question for anyone who knows--my impression of the Apple-Arm relationship was that Apple primarily needed a new ISA to do an end-run around Intel's x86 stranglehold and found an avenue via ARM? For the recent chips, I thought large majority of the silicon was custom designed in-house, with Arm's own designs serving as a loose blueprint.
Is Arm really delivering value here to other firms? And is it sustainable or meaningful in the long-term?
Apple could have made their own ISA, or used one of any number of other ISAs.
ARM provides a couple things. First, they offer a progression. You can start out with a drop-in ARM core, and as you progress you can start customizing, and then eventually you can make your own design. (Indeed, the original iPhone CPU contained an off-the-shelf ARM11 core, and the Apple A4 was an ARM-designed Cortex-A8 with some speed tweaks by Apple)
Second, obviously it's a proven starting place, you know it works well. It has been debugged, and verification is established. ARM might even license out the formal verification testing, which is a big chunk of work to develop. Designing your own ISA just takes time.
Third, there's an established software base and other developers writing for ARM already. BSD already ran on ARM when Apple started making their own core, and gcc/clang have supported ARM targets for years.
It's a little bit like licensing USB. Yeah, you could design your own connector with new cables and protocols and electricals. Maybe it would even be better. Or maybe the dominance of USB, proven design, extensive software stack, and so on is worth something.
As for providing value, as with USB, ARM will continue to develop the ISA, guide the way for the many other smaller companies, and through their inherent leadership as the sole owners of the ISA, generally keep things cohesive & interoperable.
> It's a little bit like licensing USB. Yeah, you could design your own connector with new cables and protocols and electricals. Maybe it would even be better. Or maybe the dominance of USB, proven design, extensive software stack, and so on is worth something.
It's a lot like that, except smaller companies get shut out entirely. This has been changing somewhat for the better, at least.
At Apple's scale, is there much financial advantage to switching from ARM to RISC V for iOS devices? Does Apple have to pay ARM any per-device royalties?
That's all closely guarded, I'm not even sure if that's public knowledge or not. Worth noting, however, Apple's quarterly revenue is like $50B while ARM's is $500M- split across many customers.
Apple certainly has shown a proclivity for such bravado, but I think there's no way it wouldn't be a huge setback to the business & the competitiveness of their chips, and breaking free of Intel ASAP is probably worth a lot more than any royalties they pay ARM.
>That's all closely guarded, I'm not even sure if that's public knowledge or not.
Well it isn't closely guarded really. Apple has an architecture license, which is the most expensive one time cost license but pays little to no cost per chip. Since none of the ARM designs fits Apple criteria anyway and Apple are investing into design of their own. Remember Apple has an ARMv8 64bit SoC Shipping before ARM themselves even had a ARMv8 blueprint design for PreOrder.
So in terms of Financial incentives, Apple has no reason to switch to any other ISA.
There are 70 million active MacOS users, and closing in on a billion iOS users. MacOS and the hardware that runs it is an afterthought for Apple, they just want to keep it somewhat alive.
Redeveloping for a new ISA was feasible when it was a life or death move Apple had to make to stay up with the rapidly improving PC world.
As is Apple will likely build a customized Ryzen chip with the AMD secure enclave cut out if they want to do anything for their MacOS line of products. China literally bought the full design for Ryzen for $300 million, a custom chip would be well below that cost.
Apple will almost certainly not switch to AMD rather than going all in on ARM, as much as I wish I had a Ryzen Mac option.
Also Apple having $50B in revenue vs $500m for ARM doesn't mean they would be guaranteed to out execute the entire loosely federated ecosystem starting from scratch.
It's possible I suppose, but is unlikely to be a more efficient way to allocate financial resources.
Worth noting that Apple has a long history with ARM dating back to the Newton, plus the established low-power advantages it holds.
Considering that Steve went to Intel to discuss the possibilities of them making the iPhone processor, I don’t think independence from Intel was a primary motivator.
Sustainable? I doubt anyone knows. But certainly ARM currently offers a significant value proposition.
ARM honestly has little low power advantage intrinsically speaking as an isa, it's just that most low power cpus tend to be ARM and most arm cpus tend to be low power, same with x86 but for high power and performance.
I have yet to see an x86 ISA that can scale down to the power requirements of ARM. So while the performance/watt might be comparable it seems like getting an x86 design to run at such a low wattage is difficult - otherwise Android x86 would have taken off more successfully.
FYI: He believes that the singularity is coming. Apparently he has had this dream since his childhood. In another Japanese interview, he repeatedly said that hyper-intelligence will make people happy.
singularity is mostly an exponential growth in capabilities, closed loop cascade from all technologies; of course people think they'll be able to augment humans to infinity but it's a large ignorance about humans, limits, balance and wisdom.
It's just tech pornography, unlimited blah, people will try and they will realize it's mostly a useless thrill.
They're already thinking of relisting ARM. I can't help but thinking risc V is invalidating their plans for Arm and they want out. Some big customers are designing their own CPUs now and that should be showing up in Arms sales projections.
A source for which statement? I made a few. If you mean that they are thinking of relisting ARM, the linked article that this comment thread is under says that softbank is considering relisting ARM (already!).
If you want a source for my statement that big customers are designing their own CPUs then I could offer that Samsung is a current ARM customer that is known to be developing RISC-V cores:
It seems hard to find proof of what WD will be replacing with RISC-V but it seems extremely likely that its ARM, and here is one link that indicates ARM in a WD product:
We also know NVIDIA is replacing their Falcon GPU controller with a home-grown RISC-V core, but that will not displace ARM. They did indicate a desire for "rich OS" support, which they currently do with ARM. Here's their old presentation:
Looking around, it's easy to find companies supporting risc-v but hard to find direct statements indicating its use to replace ARM.
Anyway, given all the activity I think ARM sales representatives have a good view of their future prospects. They undoubtedly have some preliminary numbers on future ARM displacement by risc-v. If I were on an investors conference call with them (which they don't have because they're owned by softbank now?) I would ask directly about this very thing.
> Having spent years in the open source technology movement, I can attest to the hunger for community-driven solutions," Art Swift, president of Wave’s MIPS IP business, said in a statement. "However, until now, there has been a lack of open source access to true industry-standard, patent-protected, and silicon-proven RISC architectures."