Apple have a long history of working with and contributing to open source projects. Where do you think Safari's rendering engine comes from?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Possibly Exploring Open-Source Alternative to Arm Architecture
- Thread starter MacRumors
- Start date
- Sort by reaction score
Sources from the industry claim that Apple had its hand in developing ARM64. They were the first to ship a working ARM64 CPU and a lot of changes in ARM64 are clearly targeted at making deep out of order CPU designs (just like what Apple had in mind) fly. There is no doubt that Apple started developing its CPU IP before ARM64 was officially announced and its very likely that there was significant information flow between Apple’s CPU architects and ARM’s ISA designers. Don’t forget that Apple was one of the original founders of ARM. Connection between these companies go way back.
And Chome, Edge, etc.
Heck, I’ve heard it said that FreeBSD would not be around without steady stream of Apple patches… not to mentions technologies like LLVM that power most popular compilers and devchains today. Or what about OpenCL or WebGPU?Gonna pass on that 2021 M1X MacBook Pro and hold on to my 2012 MacBook till the RISC MacBook Pro!
Once again, people assume and/or lack reading comprehension. It seems people are missing the point: even quoted by Apple, that it is to be used where it can be to reduce licensing fees. So, it probably won’t work for the A/M chips, but smaller chips like battery controller or any other of the other 3 billion and growing chips that are hidden in all the crap we use.
Another way to cut costs and stick it to apple customers with high priced Apple products.
How much of a bonus does Tim Cook get for this cost cutting measure??
How much of a bonus does Tim Cook get for this cost cutting measure??
Lots of Apple haters on a Mac forum! lol
Apple has a 3, 5, and 10 year road map. Apple will likely purse RISC-V and use it as a negotiating chip (pun intended) with Nvidia/ARM. When those negotiations don't go well, Apple will make the move. Just like Apple exploring cellular modems with Intel, which ultimately allowed them to negotiate a better deal with Qualcomm.
Now that Apple owns Intel's cellular modem IP, perhaps Apple is planning to roll it's own cellular modem and combine with RISC-V on a single SOC to cut off both Qualcomm and Nvidia/ARM at the same time. Given the Qualcomm-Apple deal is set to expire in 2025, I expect we might see something soon after, maybe after a short extension if Apple is not ready by then.
www.macworld.com
Now that Apple owns Intel's cellular modem IP, perhaps Apple is planning to roll it's own cellular modem and combine with RISC-V on a single SOC to cut off both Qualcomm and Nvidia/ARM at the same time. Given the Qualcomm-Apple deal is set to expire in 2025, I expect we might see something soon after, maybe after a short extension if Apple is not ready by then.
Apple working on its first cellular modem, which could bring 5G to the MacBook
A new Bloomberg report claims that Apple has begun work on its first cellular modem, which could be a game-changer for the Mac.
www.macworld.com
Last edited:
https://arstechnica.com/gadgets/2020/12/new-risc-v-cpu-claims-recordbreaking-performance-per-watt/
RISC-V has true potential for say a watch or embedded systems.
lol even Intel is looking/exploring RISC-V. You have no idea what you are talking about.
The MacBook Air today is nearly a $1,000.00 less than it cost when it debuted.
I guess it’s probably good to keep their options open, lest they get bottlenecked by someone else’s roadmap viz Intel Macs. That’s probably what this is about at this point more than anything, especially as Arm currently underpins their golden goose the iphone… though I wonder why not then go the whole hog and create their own instruction set?
ARM China recently went rogue and declared independence from the parent company/joint venture.
[Updated, Correction] Arm China Goes Rogue In Hostile Takeover And Starts-Up As An Independent Company
Allen Wu, who serves as the CEO of Arm China, has quite literally gone rogue.
Mac OS is literally built on open source. PHP, FreeBSD, Apache, MySQL, Darwin, etc.
Apple has a perpetual license so licensing is different from paying for every ARM core in each chip that Apple makes/sells, would this not change the license cost/fee?
RISC-V seems to be the best solution, I've read up on it a while back.
I think this is a knee jerk reaction to NVidia going for Arm Holdings, Inc.
There's nothing magical about ARM either. It's Apple's secret sauce that makes their implementations fly and distinct from every other corporation's designs. RISC-V is neat, and very suitable for a company like Apple that want to take a technology and wrangle it to its own liking. Their license with ARM seems to afford that now, as would using RISC-V but they don't have to pay anyone for it. I don't think the Nvidia buying Arm is something that bothers Apple as they've been on a "doing it for themselves" trajectory far longer than that.
WhenApple announced the branding "Apple silicon" I was fully into the idea that they are now masters of their own domain, not beholden to other companies tech. It's ARM for now, but they can take any CPU instruction set and still call it that. So why switch to RISC-V, what would that bring them? Why not do an ISA of their own? It's only the general purpose computing stuff that's of an ISA that's named and common to us, what the heck the myriad of accelerators are running, including the graphics stuff is completely opaque to us, and only exposed through frameworks, APIs and compilers if we're lucky. It's not like Apple is doing talks at conferences how they implemented their neural engine, their compression or encryption accelerators. These can be RISC-V right now for all we know. How is Metal implemented in hardware? Something something tile based rendering.. yeah, but not like Imagination makes their PowerVR stuff, are they? Apple's doing it by themselves.. building instruction sets that they are the master of and that only they know the details of.
ARM as an architecture is good for Apple right now as the software ecosystem is large and sufficiently on par with x86. As is availability of developers and engineers at all stages, from hardware to software. RISC-V is a budding technology right now, without nearly as large ecosystem, and that's a problem for everyone involved. But it's an interesting problem that will solve itself over time, and nothing Apple's shied away from before.. they can hop on a train early if they deem it interesting or just better.
Apple need ARM for the higher level stuff they must expose to their customers, but for the obscure stuff.. the controllers, accelerators, the custom ASICs.. they are probably either so common as they are just picking macros or IP off the shelf, or so bespoke that they have to write everything from scratch. It's this space that RISC-V comes into play for Apple, as it's in this space RISC-V is coming into play for everyone. Fairly few are taking on ARM or x86 in the RISC-V world, they are pretty much all going for the very niche things, from AI and HPC accelerators, to minuscule controllers hardy doing proper integer pipelines. Everywhere where very custom stuff would have to be implemented anyway. And they are not sharing their secret sauce. They are exploring, building know how, seeing what works, where RISC-V is better, where it makes sense. It's the ISA that's free, not the implementations. And it's right up Apple's ethos.
The argument that RISC-V isn't ready for Apple now is irrelevant, their roadmap stretches on for decades. Apple started eying the desktop space with Arm at least a decade ago, when their IC team wasn't nearly as good as it is now. If Apple wanted to do a desktop class RISC-V processor, it'll probably won't take them too much effort. I can see Apple forking ARM to their liking long before RISC-V comes into play for us to bother about. Apple seems to be heavily involved in the roadmap for ARM, at least in ARMv8 and ARMv9, but what's next.. we'll see. Apple can just deviate at will and not really care as long as they are contributing to Clang and LLVM accordingly. That's where we'll see the first real hints of another switch. The switch to ARM was hidden in plain sight as it was no secret the they had ARM in there mobile devices. It'll be harder to hide next time.
WhenApple announced the branding "Apple silicon" I was fully into the idea that they are now masters of their own domain, not beholden to other companies tech. It's ARM for now, but they can take any CPU instruction set and still call it that. So why switch to RISC-V, what would that bring them? Why not do an ISA of their own? It's only the general purpose computing stuff that's of an ISA that's named and common to us, what the heck the myriad of accelerators are running, including the graphics stuff is completely opaque to us, and only exposed through frameworks, APIs and compilers if we're lucky. It's not like Apple is doing talks at conferences how they implemented their neural engine, their compression or encryption accelerators. These can be RISC-V right now for all we know. How is Metal implemented in hardware? Something something tile based rendering.. yeah, but not like Imagination makes their PowerVR stuff, are they? Apple's doing it by themselves.. building instruction sets that they are the master of and that only they know the details of.
ARM as an architecture is good for Apple right now as the software ecosystem is large and sufficiently on par with x86. As is availability of developers and engineers at all stages, from hardware to software. RISC-V is a budding technology right now, without nearly as large ecosystem, and that's a problem for everyone involved. But it's an interesting problem that will solve itself over time, and nothing Apple's shied away from before.. they can hop on a train early if they deem it interesting or just better.
Apple need ARM for the higher level stuff they must expose to their customers, but for the obscure stuff.. the controllers, accelerators, the custom ASICs.. they are probably either so common as they are just picking macros or IP off the shelf, or so bespoke that they have to write everything from scratch. It's this space that RISC-V comes into play for Apple, as it's in this space RISC-V is coming into play for everyone. Fairly few are taking on ARM or x86 in the RISC-V world, they are pretty much all going for the very niche things, from AI and HPC accelerators, to minuscule controllers hardy doing proper integer pipelines. Everywhere where very custom stuff would have to be implemented anyway. And they are not sharing their secret sauce. They are exploring, building know how, seeing what works, where RISC-V is better, where it makes sense. It's the ISA that's free, not the implementations. And it's right up Apple's ethos.
The argument that RISC-V isn't ready for Apple now is irrelevant, their roadmap stretches on for decades. Apple started eying the desktop space with Arm at least a decade ago, when their IC team wasn't nearly as good as it is now. If Apple wanted to do a desktop class RISC-V processor, it'll probably won't take them too much effort. I can see Apple forking ARM to their liking long before RISC-V comes into play for us to bother about. Apple seems to be heavily involved in the roadmap for ARM, at least in ARMv8 and ARMv9, but what's next.. we'll see. Apple can just deviate at will and not really care as long as they are contributing to Clang and LLVM accordingly. That's where we'll see the first real hints of another switch. The switch to ARM was hidden in plain sight as it was no secret the they had ARM in there mobile devices. It'll be harder to hide next time.
Would be interesting to compare that CPU to an S6 in the Apple Watch. The benchmark you have linked has a lot of problems btw… first of all, trivial microbenchmarks are more of academic interest (I‘m not surprised that a simple core does relatively well in naive ALU tests, but how will it perform real-world code with indirect branches and cache misses?). Then the author takes 15W TDP as Ryzen‘s power consumption, which is probably closer to 30 watts for the duration of that test.
I would be much more interested in seeing some lighter SPEC or browser benchmarks. I wonder why the authors didn’t provide those.
Correct, but the savings could be plowed into additional functionality that Apple may be able to implement.
Maybe it’s due to the nature of my own work that I think this, but my guess would be that nothing is going to come of this, because the job listing is nothing more than a deliberate attempt to gain leverage in a contract negotiation by terrifying ARM that one of their biggest customers may walk. Apple just designed an entire desktop chipset and OS around ARM. It would be insane to change processor architecture now.
Supporting an established ISA benefits from having a mature development ecosystem. Compiler support and developer familiarity are worth a lot. Last I checked, LLVM still had issues with RISC-V codegen (of course, the fact that RISC-V has nothing useful for modern high-performance computing doesn’t help either).
I expect Apple to stuck with ARM64 for a while. It’s an elegant, fully featured ISA that maps very neatly to the CPU design style of Apple and it will likely remain scalable for a while. Not to mention that the new ARM vector processing ISA has tremendous potential for HPC applications. Again, from the technical point of view I don’t see a single reason why RISC-V would allow better performance scaling in the future. It’s not that different from ARM to begin with, just more verbose and less complete. In the future, it might make sense to explore some new design space ideas (like mill architecture etc.). When it comes to traditional register-based designs, ARM64 is already very close to being optimal.
Swift is open-source.