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 Signs New Deal With Arm to License Chip Designs Beyond 2040
- Thread starter MacRumors
- Start date
- Sort by reaction score
It's completely inaccurate. This article is so terrible it should have appeared on AppleInsider.
Apple is using only the ARM instruction set. They don't use any designs from ARM, nor have they for more than a decade. (Actually they may be using a few tiny arm designs as microcontrollers, but they're not using them as general CPU cores, which is what we're all talking about.)
I'm fairly certain the latest agreement involves zero dollars. (Or perhaps, ARM is even paying Apple!) "That's a nice instruction set you have there, it would be a shame if anything RISC-V happened to it!" Apple doesn't need ARM for anything, whereas if Apple adopted RISC-V it would be a huge blow to ARM. Apple's ARM chips are the fastest by far; if they went to RISC-V and ARM had no answer, the optics for ARM would be absolutely terrible.
Some of the other errors in the article have already been corrected, but:
WTF is "integrated circuit design data" supposed to be, and how is "blueprint" an explanation (as opposed to a restatement)? Try:
ARM developed many iterations of their ISA (Instruction Set Architecture), and many implementations of that ISA. Those implementations are used by most smartphone manufacturers, while nearly all the rest, including Apple, use chips designed to implement the ARM ISA.
Failing to distinguish between chip designs and ISAs is an egregious flaw, when talking about ARM.
It tells us nothing at all. There is literally zero relevance. The lack of eGPUs is not due to anything ARM-related. It's just Apple's choice.
And me 61ys old, and hopefully still alive.
Time will tell if Apple will still be a healthy company, because the next hyped company is waiting behind the corner, it’s just a matter of time till it pop ups.
My wife has one buried somewhere - it was issued to her for work. I HAVE watered it but is has never sprouted.
Technically started with the Acorn Archimedes, the successor to the BBC Micro. We had a few at Primary School before the PC and to an extent the Mac took over!
Also, I'll add that while this is partially Reuters' fault - they got this wrong in the original article - the headline is also terrible. Apple did not license "chip designs". They licensed the ISA - more a language than a technology.
To be a bit pedantic (but accurate) Apple is not going to license ARM chip designs. They are licensing the ARM instruction set. They design their own chips.
There are ARM versions of Linux and Windows, and they run on M-series Macs currently.
Intel is developing chips similar to Apples, but not sure if they are going to use the ARM instruction set like Apple does.
There is money in it for them as well. If Arm is doing the initial IPO auction where 'privileged class' gets there shares before everyone else. Apple , Google, etc will all have their blocks the 'scarcity games' start. Softbank is sitting on their 90% (at least for a reasonably long while). If Apple and the rest of the gang of six buy up another 3-4% then there is actually only 6-7 out there. The other big players ( pension fund x , y , z . Index fund Semiconductors , 'tech' , 'growth', etc.) squat on another 2-3% and down to 3-4% out there.
For most part Softbank is dumping what they transfer back from their venture cap entity. The IPO is going to cover that and make them some loss covering float ( lost of questionable investments for Softbank to cover over out there. )
If the stock holds at the price after a 1-2 year bidding war those other players can unload small portions for a profit. Softbank is going to . Dropping from 90 to 85% still leaves them grossly in charge and still a relatively small pool of float to fight over. [ If not inflated too high, Apple and the tech companies can incrementally cover that while each individually staying un 5% with very little SEC/Government drama. ]
If stock doesn't hold and things go bad. Apple and the other techs didn't have to buy in during the bidding war and loose even more money. Again if the Tech companies want control they will get more control/$ spent at that point ( and already have a position to trade from so some activity there isn't going to spike the radar much. ) .
If Softbank does something to piss off the tech companies. They all leave at the same time (and make lots of noise when they go " Softbank is crazy ... we're leaving." . several percentages points extremely quick has very good chance of causing a stampede. These companies are not so much "show of support" but price stabilizers.
Pretty good chance that first notion wasn't true. For a specific architecture Apple has license to do something forever. But that notion thrown out there that because Apple was a founder, they get everything Arm ever thinks up practically free for life probably wasn't true in first place. (this isn't at the same scale at all like giving Woz a free Mac every time release new model. Nor is that typically contractual. )
Arm's whole business is selling contract R&D . New R&D should mean incrementally more money needs to be turned over. Even if Apple was a founder , can't make that new , independent entity give away the farm. Where there some practical 'give away the farm' terms so that first Arm SoC used in the Newton. Probably. For everthing that Apple's seed money did NOT cover in the slightest, forever into the future? Probably not. ( the way Arm does contract R&D has evolved over time. What they did then and now is substantively different. So doubtful any terms directly to what Arm did then are going to stick now. )
If being pedantic Apple likely does both. Apple uses smaller embedded Arm designs as controllers. ( SSD controller, ) it isn't just "big" end user application cores. ( there is non-GUI stuff that Apple does too. Remember they bought up Dialgog's subdivision for power manangement IC (PMIC) and have weaved system management SMIC stuff in their own systems also. )
There is about zero good rational for Apple to build most of those low level controllers completely from scratch on bleeding edge fab nodes.
RISC-V is likely going to eat away at Arm's dominance in embedded controllers over time, but for now that is just as broad , if not broader , than the Personal Computer ( from smartphone to classic PC form factor) space.
ah... forgot that one. Wouldn't it be something if one could compile current apps for all these CPU's in xcode?
If no significant decisions can be taken, this could well be the end for a tech company. Tech markets still move quickly and companies have to adjust accordingly.
No, within VM. Parallels has agreement with MS to package Windows for ARM with Parallels. Works quite well on my M1 MBP.
Agreed. The author shows an almost complete lack understanding of ARM or Apple's relationship with them.
Apple "licenses" use of the ISA, but designs their own processors from it. ARM doesn't build anything; they do have "reference designs" but they're basically CAD files.
On the M series the main processors using ARM's ISA are a pretty small area of the chip. GPU, Neural Engine, RAM, cache, etc. tale up the majority of layout. Point being that ARM has almost nothing to do with the performance of the A/M/S SOCs Apple produces.
Now that Apple has proficiency in designing SOCs, they could, technically, move to any other ISA, or invent their own.
Apple "licenses" use of the ISA, but designs their own processors from it. ARM doesn't build anything; they do have "reference designs" but they're basically CAD files.
On the M series the main processors using ARM's ISA are a pretty small area of the chip. GPU, Neural Engine, RAM, cache, etc. tale up the majority of layout. Point being that ARM has almost nothing to do with the performance of the A/M/S SOCs Apple produces.
Now that Apple has proficiency in designing SOCs, they could, technically, move to any other ISA, or invent their own.
Actually, smart money is that Apple *is* designing their own controllers. They definitely had one a few iterations ago called Chinook, which they apparently built by stripping down one of their normal cores quite a lot. (And yes, that was a surprise to me - I would have expected no OoO, maybe 32-bit, etc.) They are likely using some really tiny ARM cores (like M0) where even that's too much, but we don't know. And their efforts with RISC-V suggest that, if not already, they will soon be replacing those cores with their own RV designs. Probably the Chinooks too, or whatever those have evolved into.
There *is* a good rationale for Apple to do this; while it will save pennies, more importantly it gives them total control over their design.
Since nobody got it last time, I'll say it more directly: I doubt Apple is paying ARM anything more, and may even be getting money from ARM. Seeing them start a public defection to RV from ARM would absolutely tank ARM's stock, and in the long run would likely have catastrophic effects on ARM's business. Apple's full of smart businesspeople as well as smart engineers; I doubt this was lost on them.
That's silly. True, the cores are a fraction of the total silicon budget. False that the ARM cores are almost irrelevant to performance. In 2023, the performance most people care about is still CPU, though GPU is also a major factor. Few people use or care about the NPU, and most of the rest just have to be good enough. ("Cache" as you listed it is properly part of the core, except for the SLC, which, true, is quite important.)
Or did you mean that ARM as an ISA is not important? In that case your entire argument seems to be wandering, though you'd be clearly correct as long as you're comparing to other ISAs like RISC-V, and not x86, which carries an expensive legacy.
This was exactly the point I was making in post #52 (with a somewhat clever pun that nobody seems to have picked up on, sigh). Inventing their own is certainly an option but it would be pushing uphill quite a lot, in terms of reinventing tooling, training, etc. RISC-V is probably all they need. They are certainly doing this already for the small cores that are not visible, though we don't know how far along they are, and it's vaguely possible that business considerations (ie, incentives from ARM) could slow or halt this- for now. I doubt anything can keep them from pursuing this indefinitely.
Not surprised. The RISC-V design is just too different for Apple to work with given they've been working with various ARM chip designs for over a decade.