[ChuckMcM]

It has been interesting to watch the ARM ecosystem unfold. Sort of like seeing a levy breach where just a trickle of water starts flowing over it and then more and more and more.

Apple has the resources and the motivation to build a "desktop" 64 bit ARM processor. The history here is also interesting. After debuting the Macintosh in 1984, and getting into fights with Motorola over what should be in the next generation chips, they did the unthinkable and adopted an entirely new instruction set architecture, PowerPC.

That relationship ended on a fairly rocky note and Apple did not have the IP rights to continue PowerPC development or the infrastructure access. They switched to Intel's architecture with great fan fare and started using ARM in their mobile offering. Now ARM has reached the point where it is adopting "high end" features faster than Intel can invent new ones. And Apple has both proven to itself that it can design and fab a completely bespoke CPU design without any input from the original owner.

So here we are looking down the barrel of "bespoke chip architecture round II."

It gives Apple some unique advantages that neither Microsoft nor Google can match.

[Tloewald]

ARM was bespoke round 1 -- Apple was an early investor and used ARM chips in the Newton family (also some laser printers I believe). PowerPC was round 2. Apple acquihired a Boutique chip design firm (PA Semi) that specialized in low power PowerPC designs to design its ARM cores, so the current round is almost a hybrid of the earlier two.

The 65816 (used in the IIgs) was also pretty much bespoke.

[cmrdporcupine]

Apple didn't adopt PowerPC after fighting with Motorola and it was no shock to Motorola, or the world. Motorola and Apple and IBM worked together on PowerPC in a consortium, after the Motorola 88000 RISC (which Apple also tried) failed for various reasons.

This was all in the exiled-Jobs years while Apple was thrashing about trying to build a real operating system (or systems) to replace classic Mac OS. (NeXT also played with the m88k.)

And ARM wasn't new to Apple with the iOS systems, either. The Newton was built around ARM back in the 90s.

[scholia]

Yes. In fact, the Newton was the reason ARM was formed as a separate company. The chip design was originally done by Acorn (using VTI tools). When the ARM operation was spun out of Acorn under joint ownership (Acorn, Apple and VTI), they changed the A from Acorn to Advanced.

[ben_jones]

It's really interesting how the historical strengths of companies can persist even after decades of evolution. If the people are coming and going, how is this institutionalized success being cultivated and maintained?

[ChuckMcM]

To be honest I don't know the answer to that. That said, Apple has always had corporate value of being maximally responsible for their own destiny. It's hallmark card wisdom to know the things you can change and accept the things you can't change, but in Apple's case their history has been filled with finding things we can't change and replacing them with something we can. So where one company might say "Well we have to work with what ever the microprocessor company will agree too" Apple wants to say "The microprocessor company will do what ever we require of them." And if they can't find such a company they look at becoming a microprocessor company.

A lot of that thinking was laid out fairly extensively in the lawsuit over manufacturing sapphire they got into. The manufacturer (GT Advanced) complained (reasonably I think) that Apple's contracts were so onerous as to make them employees of the company in everything but name. And that appears to be how Apple likes it, complete control of their destiny when ever possible.

[kevin_b_er]

So is the lesson taking contracts from Apple is not a good idea?

[Steko]

Apple University is one of the things Jobs set up to try and keep Apple fresh:

https://www.nytimes.com/2014/08/11/technology/-inside-apples...

http://www.asymco.com/2011/05/08/codifying-asymmetry-how-app...

[95014_refugee]

At least part of it is that the people aren't coming and going.

When new, key players arrive they're attracted and vetted by the existing community. And that community is well populated with folks that have long tenure.

There are plenty of folks at Apple in key technical positions with 10+ years at the company, and not a few with more than 20.

[beokop]

Apple isn't going down this road again because it's "in their culture", they're doing it because it puts them at an advantage and because they can. Changing the architecture means changes from the silicon level to the application level; it's not so much about engineering talent as it is controlling the entire stack. Still, 30 years later, Apple is the only company that could pull it off.

[jon-wood]

There has to be at least a little bit of culture having an impact here, even if only in the sense of people being able to say they've done a wholesale change of architecture before so they should be able to do it again. Also the culture of being responsible for the entire stack including hardware, OS, and software gives them a lot more flexibility than somewhere like Dell where they'd have to go through extensive negotiation with Microsoft to make a change to CPU architecture fly.

[sjwright]

My guess would be that such specific historical strengths are local manifestations of a more general instinct that is nurtured within a company culture. This instinct would result in similar behaviour when presented with a similar situation — for better or worse.

I also suspect that a company like Apple would have a portion of employees that stick around for the very long term, who have been there since early days and can act as the company's institutional memory.

Specifically in this instance though, it's pretty clear that Apple is going down this road because they've already started going down this road. This is all just straightforward technology and skill reuse: the touch bar is pretty much an Apple Watch in a different package.

[wmf]

Apple did not have the IP rights to continue PowerPC development

Didn't they get those rights when they bought PA Semi? I think it didn't make sense for Apple to develop their own processor in 2005 because they were just in a much weaker position than they are now.

And it was said that in the old days Exponential got their PowerPC license from Apple.

[ChuckMcM]

I don't believe PA Semi ever had a transferable license to Power. They had their own architecture which was very power efficient and they were a CPU design house with ARM experience (they did StrongARM at DEC). What it looks like from the outside looking in, is that Apple bought out a full architecture license (with derivative ownership) from ARM and bought PA Semi to be the core team to start building CPUs that they had 100% of the rights to.

[nickpsecurity]

That's how I saw it. I did recently discover PA's PPC looking into various implementations of the architecture that might be salvageable. It was quite impressive. Still better specs on 65nm than the Rocket RISC-V that's on 45nm:

https://en.wikipedia.org/wiki/PWRficient

EDIT: Funnier is it was used in a desktop for Amiga's (AmigaOne X1000) before acquired by Apple. Had to hurt what little ego the Amiga people had left.

[ChuckMcM]

Linked in there was the EE Times article (http://www.eetimes.com/document.asp?doc_id=1168406) which refreshed my understanding of the licensing issue. It reads like the Power license they had would have to be re-negotiated with IBM as a transfer if they were to keep selling and making their parts.

[nickpsecurity]

Another source I had said they do sell EOL'd parts to legacy customers. License must have went through at least for that.

[caleblloyd]

Windows RT and ChromeOS have both supported ARM for a while. In the case of Windows, adoption is low because not all applications work on ARM. Apple would hit the same problem initially and would have to wait for application developers to distribute ARM versions.

[zimpenfish]

> would have to wait for application developers to distribute ARM versions.

Or they can switch to their bitcode bundles for Mac apps like the iOS apps and transcode for Intel / ARM on delivery. Plus they have previous form in this area with the "run PPC apps on Intel" via Rhapsody (IIRC.) It'll be a stumbling block, nothing more.

[caleblloyd]

I'm mainly talking about unmanaged apps, not apps distributed through the Mac App store. I don't think that x86 emulation is an option, that would be terribly slow.

[zimpenfish]

If they still have the Rhapsody tech, it's not strict emulation, it's more like transliteration / JIT; much like Transmeta were working on, IIRC.

[caleblloyd]

Ah, Binary Translation at the silicon level. That would be very interesting. Apparently NVIDIA's Project Denver was supposed to do this for x86 but they were unable to get the proper licenses, so they shifted it to do it for ARM. https://en.wikipedia.org/wiki/Project_Denver

[eru]

> It gives Apple some unique advantages that neither Microsoft nor Google can match.

Isn't Google working on some in-house chip for training neural networks? (That's only server-side as far as I know, but still.)

[sametmax]

If I recall, Jobs also considered manufacturing the hardware was one of Apple's mistake because it's not their core value.