Hacker News Reader

Comment by twerkmonsta | Hacker News Reader

Doesn't this all feed into the idea that Apple is ditching x86/x64 and going full ARM?

It would make sense that the Apple TV is so much more powerful considering Apple pours so much into custom ARM chip design and OS development. MacOS doesn't work on ARM, and it doesn't seem like Apple is interested in porting it. Seems more likely the new Mini, if there ever is one, would be ARM based.

View on HN

I would be shocked to an extreme degree if Apple hasn’t been maintaining a version of macOS for their A-series chips for several years now. There’s precedent — they kept the lid sealed on their x86 OS X build for half a decade before the Intel transition was announced.

So while I agree that updates to the Mini are being held back by a pending ARM transition, I don’t think it’ll run iOS. It’ll run macOS for ARM.

> they kept the lid sealed on their x86 OS X build for half a decade before the Intel transition

Do you think they’re still the same Apple though?

I wouldn't be surprised if there were builds for Mac OS X on ARM ever since iPhone came out.

If you read the Walter Issacson biography, it also implies this.

I have this sitting on my shelf a long time and I'm so embarrassed I haven't read it yet.

Again, I think that Apple is a very different beast even since then ...

> ever since iPhone came out.

That's more than 10 years ago. It takes a certain level of competence to actually execute upon this. I wouldn't call it a done deal.

Per-core performance has been stagnant in the past decade. Apple already appears to beat Intel's Y-series processors in most things. Scaling the un-core parts up to compete on desktop will take some doing, but scaling up into the relatively restricted 15-28W range is much easier.

Most importantly, Intel charges a premium for ultra-low power chips. If Apple can make a chip that's 80% of the performance for a couple hundred less per unit, it's a great deal. Moving everything to a unified ISA reduces design and developer costs and increases total control (aka vendor lock-in) which also increases profits.

Apple was a co-founder of ARM Holdings (along side Acorn and VLSI Tech). I don't know how much stock they hold now, but Wikipedia says they held almost 15% in 1999.

In the 2010-ish time, there was a huge mystery corporation looking into buying MIPS before it was ripped into pieces (with the actual MIPS design team going to Imagination Technologies). It's worth considering that ARMv8 64-bit is extremely similar to the MIPS ISA.

Insiders have expressed surprise at Apple's progress. It takes 4-5 years to make and launch a new architecture (and A7 was completely new in every aspect). The starting mark for a third-party design should be from Oct 2011 when the ARMv8 ISA was announced. What's the actual time to a shipping product?

* Apple A7 -- (Sept 2013) 1 year 11 Months

* A53 -- (mid 2014) 2.5 years

* Apple A8 -- (Sept 2014) 3 years

* A57 -- (Q1 2015) 3.5 years

* Apple A9 -- (Sept 2015) 4 years

* Kryo -- (Dec 2015) 4 years 3 months

* A72 -- (Q1 2016) 4.5 years -- basically a fixed A57

How did Apple design such a fast chip in less than 2 years? They did not. Even Intel with a virtually unlimited budget, many more designers, and in-house fabs can't make that happen (PA Semi was good, but not that good). Apple had to know about the new ISA when apparently no other companies did. If ARM told only Apple about the upcoming ISA, that could easily run afoul of regulators.

Here's what I speculate happened. ARMv7 is holding back chip design. Apple looks into buying MIPS because its a great ISA, but changing architectures without backward compatibility is super risky (this was during the app war period). You don't do risky if you don't have to.

Start designing a chip using a new ISA that's very similar to MIPS (a known good starting point to reduce design time). Approach ARM with the options. You can turn that ISA into a new MIPS ISA pretty easily or you can hand over that ISA for ARM to release with guaranteed backward compatibility. Between the long-time business connections, fear of losing Apple, and the fact that it's a good design, ARM agrees. MIPS is divided up with ARM buying rights to most of their patents (basically eliminating the chances of ARM getting sued over ISA similarities).

ARM starts on their own designs and releases the ISA 6-10 months later. Apple chips launch when expected (and have a multi-year lead). ARM chips launch when expected given a few months insider info. Kryo and other third-party cores launch when expected compared to the ISA release.

If I had Apples resources then I think I'd be working on a ARM (or maybe even RISC-V) box that would replace the Apple TV and Mac Mini and serve as a new home IoT server. It could serve as a home based backend to a constellation of Apple devices and services.

The current Apple TV could serve that purpose - no reason for it to run macOS. Or even to be any faster, the 4k Apple TV is surely fast enough.

I'd agree that processing power I think is pretty reasonable, it would be nice for it to handle storage too. Imagine if one were editing a set of video on an ipad pro, or fast backup/restore of other idevices, or you wanted security video peripherals, etc.

Sounds like you just described HomePod.

Yep, but the AppleTV already acts as a "Home IoT server" alongside the HomePod. (and an iPad can it too.)

HomeKit uses one of those 3 devices to allow external access to your local sensors. For instance, I get notifications when the front door is opened and closed. I can unlock it remotely too. Am slowly adding extra sensors.

From the cousin comment: > security video peripherals See HomeCam https://homecam.app

Unfortunately you can't run any old app in the background on any of these closed devices. I have a quad-core Mini that does duty as media library, software update caching, video encoding and photo export... if I could do those on an ATV it would be great.

It seems like for the homecam.app there's no storage of the video stream, or one would need to add another box/device for that?

I haven't bought into any home automation hubs yet, it just doesn't seem quite there in terms of being worth it from any of the vendors, and I think if it were, I'd just run a full up server for a central hub.

> Doesn't this all feed into the idea that Apple is ditching Mac and going full iOS?

FTFY

This theory actually got Jossed in light of this year's WWDC keynote

Sounds familiar...

https://www.youtube.com/watch?v=sPAsWAcmrlE

https://www.theguardian.com/technology/2012/jul/19/ipad-prot...

Apple has a history of denying something until it is working to their satisfaction.

That doesn’t seem like the best example. I’m not sure what the context was for that video, hit it doesn’t look like Steve said that tablets would fail. He said that a tablet made in that era would only be useful to rich people who want an extra device just for content consumption. That’s still most of the market for tablets today.

That's the point. Apple generally doesn't make forward-looking statements about its tech. They say that everything they don't do yet is wrong, and when they get it working they say it's so right. The "right now" part is implied.

I missed the keynote. What happened?

Craig asked the question, then advanced to a slide that just said, “No.”

Reply on news.ycombinator.com