This week on AI: No Mac Pro until 2019, Apple to ditch Intel CPUs, 2018 iPad reviewed & more

Some of us remember what a dead end the Power PC chips turned out to be. Apple moving back to non-Intel chips scares me.

alpha10711 said:

Some of us remember what a dead end the Power PC chips turned out to be. Apple moving back to non-Intel chips scares me.

Well that’s a lot less likely if they’re designing the chips themselves.  

alpha10711 said:

Some of us remember what a dead end the Power PC chips turned out to be. Apple moving back to non-Intel chips scares me.

Times have changed since this happened. Today's computers are roughly 8x faster than they were back then and have 4-8x more RAM. What's become evident the more ubiquitous that computers have become is that all CPU development is a dead-end for all but the most specialist fields of work. If you gave 99% of computing users today a chip that was 100x faster, they wouldn't know what to do with it and likely wouldn't notice the difference beyond the battery life being better. A lot of heavy computing has been moved to the GPU.

Apple today also sells around 5x the amount of Macs since they switched to Intel and while that was partly helped by the switch, it was massively helped by the iPhone, which put Apple back into everyday conversation in a positive way e.g 'hey, this company makes awesome phones, I wonder if they also make awesome computers. Well look at that, they do'. The iPhone brought developers back to the Mac.

ARM chips would likely mean they won't have to emulate iOS development on Macs, which would be better than it is now.

Windows will soon support 64-bit apps on ARM so if Bootcamp is still an option, they can support this version of Windows:

https://www.engadget.com/2018/04/05/microsoft-windows-arm-pc-64-bit-sdk-build/

The main reason Apple uses other manufacturers' chips in the first place is because they originally didn't make their own chips, just like other computer companies such as HP, Dell and Lenovo.

Another difference today with ARM compared to PPC is that hardly anyone used or uses PPC chips. Even in servers, they have single digit marketshare:

https://www.pcworld.com/article/3087817/battle-lines-are-drawn-ibm-prepares-power9-to-take-on-intel-and-arm.html

More than 3 billion people use ARM chips for everyday computing tasks already.

Nobody wants compatibility headaches and I'm certain Apple doesn't either. They wouldn't put in efforts to support eGPUs over Thunderbolt to then throw it all out. They could get Intel to manufacture custom chips for them if necessary. There's plenty of preparation being made for such a transition. Dropping 32-bit draws a line for support of plugins and binaries so people will be well aware of what is supported going forward before any major transition happens and everything that is supported after that should be easily supported on another set of hardware.

I don't really see much to be gained in making this kind of switch for the higher-end Macs (below iMac Pro / Mac Pro) because it would be a small fraction of the price but $100-200 on the low-end can make the difference for large purchasing decisions in education and business and it would save a lot for the iMac and Mac Pro. If that's what they want to target, it's best to take the whole lineup over rather than have two sets of software. It's far enough away that it's not worth bothering about now because CPUs and GPUs will be faster again by the point any switch is due to happen and people will have a different vantage point from which to assess any potential problems. People can easily buy a 2019 x86 Mac and use it until 2029.

Marvin said:

alpha10711 said:

Some of us remember what a dead end the Power PC chips turned out to be. Apple moving back to non-Intel chips scares me.

Times have changed since this happened.

It can’t be reiterated enough. The most important thing about chip performance, strategic importance, are the fabs. In the past, Intel was always the leader, by years sometimes, because of favorable money flows due to their dominance in PC chips. They had the most money to keep pushing to the next node the fastest.

This looks like it is not true anymore because the phone market has a lot more money it, a lot more users, a lot more use. It’s been flat out amazing to watch over the last ten years and seeing TSMC and Samsung catch up. Something a lot of people thought couldn’t happen.

They were at least a full node to a node and a half behind Intel 7 to 8 years ago. The A5X was a 165 mm^2 chip on Samsung’s 45 nm process in 2012. That was a PC chip sized SoC. Intel was shipping 32 nm processors in 2010, basically two years earlier. By the end of this year, Apple may have A12 and A11X SoCs on TSMC 7 nm and Intel will have Cannon Lake Y series, maybe U series, processors out on 10 nm. The two fabs should be pretty equivalent in transistor densities.

My bet is Apple’s SoCs this fall will outright outperform those Cannon Lake processors, and will match Coffee Lake at 15 W even though they are sitting in 5 W TDP devices.

After that, the race is on to 5 nm and Intel doesn’t have much provocation to get there fast, while the ARM world will be eating themselves alive to get the next node. I bet Intel is going be at 10 nm for 5 years. They’ve been at 14 nm for 4 years.

Whenever Apple moves to in-house CPU/GPU for MACs, it still have to maintain MacOS releases for both platforms for a while that may be cost prohibitive to ditch Intel..