Apple Mac computers make jump to its own chips

[dayjob]

i found this interesting/informative. history of chips.. RISC/CISC/ARM/x86 etc etc. as relevant to apple's move

[FarleyCZ]

tl;dr: I'm not concerned about the processor architecture itself.

Got it. Thanks for the reply. Can I pick your brain a bit more about this pls?

How plausible do you see a possibility of developers just adding and adding and adding new plugin format supports? I mean ... even right now you kinda have to keep support for VST 2.4, VST 3, RTAS, AU, RE ... and god knows what else. There are some plugin formats for ARM already, but AFAIK none of them is so heavily spreaded out that you can build a proper bussines on top of it. So likely in next few years, ARM DAWs will compete to push their own plugin formats to the industry standard status, which might very well end up in the market forcing you to add many more of plugin formats to the list.

Do you think this might happen and if so, will this be managable by an average developer team, or will we see a major shift onto some JUCE/SOUL-like umbrella framework?

[rezoneight]

And for the view from a former MS employee of some repute check out this thread:

https://twitter.com/stevesi/status/1275311056672325633

Is the former MS employee perhaps working now for..Apple? For some reason the first sentence gave it all away..

One need only Google. Lazy lot around here. No he doesn't work for Apple.

[tooneba]

The significant difference between PowerPC-Intel and Intel-ARM transition is ARM has never been able to out-perform latest generation i9, Xeon, Ryzen, Threadripper in general purpose computing.

[Tronam]

The significant difference between PowerPC-Intel and Intel-ARM transition is ARM has never been able to out-perform latest generation i9, Xeon, Ryzen, Threadripper in general purpose computing.

Apple doesn't use ARM reference designs, so past examples may not be a very good reference for comparison; They only license the instruction set. They've been designing their own microarchitecture since the iPad/iPhone 4 onward and for the past couple of years their own GPUs as well. I won't be surprised at all if Apple silicon Macs over the next 24 months match or exceed most x86 systems in performance, power per watt and heat, especially in form factors with fewer physical, power and thermal constraints. They wouldn't be doing this if they didn't feel confident about their roadmap. They've probably been prototyping this for 3-5 years already, and likely accelerated development ever since Intel started stumbling, failing year after year to make the transition to 10nm.

[rezoneight]

The significant difference between PowerPC-Intel and Intel-ARM transition is ARM has never been able to out-perform latest generation i9, Xeon, Ryzen, Threadripper in general purpose computing.

Apple doesn't use ARM reference designs, so past examples may not be a very good reference for comparison; They only license the instruction set. They've been designing their own microarchitecture since the iPad/iPhone 4 onward and for the past couple of years their own GPUs as well. I won't be surprised at all if Apple silicon Macs over the next 24 months match or exceed most x86 systems in performance, power per watt and heat, especially in form factors with fewer physical, power and thermal constraints. They wouldn't be doing this if they didn't feel confident about their roadmap. They've probably been prototyping this for 3-5 years already, and likely accelerated development ever since Intel started stumbling, failing year after year to make the transition to 10nm.

How in the hell can someone downvote this? Totally spot on.

[tooneba]

The significant difference between PowerPC-Intel and Intel-ARM transition is ARM has never been able to out-perform latest generation i9, Xeon, Ryzen, Threadripper in general purpose computing.

Apple doesn't use ARM reference designs, so past examples may not be a very good reference for comparison; They only license the instruction set. They've been designing their own microarchitecture since the iPad/iPhone 4 onward and for the past couple of years their own GPUs as well. I won't be surprised at all if Apple silicon Macs over the next 24 months match or exceed most x86 systems in performance, power per watt and heat, especially in form factors with fewer physical, power and thermal constraints. They wouldn't be doing this if they didn't feel confident about their roadmap. They've probably been prototyping this for 3-5 years already, and likely accelerated development ever since Intel started stumbling, failing year after year to make the transition to 10nm.

Are there any example of non-reference model outperformed the reference model twofold in the history of processor unit? It basically needs doubled power to compete with intel/AMD's multi-purpose CPU in terms of replacing whole product line's CPU. I'm just wondering if you are talking about it based on past examples or pure speculation.

iMac (27-inch Retina Early 2019)
Intel Core i9-9900K @ 3.6 GHz (8 cores)
8285

[Burillo]

i don't think performance is what they're going for. their goal is control. performance has always been secondary on macs, which is perfectly illustrated by the underpowered mac pro's (underpowered for the amount of money you pay for them).

[Michael L]

their goal is control.

Very important. Apple have a different priority than many companies. From the beginning, their goal was to make you want a new product, rather than give you what you need. To regularly ship those products, they need control of their supply chain.

[JCJR]

Re arm performance:
Disclaimers
_1_ Personally I boycott apple and do not care what they do.
_2_ The following link re arm supercomputers not real relevant. Maybe compared to apple arm, comparing Nukes with Firecrackers.
https://www.pcmag.com/news/japans-arm-b ... ercomputer

The current world's fastest supercomputer is arm based, which might imply performance potential for the architecture.

OTOH as someone said, even if you stack up enough PIC chips then eventually it will run purt fast. Ye Olde Infinite Monkeys Design Pattern.

[Tronam]

i don't think performance is what they're going for. their goal is control. performance has always been secondary on macs, which is perfectly illustrated by the underpowered mac pro's (underpowered for the amount of money you pay for them).

I don’t follow this logic. In nearly every device category where they’re using their own custom A-series processors, they have consistently led the industry in both performance and power efficiency. To assume that these upcoming desktop CPUs won’t emphasize performance is inconsistent with how their silicon team has operated over the past decade. They introduced the industry’s first 64-bit smartphone processor in the iPhone 5S and have aggressively released increasingly faster SoCs every single year like clockwork, always ahead of the competition and often dramatically so. Being free of Intel’s rocky and inconsistent roadmap will allow them to ramp up Mac processors at their own pace and incorporate any custom architecture blocks they choose. Just look at how powerful the iPad Pro is at H.265 hardware encoding. This will also allow Apple to sidestep all the security vulnerabilities and zero-day exploits that have been plaguing Intel over the past 5 years.

[paulm12]

The significant difference between PowerPC-Intel and Intel-ARM transition is ARM has never been able to out-perform latest generation i9, Xeon, Ryzen, Threadripper in general purpose computing.

Apple doesn't use ARM reference designs, so past examples may not be a very good reference for comparison; They only license the instruction set. They've been designing their own microarchitecture since the iPad/iPhone 4 onward and for the past couple of years their own GPUs as well. I won't be surprised at all if Apple silicon Macs over the next 24 months match or exceed most x86 systems in performance, power per watt and heat, especially in form factors with fewer physical, power and thermal constraints. They wouldn't be doing this if they didn't feel confident about their roadmap. They've probably been prototyping this for 3-5 years already, and likely accelerated development ever since Intel started stumbling, failing year after year to make the transition to 10nm.

I'm not so sure about that. Apple is still a relatively new player to the silicon game, and making CPUs for computers that actually have a full filesystem and need to support applications (including legacy ones) instead of the stripped-down OS that iPhones/iPads run on (and I am a huge iPhone/iPad fan) is a huge undertaking. I do think this is something they've been working on for over 10 years actually. Of course, Intel has had some problems with their chips, mainly security things, and they've been doing hardware for a long time.
But Apple has also had problems with their hardware as of late too. The T2 was a complete disaster for audio professionals.
Of course, the main reason Apple is doing this is simply for the $$$. If Apple made their own x86 chips, they would have to pay Intel for licensing fees since Intel owns x86. The transition would be easier, and Apple has a lot more desktop/laptop experience with x86. Apple will no longer be tied to Intel and Intel fabs, and will instead likely use TSMC.

It makes sense from a high level to go for a "unified" architecture using ARM. And it would make the most sense if Apple's first ARM computer would be a cheaper netbook/chromebook replacement that can run iOS apps and also desktop apps. But for high-performance computing, the advantages of ARM are less pronounced. For iOS and mobile devices, having a RISC architecture means less power consumption, which really matters. But for high performance tasks, CISCy-RISC like x86 (in my experience) usually wins out (if you don't care about power consumption).

[Passing Bye]

I'm interested in new non Intel Apple laptop and really happy to see that Logic is already supported and U He is pretty much on the ball from the get go, sincerely couldn't care less if everybody else rejected whole transition.

[IncarnateX]

To performance: I bought a Macbook Air late 2017, 1,8 ghz (Intel i5), 8 gb ram and Catalina. While it performed in Logic Pro X as I would have expected any computer to with such stats, that is, not too many VSTis in play before it starts hicking, I was rather surprised that it runs 42 + modules, including several drummachines, samplers and synths, in Reason 11. It is about the same as my pc with better processer (i7) and more speed (2,7) does. I have not taken any of them to their limits yet, so it is not really a comparison. Some magic must have happened in Reason since last time I was on it, namely with the release of 6.5 where REs were more cpu heavy to my knowledge. Anyway, I do not think I will be missing something with such a Reason performance. Would never have expected this by default.

[tooneba]

_2_ The following link re arm supercomputers not real relevant. Maybe compared to apple arm, comparing Nukes with Firecrackers.
https://www.pcmag.com/news/japans-arm-b ... ercomputer

The current world's fastest supercomputer is arm based, which might imply performance potential for the architecture.

Interesting. Thx for the link.