Intel Expecting Apple to Transition to Custom ARM-Based Chips Starting in 2020

[techwhiz]

It could be that the main problem from Apple's point of view is Intel's spotty upgrades. Intel still hasn't worked the kinks out of their 10 nm processor. We're back to the dilemma in the old G5 days: IBM couldn't (or wouldn't) upgrade fast enough to suit Apple's needs. By doing it all in-house, Apple controls the timeline.

I'm wondering if Apple will retain Intel for their "pro" line of machines, to ensure maximum compatibility. The bigger-selling consumer models like the MacBook line could transition to ARM, but the more specialized Mac Pro might not. Then again, Apple might need a separate OS for each type. Who knows? If this is true, though, I just hope they get it right.

Apple will be at the mercy of TSMC and their fab cycles.
So if TSMC misses or delays a process node they are no better than Intel.
If TSMC has fab constraints then Apple has delivery constraints.

Intel owns their own FABs, Apple does not. That immediately puts them at a disadvantage when trying to compete in a different arena than an embedded mobile CPU.

Building a multithreaded cache coherent processor is a lot different than building a small embedded CPU with limited I/O capability.

Just for starters they need a cache coherent interconnect for their new CPU complex, CHI won't be enough.
The cores will need to be much higher performance than what ARM has now.

I can list the number of companies that have successfully went head to head with Intel -> AMD.
I can list the others that missed: Sun SPARC, HP PA-RISC, SGI MIPS, AIM (Apple, IBM, Motorola) PPC, Motorola 68K and 88K, Qualcomm ARM server, Broadcom ARM server, Samsung ARM server, AMD (they sold their FAB and went dormant for years until they defined a new architecture that could scale) and belong in both categories.

 

[cppguy]

I wonder if they're also going to transition the Pro line, or just the mobile devices. If a Mac Pro is coming around September, and a few months later they're switching everyone to ARM, that's bad timing. Or the Mac Pro may be the first computer with an array of ARM processors.

I don't think it's all bad news, other than for Windows in Bootcamp/VMware. On the other hand, Microsoft might be pushing ARM as well. Imagine how much less heat an ARM Surface Pro would generate. Most people wouldn't complain about the lack of fans and the multi-day battery life.

I'm not worried about professional software. Adobe and the big brands will follow Apple very quickly. Lightroom CC will eventually replace the Lightroom Classic anyway, Photoshop is coming to ARM regardless, and Premier already has an iPad version. They may be crippled initially, but gradually they're catching up.

I only feel sorry for travelers who need both Windows and Mac, and they'll be forced to take two ultrabooks now. Or connect remotely via VPN. Businesses use Windows, because they can run everything from the 1990s and beyond. So Windows will have to be supported on Intel for a long time.

Apple, on the other hand, doesn't care so much what they're breaking. Remember cool 32-bit games like Dark Nebula? You cannot even play it on any modern device anymore. Apple won't feel sorry for "lazy" developers who won't recompile for ARM.

 

[cmaier]

Apple will be at the mercy of TSMC and their fab cycles.
So if TSMC misses or delays a process node they are no better than Intel.
If TSMC has fab constraints then Apple has delivery constraints.

Intel owns their own FABs, Apple does not. That immediately puts them at a disadvantage when trying to compete in a different arena than an embedded mobile CPU.

Building a multithreaded cache coherent processor is a lot different than building a small embedded CPU with limited I/O capability.

Just for starters they need a cache coherent interconnect for their new CPU complex, CHI won't be enough.
The cores will need to be much higher performance than what ARM has now.

I can list the number of companies that have successfully went head to head with Intel -> AMD.
I can list the others that missed: Sun SPARC, HP PA-RISC, SGI MIPS, AIM (Apple, IBM, Motorola) PPC, Motorola 68K and 88K, Qualcomm ARM server, Broadcom ARM server, Samsung ARM server, AMD (they sold their FAB and went dormant for years until they defined a new architecture that could scale) and belong in both categories.

My PhD dissertation was on CPU caches. I was a designer on the original Opteron, UltraSparc V, Exponential x704, etc. Designing A12 is not that different from designing a multi core x86.

 

[j569mxr]

I'd wager that there won't be a full transition at first. Perhaps, migrating the non-Pro models to these Apple based chips will be the plan. Pro models will probably still utilize Intel chips.

This ain’t Steve Jobs’ Apple anymore. It’ll probably happen all at once. Rosetta 2.0 included for emulating x64 and no more VM or BootCamp. Not looking forward to the MacPadPhone...

 

[Inutopia]

Well if this is true then I dearly hope Apple can pull it off.

I am old and ugly enough to remember the transition to Intel from PPC; now that was a great move. No question.
But, Apple's base was far smaller… and even then it took a lot of heartbreak.

Not really looking forward to the "new" Rosetta as companies suck teeth while deciding if they are porting it all over.

Of course there was BootCamp — which was great — but this time round?

Anyway… deep breaths and let's see if Apple come up with some magic.

That only happened 12 years ago.

 

[Reindeer_Games]

My PhD dissertation was on CPU caches. I was a designer on the original Opteron, UltraSparc V, Exponential x704, etc. Designing A12 is not that different from designing a multi core x86.

But you would agree that going from one dedicated chipmaker to another does not guarantee freedom of production bottlenecking, right?

 

[Juicy Box]

but a win for customers in the long run.

I think it is way too early to be making this statement.

 

[cmaier]

But you would agree that going from one dedicated chipmaker to another does not guarantee production bottlenecking, right?

Sure. One advantage of TSMC over Intel, though, is that Intel only (more or less) services Intel. TSMC has to answer to many customers. At AMD we had fabs. That didn’t do us much good. Apple can also switch fabs to Global Foundries or Samsung or the like if it turns out TSMC hits a roadblock. There are ways to design cell libraries to make porting fairly easy (requires a lot of work up front, to be sure, but once you do it you are good). Even at AMD we designed a number of parts to be used on multiple fab lines or multiple fabs, just in case. Quite doable.

 

[GoodWheaties]

Microsoft also now have windows for ARM with a baked in x86 emulator (works pretty well from what I have seen)

What??!!! We clearly haven’t seen the same benchmarks...

 

[Reindeer_Games]

Sure. One advantage of TSMC over Intel, though, is that Intel only (more or less) services Intel. TSMC has to answer to many customers. At AMD we had fabs. That didn’t do us much good. Apple can also switch fabs to Global Foundries or Samsung or the like if it turns out TSMC hits a roadblock. There are ways to design cell libraries to make porting fairly easy (requires a lot of work up front, to be sure, but once you do it you are good). Even at AMD we designed a number of parts to be used on multiple fab lines or multiple fabs, just in case. Quite doable.

Fair enough-but ARM architecture isn't as exploitable, and why many multi-boot individuals much prefer Intel's chipsets. I'm not saying the technology is a dead-end, just for a lot of people that are upset-it looks like it may be. We are upset we have to leave Apple hardware-that should say something.

 

[cppguy]

This ain’t Steve Jobs’ Apple anymore. It’ll probably happen all at once. Rosetta 2.0 included for emulating x64 and no more VM or BootCamp. Not looking forward to the MacPadPhone...

If the Mac Pro is Intel-based, they're not going to stop supporting that a few months later. On the other hand, if the Mac Pro is ARM based, no question that the days of Intel macOS are numbered.

 

[just21]

Is the ARM roadmap the Mac roadmap?

https://www.arm.com/company/news/2018/08/accelerating-mobile-and-laptop-performance

• Client CPU’s expected to deliver year-over-year performance improvements of >15% for compute through 2020
• Arm positioned for laptop share gain with roadmap designed for 5G always-on, always-connected devices

• The follow-up to Cortex-A76 will be codenamed ‘Deimos’ and delivered to our partners in 2018. Optimized for the latest 7nm nodes, ‘Deimos’ is based on Arm DynamIQ technology and is expected to deliver a 15+ percent increase in compute performance.

• In 2019, the CPU codenamed ‘Hercules’ will be available to Arm partners. ‘Hercules’, also based on DynamIQ technology, will be optimized for both the latest 5nm and 7nm nodes. ‘Hercules’ continues the trajectory of increased compute performance, while also improving power and area efficiency by 10 percent (in addition to the efficiency gains achievable from the 5nm process node).

 

[gsurf123]

Well-I guess that settles it. Apple was nice for decade and a half, but an ARM will never be an i7.

That's a bold call. I've gone from Motorola to Intel so moving forward won't hurt. Intel has watched as they have been passed up. They have no one to blame but themselves. I don't care who makes the chip so long as it works as is as fast and efficient as possible. Intel has let everyone down multiple times in the last decade with shipping delays, marginal performance increases and heat issues.

 

[Michael Scrip]

“and they have ramped up given the many Intel chip delays that have resulted in subsequent delays for Mac products. With its own ARM-based chips, Apple will not be tied to Intel's chip release cycles. ”

I dont understand why you keep repeating this, which is false! Apple updates their macs when they want, not when Intel releases cycles!! Hahahahha! Maaan, where are the iMacs with the 5ghz i9??
Maybe i9 are to expensive or too hot for iMacs, but this is not the only example, as rarely Apple updates right after Intels releases, they hold and hold and (mac mini, mac pro ahem) hold

Exactly.

It's funny how Dell, HP, Lenovo, Acer, Asus and all other Intel-using PC manufacturers make frequent updates to their machines...

...yet Apple sometimes goes YEARS without updates.

I know Intel has had their share of problems. I'm not gonna deny that.

But to specifically call-out Intel as the reason for Apple's delays? Hogwash.

Intel themselves launched 50 different NUC models in the time between the 2014 and the 2018 Mac Mini.

Think about that for a moment. Intel... the chipmaker... makes entire PCs.

But we're gonna blame Intel because Apple can't update their machines? Sure.

 

[cmaier]

Fair enough-but ARM architecture isn't as exploitable, and why many multi-boot individuals much prefer Intel's chipsets. I'm not saying the technology is a dead-end, just for a lot of people that are upset-it looks like it may be. We are upset we have to leave Apple hardware-that should say something.

Also remember that “apple arm” isn’t necessarily arm. They have an architecture license but they also control the programming language, the compiler, the App Store, the OS, etc. They can add instructions, add coprocessors, etc. They can even remove stuff they don’t want to use. Gives them a ton of power to make a chip that sings with their ecosystem, as opposed to Intel x86 which has to run windows, a million flavors of Linux, ancient OS’s that hardly anybody cares about anymore, old versions of windows, etc.

 

[cppguy]

What??!!! We clearly haven’t seen the same benchmarks...

Windows is actually two platforms in one. There are legacy x86 (C++) apps, and then there are .NET (Universal) apps, which are machine-independent. .NET code is JITed much like Java, which means it doesn't need emulation. Of course native x86 code is terribly slow on ARM when emulated.

 

[cmaier]

Is the ARM roadmap the Mac roadmap?

https://www.arm.com/company/news/2018/08/accelerating-mobile-and-laptop-performance

• Client CPU’s expected to deliver year-over-year performance improvements of >15% for compute through 2020
• Arm positioned for laptop share gain with roadmap designed for 5G always-on, always-connected devices

• The follow-up to Cortex-A76 will be codenamed ‘Deimos’ and delivered to our partners in 2018. Optimized for the latest 7nm nodes, ‘Deimos’ is based on Arm DynamIQ technology and is expected to deliver a 15+ percent increase in compute performance.

• In 2019, the CPU codenamed ‘Hercules’ will be available to Arm partners. ‘Hercules’, also based on DynamIQ technology, will be optimized for both the latest 5nm and 7nm nodes. ‘Hercules’ continues the trajectory of increased compute performance, while also improving power and area efficiency by 10 percent (in addition to the efficiency gains achievable from the 5nm process node).

No. Apple doesnt need to rely on anything they get from ARM. They already do their own micro architecture.

 

[gsurf123]

My only concern with this is will this change still allow us to run Boot Camp and Parallels Desktop on ARM based Macs?

I don't know about boot camp, but there were X86 emulators long before Intel. They were not great, but they did work. If the ARM chip is a screamer then there won't be anything to worry about. Not too mention there will be plenty of Intel Macs in the wild through at least 2025. These machines have a long lifespan.

 

[cube]

Windows is actually two platforms in one. There are legacy x86 (C++) apps, and then there are .NET (Universal) apps, which are machine-independent. .NET code is JITed much like Java, which means it doesn't need emulation. Of course native x86 code is terribly slow on ARM when emulated.

You can compile C++/CLI programs as native, mixed, or pure .NET .

 

[TSE]

Read a bit further, buddy.

I fail to see the post you are talking about. All I know is defending Intel is asinine. They have rested on their laurels and will pay the price! They have basically been on Skylake since 2015 and have utterly failed on the manufacturing side of things as well, which was at one time their bread and butter.

I welcome Apple-powered Macs with open arms (no pun intended). Hopefully they keep up the significant annual gains in terms of performance-per-watt as they have for the past 10 years.

 

[zoomp]

Maybe the modular Mac Pro could run both processors? Maybe one can switch its brain since it's modular? That would make sense for iOS developers right now.

 

[cycomiko]

Can't afford a Mac but on a Mac-centric forum. You don't feel dirty or ashamed? I often stand outside of nice steakhouses myself sniffing that beautiful aroma because I won't go in as it is overpriced.

Or, in an less obnoxious version, doesnt think the mac is worth that much, so will not pay apple that much, regardless of their own personal version of affordability.

 

[Reindeer_Games]

That's a bold call. I've gone from Motorola to Intel so moving forward won't hurt. Intel has watched as they have been passed up. They have no one to blame but themselves. I don't care who makes the chip so long as it works as is as fast and efficient as possible. Intel has let everyone down multiple times in the last decade with shipping delays, marginal performance increases and heat issues.

I agree, but I'm speaking more to my application of dual boot; its looking dark. It's why I've spent the money I have with them.

Also remember that “apple arm” isn’t necessarily arm.

Yes, but multi-boot what sold me and is allowed me in the eco-system.

I fail to see the post you are talking about. All I know is defending Intel is asinine.

^See it now? I'm a gamer.

 

[falainber]

This sounds like a dead end for Appe to me. They are trying to develop everything in house. As big as they are they are just a little fraction of the world capital and brain power. They may succeed for a while by being able to better optimize the whole package but in a long run they will lose. The time of computer giants doing everything in house (IBM, DEC, HP etc.) has long past. There was a technological and economical reason for that. Apple is not going to reverse the course of history.

 

[cmaier]

I agree, but I'm speaking more to my application of dual boot; its looking dark. It's why I've spent the money I have with them.



Yes, but multi-boot what sold me and is allowed me in the eco-system.



^See it now?

Yep. Sold me too. Back in the leopard days i was impressed by MacOS’s Unix underpinnings (having use Solaris and other Unix OSs for work), but i was afraid of losing windows until boot camp came out. So with leopard i finally bought a mac. It was my comfort blanket, knowing i could run windows. After about 4 months i never booted windows again. I suspect many people who were initially sold based on this functionality are now not using it. And many more people find MacOS familiar and non-scary because it is iOS-like in many ways.

The number of people who need multi-boot is tiny. A fraction of the 5million or so who buy macs in A quarter. Compared with the 100 million iPhone purchasers who apple hopes to mine as future mac purchasers. They just don’t care about you.