Intel Says 10nm Chip Development is On Track

[Tech198]

I think Apple and Microsoft will work together to develop a solution for this (Windows on ARM). Those two companies realize that they gain a lot more from a partnership today.

Ya, they probably will, but i'm just coming from iOS and ARM...

The fact Apple and MS would pull a larger audience on Mac is probably more of a win, because of Office on the Mac as well, but remember Apple is still it's own ecosystem... it could flip the other way if Apple decided to.

It's always been this way on iOS... Although some of the same software are available, its always been somewhat limited..

I don't think the only reason is because Apple decided...it is probably in part due to hardware as well. That's why we went from PowerPC to Intel..

 

[Unregistered 4U]

I think Apple and Microsoft will work together to develop a solution for this (Windows on ARM).

I don’t think so. With Intel’s chips, you not only choose the CPU but the motherboard as well. And, those motherboard components have to adhere to the x86 architecture. Because the main differences between a macOS motherboard and a Windows motherboard could be resolved at a software level (drivers and such) it meant that the effort to provide BootCamp was probably just slightly more than trivial.

The differences between ARM/Windows and ARM/macOS could likely be much greater such that it’s not feasible to create a system that can execute macOS and then in an ARM boot camp, support Windows (especially since Apple’s custom chip doesn’t support 32-bit instructions at all and Windows ARM does).

they'll probably start out on MacBooks only

I really think they’re going to start at the high end. Because, I think their selling point to professionals will be, “Do you want a system that can run Final Cut Pro, Logic, Motion and our other apps the fastest, thus saving you time and money? Then, look at the performance of this A-processor macOS system versus this Intel macOS system.” I’m looking forward to a future “bake-off” presentation where Apple chips boot faster, restart faster, run cooler and provides greater performance at macOS tasks.

 

[DNichter]

I don’t think so. With Intel’s chips, you not only choose the CPU but the motherboard as well. And, those motherboard components have to adhere to the x86 architecture. Because the main differences between a macOS motherboard and a Windows motherboard could be resolved at a software level (drivers and such) it meant that the effort to provide BootCamp was probably just slightly more than trivial.

The differences between ARM/Windows and ARM/macOS could likely be much greater such that it’s not feasible to create a system that can execute macOS and then in an ARM boot camp, support Windows (especially since Apple’s custom chip doesn’t support 32-bit instructions at all and Windows ARM does).

I really think they’re going to start at the high end. Because, I think their selling point to professionals will be, “Do you want a system that can run Final Cut Pro, Logic, Motion and our other apps the fastest, thus saving you time and money? Then, look at the performance of this A-processor macOS system versus this Intel macOS system.” I’m looking forward to a future “bake-off” presentation where Apple chips boot faster, restart faster, run cooler and provides greater performance at macOS tasks.

You’re not wrong, but if both companies want to move forward with their OS on ARM, they of course could work to do that. How easy it is to do, I have no idea, but there are smarter people than me to figure that out.

 

[GeneralChang]

I really think they’re going to start at the high end. Because, I think their selling point to professionals will be, “Do you want a system that can run Final Cut Pro, Logic, Motion and our other apps the fastest, thus saving you time and money? Then, look at the performance of this A-processor macOS system versus this Intel macOS system.” I’m looking forward to a future “bake-off” presentation where Apple chips boot faster, restart faster, run cooler and provides greater performance at macOS tasks.

I think there's two main reasons why this probably won't be the case. First, professional users who favor high end Apple hardware will be the hardest hit by this transition and any resulting bugs that come with it. Even if your entire workflow is Apple-centric (Logic, Final Cut, etc), even Apple is going to want to try testing this change at scale with stuff like iMovie and Garageband before they're ready to push their professional tools into this space. Pushing it to the average consumer first allows them to fine-tune their OS and processes for the new architecture without breaking the high-profile workflows of, say, people working in Hollywood.

The second reason is simply because while A-series chips now rival Intel processors in raw power, that's only when talking about Intel's lower power chips. An A-series chip that beats out a Core-M CPU is basically already shipping. An A-series chip that can throw down against a Xeon W... there's gonna be a few more iterations before we get to that point, I'm pretty sure.

 

[Unregistered 4U]

high-profile workflows of, say, people working in Hollywood.

This is where I think the professionals are a tiny, captive market that, if they’re willing to buy in to Apple’s vision, then they’d be willing to run the new system side by side with the old one while the kinks are worked out. Even when replacing old with new of the same general hardware, no real professional goes all-in on day one. They bring their workflow to the new system and over time they figure out what’s broken. This way, they can report the bug and always finish their work on their tried and true system. If those users can push the system hard and get to the point where there’s no blocking issues, then rolling out to the general public would be child’s play. One main reason why I think this is because it would be a VERY valid reason as to why the new MacPro has taken SO long to be ready for the market.

And, synthetic benchmarks only tell part of the story. Apple knows what every class, every library does, knows how it should compile and has a good idea of how fast it should execute as they’re evaluating any new chip architecture changes. The raw speed numbers could very well be worse than a Xeon. But, it could still run macOS and macOS apps faster than any Intel processor. It’s a CPU custom designed to run the software and not a generic part like Intel’s. Additionally, given that Apple has been quite adept at beating any and all comers on a performance per watt basis, I don’t think it’s too much of a stretch to think that they’d be able to scale that success.

 

[GeneralChang]

This is where I think the professionals are a tiny, captive market that, if they’re willing to buy in to Apple’s vision, then they’d be willing to run the new system side by side with the old one while the kinks are worked out. Even when replacing old with new of the same general hardware, no real professional goes all-in on day one. They bring their workflow to the new system and over time they figure out what’s broken. This way, they can report the bug and always finish their work on their tried and true system. If those users can push the system hard and get to the point where there’s no blocking issues, then rolling out to the general public would be child’s play. One main reason why I think this is because it would be a VERY valid reason as to why the new MacPro has taken SO long to be ready for the market.

And, synthetic benchmarks only tell part of the story. Apple knows what every class, every library does, knows how it should compile and has a good idea of how fast it should execute as they’re evaluating any new chip architecture changes. The raw speed numbers could very well be worse than a Xeon. But, it could still run macOS and macOS apps faster than any Intel processor. It’s a CPU custom designed to run the software and not a generic part like Intel’s. Additionally, given that Apple has been quite adept at beating any and all comers on a performance per watt basis, I don’t think it’s too much of a stretch to think that they’d be able to scale that success.

The mass exodus away from Final Cut back when Final Cut Pro X was first released would seem to disagree with your assessment. And I'm pretty sure the reason the new Mac Pro has taken so long is because Apple designed themselves into a very uncomfortable corner with the trash can and it took them a few years to figure out that the market was not going to follow them this time.

I'm not saying Apple can't produce a chip based on ARM that'll beat out Xeons, I'm just saying Apple isn't going to jump directly from a less-than-5 watt chip to a 100-or-more watt chip without tackling a few tiers in the middle first. That'd be like Estes bidding on the solid-fuel boosters for the SLS based on having built model rocket engines for all those years. I'm sure internally Apple is testing a wide range of chips of various power, possibly even some that'll rival Xeons and Threadrippers. But I suspect their first version in a Mac will be low-powered, because that's the version they're already confident in.

 

[Unregistered 4U]

The mass exodus away from Final Cut back when Final Cut Pro X was first released would seem to disagree with your assessment.

There was indeed a mass exodus, but, some of those folks came back or they grabbed a ton of new users with lowered price. FCP7 was at 2 million users when FCPX was released and there’s now 2.5+ million users of FCPX now. And, as others have said, a quick way to get themselves out of that corner was to make the “big ol’ Intel box” users have been craving for years. If they’re building another constrained tiny Intel box (which also would take time to develop), they may miss the mark with a lot of folks again.

If they’ve been testing high wattage chips for years, they’re already aware, probably for some years now, that they’ve got the hardware question answered... they just need to bring the development software up to speed... Hence the recent reports of Marzipan “Universal” app writing. That is set for 2021, though.

So, If Apple delays the MacPro to 2021 to “really deliver the modular performance our customer need” then I’d lean even further into a Pro-first direction.

 

[GeneralChang]

There was indeed a mass exodus, but, some of those folks came back or they grabbed a ton of new users with lowered price. FCP7 was at 2 million users when FCPX was released and there’s now 2.5+ million users of FCPX now. And, as others have said, a quick way to get themselves out of that corner was to make the “big ol’ Intel box” users have been craving for years. If they’re building another constrained tiny Intel box (which also would take time to develop), they may miss the mark with a lot of folks again.

If they’ve been testing high wattage chips for years, they’re already aware, probably for some years now, that they’ve got the hardware question answered... they just need to bring the development software up to speed... Hence the recent reports of Marzipan “Universal” app writing. That is set for 2021, though.

So, If Apple delays the MacPro to 2021 to “really deliver the modular performance our customer need” then I’d lean even further into a Pro-first direction.

The universal app rumors actually support my argument nicely, considering the people that will benefit most from that are the folks developing for iOS, and they're hardly targeting the same market as, say, the iMac Pro. And it takes a couple of years for OEMs to develop and test a new large tower chassis, even if it is just a "big ol' Intel box". I'd be willing to bet Apple was trying to save the trash can design right up till Phil apologized for the thing in 2017. They're probably just about finished with their cheese-grater redesign.

I'm gonna just go with Apple releasing a Xeon or HEDP powered Mac Pro this year, with the first ARM powered Macs being the MacBook and MacBook Air in 2020. Everything I know about silicon development, software development, and Apple's own product strategy & history says we'll either see the entire line go at the same time (nothing particularly "pro first" about that), or they'll start pushing the new paradigm from the bottom up.

 

[cmaier]

The universal app rumors actually support my argument nicely, considering the people that will benefit most from that are the folks developing for iOS, and they're hardly targeting the same market as, say, the iMac Pro. And it takes a couple of years for OEMs to develop and test a new large tower chassis, even if it is just a "big ol' Intel box". I'd be willing to bet Apple was trying to save the trash can design right up till Phil apologized for the thing in 2017. They're probably just about finished with their cheese-grater redesign.

I'm gonna just go with Apple releasing a Xeon or HEDP powered Mac Pro this year, with the first ARM powered Macs being the MacBook and MacBook Air in 2020. Everything I know about silicon development, software development, and Apple's own product strategy & history says we'll either see the entire line go at the same time (nothing particularly "pro first" about that), or they'll start pushing the new paradigm from the bottom up.

I agree. But it will be embarrassing when the low end arm machines run faster than the pro machines.

 

[GeneralChang]

I agree. But it will be embarrassing when the low end arm machines run faster than the pro machines.

For certain definitions of "faster". Watt-for-watt Apple's chips are already poised to destroy x86 chips, and I even suspect they've got a real shot at taking the single-core performance crown away from Intel's i7's finally. Massively parallel high-powered multi-threaded workloads, though... that's where I'm less confident.

 

[cmaier]

For certain definitions of "faster". Watt-for-watt Apple's chips are already poised to destroy x86 chips, and I even suspect they've got a real shot at taking the single-core performance crown away from Intel's i7's finally. Massively parallel high-powered multi-threaded workloads, though... that's where I'm less confident.

as a cpu designer I don’t see any ceiling that prevents arm from handling such workloads. Particularly since apple can modify arm anyway it wishes, add new instructions, new coprocessors, etc. It owns the whole stack - language, compiler, OS, distribution/store.

 

[GeneralChang]

as a cpu designer I don’t see any ceiling that prevents arm from handling such workloads. Particularly since apple can modify arm anyway it wishes, add new instructions, new coprocessors, etc. It owns the whole stack - language, compiler, OS, distribution/store.

Absolutely, the only "ceiling" they've got is their available development bandwidth and roadmap. As I've previously said, I don't think they can't, I just don't think that's where they'll want to start.

 

[cmaier]

Absolutely, the only "ceiling" they've got is their available development bandwidth and roadmap. As I've previously said, I don't think they can't, I just don't think that's where they'll want to start.

Chip designers like to start at the high end. Easier to scale down than up

 

[k27]

And I'm pretty sure the reason the new Mac Pro has taken so long is because Apple designed themselves into a very uncomfortable corner with the trash can and it took them a few years to figure out that the market was not going to follow them this time.

This is Apple PR. The trash can was limited from the beginning. There is simply more space in classic enclosures (also for cooling and graphics cards).

And it takes a couple of years for OEMs to develop and test a new large tower chassis, even if it is just a "big ol' Intel box".

Really? I don't think so... you don't need years for that.
Apple wants to reinvent the wheel. Or it is ARM...

 

[GeneralChang]

This is Apple PR. The trash can was limited from the beginning. There is simply more space in classic enclosures (also for cooling and graphics cards).


Really? I don't think so... you don't need years for that.
Apple wants to reinvent the wheel. Or it is ARM...

I mean, this is Apple. Reinventing the wheel does not mean it won't be a big Intel box. It just won't be as straightforward as stuffing the cheese-grater with new parts. The pessimist in me says that would be admitting defeat. The PC builder in me who was just working inside that case a week ago would say it's actually kind of painfully designed and needs to be reimagined.
[doublepost=1550786410][/doublepost]

Chip designers like to start at the high end. Easier to scale down than up

Except isn't that kinda the opposite approach to ARM's entire development path thus far?

 

[k27]

It just won't be as straightforward as stuffing the cheese-grater with new parts.

That's exactly what Apple should have done. Then users could have bought a classic Mac Pro with the latest hardware. In the meantime Apple could have reinvented the wheel. But as it is now, users have to wait or buy the outdated and limited trash can. Or they leave the Mac. OK, the iMac Pro is still there.
Nobody asked for the trash can. And nobody asked for the new keyboards... that's the real problem.
The problem is not the CPU architecture but Apple with stupid and unnecessary hardware design decisions. Pricing policy. Quality problems, missing updates and so on.

 

[GeneralChang]

That's exactly what Apple should have done. Then users could have bought a classic Mac Pro with the latest hardware. In the meantime Apple could have reinvented the wheel. But as it is now, users have to wait or buy the outdated and limited trash can. Or they leave the Mac. OK, the iMac Pro is still there.
Nobody asked for the trash can. And nobody asked for the new keyboards... that's the real problem.
The problem is not the CPU architecture but Apple with stupid and unnecessary hardware design decisions. Pricing policy. Quality problems, missing updates and so on.

I hate to be the odd one out, but the third-gen butterfly keyboards, with the silicone membrane under the caps? I did actually ask for that.

 

[cmaier]

I hate to be the odd one out, but the third-gen butterfly keyboards, with the silicone membrane under the caps? I did actually ask for that.

Next time can you ask for completely new keyboards? Because the membrane didn’t solve the problem.

 

[GeneralChang]

Next time can you ask for completely new keyboards? Because the membrane didn’t solve the problem.

You probably don't want me asking. My ideal keyboard is a textured sheet of glass with a Taptic Engine under every keycap.