Intel is becoming an anchor for Apple. I know Apple producing an ARM-based iMac most likely will not happen, but it would be a way for Apple to begin unraveling the binding that Intel has on Apple's innovation.
Give us a break. Your ilk said the same thing about the IBM/Motorola Power PC and the 6502 before that (why oh why did Woz choose that CPU? Answer: It was cheap). Somebody is always a boat anchor for Apple and if Apple would just follow your advice all would be well. I think I’ll put my faith in Apple engineers who actually know what’s going on in the CPU world.
Well I know processor folks at both Intel and Apple and indeed Intel is an anchor for them... the OP said nothing about "all would be well", but for sure the Apple processor folks have a lot more runway to build a better platform.
Guess saying "I think I'll put my faith..." is a good way for anyone without a clue to start a sentence
"but for sure the Apple processor folks have a lot more runway to build a better platform"
Honest question: What do you mean by that?
I'm not a hardware guy, but I'd like to know if Apple's ARM AxX platform (possibly IBMs Power9) or other RISC solution could be used to remove the dependency on Intel.
To my mind, an Apple [non-Intel] platform would have to provide good enough x86 emulation to acceptably run Windows VMs. But the real advantages would be running apps written specifically for the Apple hardware and OS. (I'm sure OSS Swift is somewhere in the equation)
As I understand it, recent Intel chips convert the x86 CISC instructions to RISC for execution.
But, likely, there is a lot more than the CPU... Memory bandwidth, multiple threads, multiple cores, I/O capability, GPUs, and a lot of other things I don't understand all I know about!
Maybe a better way to ask the question: If you are Apple, given a clean slate, what kind of chip would you have to build to remove your dependency on Intel?
Intel has been dropping the ball for 3(?) years now. Shouldn't that have been enough time for AMD to catch up again? Does anybody know if AMD offers anything close to/better than Intel?
The PC market is rapidly changing, in part because Intel has slowed down. I think the larger factor is just that most people don't really need a faster computer anymore, so they keep their existing computers for a lot longer. I used to get a new computer every year, but my current computer is four years old and still doesn't feel slow (added a SSD).
Lots of nerds complain about this, but Apple has exactly the right idea. They're slowing down their release cycles because the vast majority of people are slowing down their purchases, and the current crop of computers is fast enough. Personally, at the current speeds, Skylake (just bought a 15" MBP) is plenty fast for everything I do, and I'd rather have either lighter weight or longer battery life. I probably won't buy my next computer for four or five more years, at which point Apple will probably have something new and cool for me to buy.
The only real issue I see is the Mac Pro, which truly is behind the times in every way. I suspect Apple will be completely changing the design as the current one isn't much of a Pro computer anyway.
I don't know how credible they are but Macworld UK says a Mac Pro update is coming end of November (per their sources)
Intel is becoming an anchor for Apple. I know Apple producing an ARM-based iMac most likely will not happen, but it would be a way for Apple to begin unraveling the binding that Intel has on Apple's innovation.
Give us a break. Your ilk said the same thing about the IBM/Motorola Power PC and the 6502 before that (why oh why did Woz choose that CPU? Answer: It was cheap). Somebody is always a boat anchor for Apple and if Apple would just follow your advice all would be well. I think I’ll put my faith in Apple engineers who actually know what’s going on in the CPU world.
Well I know processor folks at both Intel and Apple and indeed Intel is an anchor for them... the OP said nothing about "all would be well", but for sure the Apple processor folks have a lot more runway to build a better platform.
Guess saying "I think I'll put my faith..." is a good way for anyone without a clue to start a sentence
"but for sure the Apple processor folks have a lot more runway to build a better platform"
Honest question: What do you mean by that?
I'm not a hardware guy, but I'd like to know if Apple's ARM AxX platform (possibly IBMs Power9) or other RISC solution could be used to remove the dependency on Intel.
To my mind, an Apple [non-Intel] platform would have to provide good enough x86 emulation to acceptably run Windows VMs. But the real advantages would be running apps written specifically for the Apple hardware and OS. (I'm sure OSS Swift is somewhere in the equation)
As I understand it, recent Intel chips convert the x86 CISC instructions to RISC for execution.
But, likely, there is a lot more than the CPU... Memory bandwidth, multiple threads, multiple cores, I/O capability, GPUs, and a lot of other things I don't understand all I know about!
Maybe a better way to ask the question: If you are Apple, given a clean slate, what kind of chip would you have to build to remove your dependency on Intel?
Why does Apple have to choose "this or that," Intel or ARM?
What I mean is that I don't get this assumption or suggestion that ARM chips have to replace Intel. Apple currently uses both, why not extend that paradigm and always have both? That way you never have to worry about WIndows VM crowds leaving Apple (or app developers with no interest in completely re-writing their apps for yet another new chip running the same OS), you still have fast chips compatible with all the current macOS apps, you build Intel based systems as you do today until Intel is no longer a company or you can't sell those machines any longer (neither of which I would assume are near term worries).
What you do with your ARM chips is extend those into new devices and leave the Intel/macOS machines alone. You use ARM chips to power new "iOS" laptops (as one example) and other new devices designed for the masses which wouldn't have to encroach on the Intel/macOS territory (much) to begin with. Over time you allow customers to choose and if in the distant future Intel machines seem no longer viable and you have adequate overlap or ARM device replacements, let the market of Intel machines die naturally.
In the meantime keep both as they have today, everyone is happy, no customers leave Apple (or more importantly, app developers), you build devices with your own chips and Intel's for different markets, different use cases. This is what I'd expect from them, not an Intel replacement.
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
You believe Mac Pro will come out next year?
I do. There are months of development time, but I think if they were going to use the Broadwell Xeon E5v4 family, Apple would have done so by now. Intel finished the rollout in June, but many of them were available in March. And while the controllers for Thunderbolt 3 are backward compatible to Broadwell, it is really Skylake that is designed for USB 3.1 and Thunderbolt 3.
There is a leaked set of Intel presentation slides from May 2015 that give some information on this. The Skylake Xeon architecture will feature a new platform (and socket) codenamed Purley with significant advantages over what has come before. Just search for skylake and purley and you'll find various takes, mostly repeating the same things from different points of view. Investors: here and here; Servers: here; Gaming: here and here.
Now, whether the Purley platform would make it into the Mac Pro is another question, because if you look at the "1S Workstation" (1S = single-socket, like the current Mac Pro) line across the bottom of the "Purley RoadMap Positioning" slide, you'll find something called "Basin Falls 1S Workstation Platform" that appears to be sort of a hybrid approach. I won't try to speculate -- I'm just pointing out that there are two directions they can go in here. Apple does use a "2S" processor in the highest-end configuration of the current Mac Pro, so a Mac Pro with Skylake architecture on the Purley platform is not out of the question. Nor is a new, dual-socket Mac Pro design. [Okay, now I'm speculating.] But if I were a betting man (I'm not), I think I would be cautious and put my money on the 1S Workstation family. After all, that's really what the Mac Pro is -- the core of a workstation -- it's not a server.
The most important thing to keep in mind is that the Mac Pro is a very forward-looking design. Apple isn't in a hurry here. They are already way out ahead. Some might say too far. Regardless, these new processor architectures and platforms require new approaches to the increasing heat -- that is what the Mac Pro is all about. In addition, USB 3.1 and Thunderbolt 3 (and beyond) mean the "workstation" is changing, moving toward the more flexible, modular world exemplified by the current Mac Pro.
Best case is a pre-production intro at WWDC, with availability later in the year. Much like 2013. More likely everything gets pushed back, but it will still happen next year.
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
You believe Mac Pro will come out next year?
I do. There are months of development time, but I think if they were going to use the Broadwell Xeon E5v4 family, Apple would have done so by now. Intel finished the rollout in June, but many of them were available in March. And while the controllers for Thunderbolt 3 are backward compatible to Broadwell, it is really Skylake that is designed for USB 3.1 and Thunderbolt 3.
There is a leaked set of Intel presentation slides from May 2015 that give some information on this. The Skylake Xeon architecture will feature a new platform (and socket) codenamed Purley with significant advantages over what has come before. Just search for skylake and purley and you'll find various takes, mostly repeating the same things from different points of view. Investors: here and here; Servers: here; Gaming: here and here.
Now, whether the Purley platform would make it into the Mac Pro is another question, because if you look at the "1S Workstation" (1S = single-socket, like the current Mac Pro) line across the bottom of the "Purley RoadMap Positioning" slide, you'll find something called "Basin Falls 1S Workstation Platform" that appears to be sort of a hybrid approach. I won't try to speculate -- I'm just pointing out that there are two directions they can go in here. Apple does use a "2S" processor in the highest-end configuration of the current Mac Pro, so a Mac Pro with Skylake architecture on the Purley platform is not out of the question. Nor is a new, dual-socket Mac Pro design. [Okay, now I'm speculating.] But if I were a betting man (I'm not), I think I would be cautious and put my money on the 1S Workstation family. After all, that's really what the Mac Pro is -- the core of a workstation -- it's not a server.
The most important thing to keep in mind is that the Mac Pro is a very forward-looking design. Apple isn't in a hurry here. They are already way out ahead. Some might say too far. Regardless, these new processor architectures and platforms require new approaches to the increasing heat -- that is what the Mac Pro is all about. In addition, USB 3.1 and Thunderbolt 3 (and beyond) mean the "workstation" is changing, moving toward the more flexible, modular world exemplified by the current Mac Pro.
Best case is a pre-production intro at WWDC, with availability later in the year. Much like 2013. More likely everything gets pushed back, but it will still happen next year.
Interesting. Then...would this be another gpu soldered up or can it be upgraded? So then, sky lake it is then. broadwell wouldn't be compatible with 3.1?
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
You believe Mac Pro will come out next year?
I do. There are months of development time, but I think if they were going to use the Broadwell Xeon E5v4 family, Apple would have done so by now. Intel finished the rollout in June, but many of them were available in March. And while the controllers for Thunderbolt 3 are backward compatible to Broadwell, it is really Skylake that is designed for USB 3.1 and Thunderbolt 3.
There is a leaked set of Intel presentation slides from May 2015 that give some information on this. The Skylake Xeon architecture will feature a new platform (and socket) codenamed Purley with significant advantages over what has come before. Just search for skylake and purley and you'll find various takes, mostly repeating the same things from different points of view. Investors: here and here; Servers: here; Gaming: here and here.
Now, whether the Purley platform would make it into the Mac Pro is another question, because if you look at the "1S Workstation" (1S = single-socket, like the current Mac Pro) line across the bottom of the "Purley RoadMap Positioning" slide, you'll find something called "Basin Falls 1S Workstation Platform" that appears to be sort of a hybrid approach. I won't try to speculate -- I'm just pointing out that there are two directions they can go in here. Apple does use a "2S" processor in the highest-end configuration of the current Mac Pro, so a Mac Pro with Skylake architecture on the Purley platform is not out of the question. Nor is a new, dual-socket Mac Pro design. [Okay, now I'm speculating.] But if I were a betting man (I'm not), I think I would be cautious and put my money on the 1S Workstation family. After all, that's really what the Mac Pro is -- the core of a workstation -- it's not a server.
The most important thing to keep in mind is that the Mac Pro is a very forward-looking design. Apple isn't in a hurry here. They are already way out ahead. Some might say too far. Regardless, these new processor architectures and platforms require new approaches to the increasing heat -- that is what the Mac Pro is all about. In addition, USB 3.1 and Thunderbolt 3 (and beyond) mean the "workstation" is changing, moving toward the more flexible, modular world exemplified by the current Mac Pro.
Best case is a pre-production intro at WWDC, with availability later in the year. Much like 2013. More likely everything gets pushed back, but it will still happen next year.
Interesting. Then...would this be another gpu soldered up or can it be upgraded? So then, sky lake it is then. broadwell wouldn't be compatible with 3.1?
On the first question, recent history would suggest soldered. But with the new platform and new socket, I don't know. I'm not clear on what factors go into Apple's decision-making there. There may be improvements that mitigate their concerns.
With regard to the second question, see here. Intel's Thunderbolt 3 controllers support USB 3.1 and are compatible with Broadwell and Haswell even though they "ship" (i.e., are the default) with Skylake. So Broadwell is compatible with USB 3.1. There is nothing to prevent Apple from using Broadwell in a revised TB3/USB-C Mac Pro, but as each day passes, it becomes less and less likely. Rumors said end of November, but that was probably just someone's common-sense guess, projecting six months development time from when the appropriate Broadwell chips became available. Almost certainly not based on insider information.
The PC market is rapidly changing, in part because Intel has slowed down. I think the larger factor is just that most people don't really need a faster computer anymore, so they keep their existing computers for a lot longer. I used to get a new computer every year, but my current computer is four years old and still doesn't feel slow (added a SSD).
Lots of nerds complain about this, but Apple has exactly the right idea. They're slowing down their release cycles because the vast majority of people are slowing down their purchases, and the current crop of computers is fast enough. Personally, at the current speeds, Skylake (just bought a 15" MBP) is plenty fast for everything I do, and I'd rather have either lighter weight or longer battery life. I probably won't buy my next computer for four or five more years, at which point Apple will probably have something new and cool for me to buy.
The only real issue I see is the Mac Pro, which truly is behind the times in every way. I suspect Apple will be completely changing the design as the current one isn't much of a Pro computer anyway.
"Pro" has become a fashion statement for executives and is now less useful as an identifier actual professionals can rely on.
How is this related to Mac Pro? I thought it uses e5 type? I didn't see e5 on it.
It's not. Putting "Mac Pro" in the title and having a section called "The Mac Pro is way overdue" is just an attempt to get you and I to click. It worked, too!
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
You believe Mac Pro will come out next year?
I do. There are months of development time, but I think if they were going to use the Broadwell Xeon E5v4 family, Apple would have done so by now. Intel finished the rollout in June, but many of them were available in March. And while the controllers for Thunderbolt 3 are backward compatible to Broadwell, it is really Skylake that is designed for USB 3.1 and Thunderbolt 3.
There is a leaked set of Intel presentation slides from May 2015 that give some information on this. The Skylake Xeon architecture will feature a new platform (and socket) codenamed Purley with significant advantages over what has come before. Just search for skylake and purley and you'll find various takes, mostly repeating the same things from different points of view. Investors: here and here; Servers: here; Gaming: here and here.
Now, whether the Purley platform would make it into the Mac Pro is another question, because if you look at the "1S Workstation" (1S = single-socket, like the current Mac Pro) line across the bottom of the "Purley RoadMap Positioning" slide, you'll find something called "Basin Falls 1S Workstation Platform" that appears to be sort of a hybrid approach. I won't try to speculate -- I'm just pointing out that there are two directions they can go in here. Apple does use a "2S" processor in the highest-end configuration of the current Mac Pro, so a Mac Pro with Skylake architecture on the Purley platform is not out of the question. Nor is a new, dual-socket Mac Pro design. [Okay, now I'm speculating.] But if I were a betting man (I'm not), I think I would be cautious and put my money on the 1S Workstation family. After all, that's really what the Mac Pro is -- the core of a workstation -- it's not a server.
The most important thing to keep in mind is that the Mac Pro is a very forward-looking design. Apple isn't in a hurry here. They are already way out ahead. Some might say too far. Regardless, these new processor architectures and platforms require new approaches to the increasing heat -- that is what the Mac Pro is all about. In addition, USB 3.1 and Thunderbolt 3 (and beyond) mean the "workstation" is changing, moving toward the more flexible, modular world exemplified by the current Mac Pro.
Best case is a pre-production intro at WWDC, with availability later in the year. Much like 2013. More likely everything gets pushed back, but it will still happen next year.
Interesting. Then...would this be another gpu soldered up or can it be upgraded? So then, sky lake it is then. broadwell wouldn't be compatible with 3.1?
On the first question, recent history would suggest soldered. But with the new platform and new socket, I don't know. I'm not clear on what factors go into Apple's decision-making there. There may be improvements that mitigate their concerns.
With regard to the second question, see here. Intel's Thunderbolt 3 controllers support USB 3.1 and are compatible with Broadwell and Haswell even though they "ship" (i.e., are the default) with Skylake. So Broadwell is compatible with USB 3.1. There is nothing to prevent Apple from using Broadwell in a revised TB3/USB-C Mac Pro, but as each day passes, it becomes less and less likely. Rumors said end of November, but that was probably just someone's common-sense guess, projecting six months development time from when the appropriate Broadwell chips became available. Almost certainly not based on insider information.
[...] Then...would this be another gpu soldered up or can it be upgraded? [...]
On the first question, recent history would suggest soldered. But with the new platform and new socket, I don't know. I'm not clear on what factors go into Apple's decision-making there. There may be improvements that mitigate their concerns.
Thanks for clarification.
I'm sorry, I was wrong about this. The 2013 Mac Pro GPUs aren't soldered, they are non-standard, custom cards that only fit the Mac Pro. So one reason you can't upgrade them is simply because nothing else will fit. There may be other reasons, like heat-dissipation concerns re: performance. Nonetheless, the upshot doesn't change -- how it all would fit together in a Purley-based Mac Pro is not clear, at least not to me.
The PC market is rapidly changing, in part because Intel has slowed down. I think the larger factor is just that most people don't really need a faster computer anymore, so they keep their existing computers for a lot longer.
You conception of the world may be inverted: people keep their computers for much longer because tech advances in processor speeds have stagnated ... not because they don't need a faster computer. The simple fact is that significantly faster mainstream computers are not available ... and the only real area to see gains is in multi-processing ... and unless you need a 20000 processor computer, that's not very helpful for most people.
im not so sure you're correct -- as a non-gamer, i find my 2011 iMac w/ SSD and maxed RAM and VRAM is still adequate for my pro needs (software dev, VMs, etc). a faster desktop won't really do that much for my use case. I'm certainly not alone...normal consumers' demands would be even less than my own.
I believe you're both partially right and wrong : yep, many normal usage do not require lot more power than 5 years ago (office suite, internet, cute cats vids, ...) but also, OS updates on a regular basis my old iMac makes it smoothly slower and slower. Also, normal usage IS more and more demanding : a few years ago, to learn something new, was reading a wikipedia article, now I'm watching a video conf/presentation. And to balance the previous statement, there is also the switching cost : I won't invest x k$ if I value the advantages I'll get for x at a small part of x. Since the progression between generations of PCs is less impressive it has been in the past, the lever to switch is less important.
Any chance Apple are just waiting for PCIE 4.0 to be finalised? And yes, I'm assuming slots are making a comeback
It's possible, but only if Intel is also waiting for it to be finalized... Skylake and/or Kaby Lake can't support it, at least not yet. That is, they can't be "forward-compatible" to an upcoming standard like that. That would completely upend the whole point of the standards process.
I don't know, but I guess it's possible that Purley and the Skylake Xeons will support it, as they are still in development and thus can implement it and test it while waiting for it to be finalized. Same goes for the single-processor "Basin Falls" workstation CPUs mentioned earlier.
Dunno about slots. Does anyone remember what Apple said (if anything) to explain dropping them in the first place? It does seem like the new standard is greatly improved when it comes to efficiency of power consumption, so maybe? But I wouldn't bet on it.
I wonder if we won’t see scalable chipsets in the future while we bridge the gap between transistors that can’t get any smaller and q-bits or whatever we have to use next. Need more processing power? Buy a stick of CPU to plug in next to your stick of RAM!
That sounds like the IBM cell chip from a few years back.
That's the one in the PS3, right? The one that was notoriously difficult to develop for? That doesn't sound like a rosy future, where the only way to improve hardware performance is to dramatically increase software complexity.
Comments
"but for sure the Apple processor folks have a lot more runway to build a better platform"
Honest question: What do you mean by that?
I'm not a hardware guy, but I'd like to know if Apple's ARM AxX platform (possibly IBMs Power9) or other RISC solution could be used to remove the dependency on Intel.
To my mind, an Apple [non-Intel] platform would have to provide good enough x86 emulation to acceptably run Windows VMs. But the real advantages would be running apps written specifically for the Apple hardware and OS. (I'm sure OSS Swift is somewhere in the equation)
As I understand it, recent Intel chips convert the x86 CISC instructions to RISC for execution.
But, likely, there is a lot more than the CPU... Memory bandwidth, multiple threads, multiple cores, I/O capability, GPUs, and a lot of other things I don't understand all I know about!
Maybe a better way to ask the question: If you are Apple, given a clean slate, what kind of chip would you have to build to remove your dependency on Intel?
Does anybody know if AMD offers anything close to/better than Intel?
What I mean is that I don't get this assumption or suggestion that ARM chips have to replace Intel. Apple currently uses both, why not extend that paradigm and always have both? That way you never have to worry about WIndows VM crowds leaving Apple (or app developers with no interest in completely re-writing their apps for yet another new chip running the same OS), you still have fast chips compatible with all the current macOS apps, you build Intel based systems as you do today until Intel is no longer a company or you can't sell those machines any longer (neither of which I would assume are near term worries).
What you do with your ARM chips is extend those into new devices and leave the Intel/macOS machines alone. You use ARM chips to power new "iOS" laptops (as one example) and other new devices designed for the masses which wouldn't have to encroach on the Intel/macOS territory (much) to begin with. Over time you allow customers to choose and if in the distant future Intel machines seem no longer viable and you have adequate overlap or ARM device replacements, let the market of Intel machines die naturally.
In the meantime keep both as they have today, everyone is happy, no customers leave Apple (or more importantly, app developers), you build devices with your own chips and Intel's for different markets, different use cases. This is what I'd expect from them, not an Intel replacement.
Let me try to rewrite that section:
"The Mac Pro is finally due"
This document shows only one Xeon processor in the "v6" Kaby Lake family, the 3.0GHz E3-1205v6. Beyond that, the number of cores and threads is not known at this time, but it is not relevant to the Mac Pro.
The 2013 Mac Pro utilizes the E5-1620, E5-1650, E5-1680, or E5-2697, all based on the "v2" Ivy Bridge technology that is at this point several two generations old. Given how the Xeon line progresses, there is no specific predecessor to compare the new model number to. In June 2016, Intel completed its delayed release of the "v4" Broadwell editions of the Xeon E5-1600 and E5-2600 families. However, to date, Apple has not used them to revise the Mac Pro, despite their availability.
It is possible that Apple will skip v4 and move to the v5 Skylake architecture, expected in the first half of 2017. Because of the delay, the Broadwell lifecycle will be shorter than usual.
There is a leaked set of Intel presentation slides from May 2015 that give some information on this. The Skylake Xeon architecture will feature a new platform (and socket) codenamed Purley with significant advantages over what has come before. Just search for skylake and purley and you'll find various takes, mostly repeating the same things from different points of view. Investors: here and here; Servers: here; Gaming: here and here.
Now, whether the Purley platform would make it into the Mac Pro is another question, because if you look at the "1S Workstation" (1S = single-socket, like the current Mac Pro) line across the bottom of the "Purley RoadMap Positioning" slide, you'll find something called "Basin Falls 1S Workstation Platform" that appears to be sort of a hybrid approach. I won't try to speculate -- I'm just pointing out that there are two directions they can go in here. Apple does use a "2S" processor in the highest-end configuration of the current Mac Pro, so a Mac Pro with Skylake architecture on the Purley platform is not out of the question. Nor is a new, dual-socket Mac Pro design. [Okay, now I'm speculating.] But if I were a betting man (I'm not), I think I would be cautious and put my money on the 1S Workstation family. After all, that's really what the Mac Pro is -- the core of a workstation -- it's not a server.
The most important thing to keep in mind is that the Mac Pro is a very forward-looking design. Apple isn't in a hurry here. They are already way out ahead. Some might say too far. Regardless, these new processor architectures and platforms require new approaches to the increasing heat -- that is what the Mac Pro is all about. In addition, USB 3.1 and Thunderbolt 3 (and beyond) mean the "workstation" is changing, moving toward the more flexible, modular world exemplified by the current Mac Pro.
Best case is a pre-production intro at WWDC, with availability later in the year. Much like 2013. More likely everything gets pushed back, but it will still happen next year.
With regard to the second question, see here. Intel's Thunderbolt 3 controllers support USB 3.1 and are compatible with Broadwell and Haswell even though they "ship" (i.e., are the default) with Skylake. So Broadwell is compatible with USB 3.1. There is nothing to prevent Apple from using Broadwell in a revised TB3/USB-C Mac Pro, but as each day passes, it becomes less and less likely. Rumors said end of November, but that was probably just someone's common-sense guess, projecting six months development time from when the appropriate Broadwell chips became available. Almost certainly not based on insider information.
I don't know, but I guess it's possible that Purley and the Skylake Xeons will support it, as they are still in development and thus can implement it and test it while waiting for it to be finalized. Same goes for the single-processor "Basin Falls" workstation CPUs mentioned earlier.
Dunno about slots. Does anyone remember what Apple said (if anything) to explain dropping them in the first place? It does seem like the new standard is greatly improved when it comes to efficiency of power consumption, so maybe? But I wouldn't bet on it.