The switch to ARM may mean that Mac is willing to give up workstation crowd too. (Because, er, making workstations will mean that Apple CAN'T put the same chips in workstations that they put in iPhones, ok? There are power/heating/expense constraints that smartphones have to work within. Make custom Apple Silicon to run in workstations? Yeah ... that's a worthwhile expense for the 250-500k Mac Pros that they sell a year. Not to mention it would drive up the cost.)
The purpose of Mac switching to ARM may well to increase convergence with iPad and iPhone users. Not to match Intel i9s and Xeons on computing power.
Why would the thermal constraints of a mobile device dictate the design of a SoC destined for a desktop Mac? Why would you think they'd be using the same processors in both phones and workstations to begin with? You seem to be operating under some assumptions which are hobbling your ability to picture Apple designing chips that are competitive with the higher end Intel chips we've been enjoying. I've seen no evidence to support this, and the fact they intend to complete the transition by the end of next year makes me think they've already got the high end figured out. Mac Pro might be last, but I certainly don't think they spent the years they did on that platform to release it once with Xeons and ditch it, and I certainly don't expect them to continue to release Intel Macs after next year.
The "increase convergence with iPad and iPhone users" thing is, what, regarding cross compatibility with software? 99% of consumers won't know or care anything about any of this. The chip is irrelevant to them.
I think you read that backwards. Linuxplatform is saying the Mac Pro sales volume may not justify the work involved in building a chip only for it. Chips above ~80W TDP wouldn't be suitable for anything but a Mac Pro (and maybe iMac Pro if they keep that around). None of the other current designs can dissipate that much heat well enough.
Of course, it ignores multi-socket systems. There have been two-socket PowerMacs and Mac Pros before, and I suspect there will be again with the switch to ARM. iMac with a single 65W processor, Mac Pro with 2 to 6. It would take a little extra engineering on the inter-processor bus, but that bus is a slower-moving target, so the investment has more time to pay itself off.
As stated in the Platforms State of the Union video at WWDC 2020, they 1) state that they are building chips that are tailored to the unique needs of each Mac, and 2) how they built a scalable architecture that allowed for all the Apple Silicon to date and which will be scalable for the variety of Mac types and uses going forward.
This starts at around 8 minute in with Sri Santhanam:
I think the Mac Pro will likely be the last machine to get this treatment and I hope that it's socketed. I have no idea how Apple will introduce their scalable architecture for high-powered desktops. Maybe they'll design it so that multiple chips that would be found in the iMac or even the MBP are used or maybe the cost and efficiency will dictate having a specific, low-volume chip for the Mac Pro. There are too many unknown factors for use to make any determination.
One thing I look forward to is the potential for Apple to design a Mac Pro that can be updated with a decent regularity (hopefully an annual cycle), and allow for component upgrades. As I recall, the "trashcan" Mac Pro's shortcoming was the lack of available GPU options. I seem to recall that other socketed Macs had CPU limitations. Apple would be able to better control these options going forward, even if only allowing for better accessibility with TB4.
Intel won't be at 10 nm forever. AMD is at 7 nm, after all, and is expected to reach 5 nm as early as 2021.
That's the problem. It seems like Intel will be stuck at 10nm well past Apple's transition to 3nm. There are very few Intel processors that are at 10nm now, with most at 14nm. Apple will be at 5 nm and likely 5nm + nodes by the time Intel completes its product line transition to 10 nm. The node disadvantage is killing Intel, and AMD would certainly agree with that.
You surmise that Apple won't be able to match the priciest Intel i9 and Xeon products lines, but I'm not sure that will be true at the end of Appel's transition.
Either way, defending Intel at this point in time is a hilll I don't value dying on, but make yourself at home.
With all due respect, it is less "knowledge of business, supply chain and vertical integration" that gets bandied about by journalism graduates who have never worked a day in engineering, manufacturing, product management etc. than it is Apple boosterism. Simply put: do you folks honestly think that Apple will be the first company to make their own CPUs? Because if you want vertical integration, Samsung has that in spades. They make the CPUs, RAM, SSD, screens and cameras! And unlike Apple, Samsung actually MAKES these components in their own factories and foundries where "Apple Silicon" is actually made by TSMC (and was previously by Samsung). And before Samsung, IBM and Motorola used to make their own components too.
We get it: iOS is faster than Android and Apple Silicon is faster than Qualcomm (and Exynos, MediaTek and Kirin). But that doesn't translate everywhere. Allow me to say that I have long been a fan of RISC, which ARM is a subset of. I remember when Sun SPARC and Motorola 68xxx UNIX workstations and servers could crush anything that Wintel was capable of. I have also been keeping up with ARM-based servers, which some quarters have been hyping for years. Linus Torvalds claims that Mac switching to ARM will be the catalyst for ARM-based servers really taking off.
But please know this: not even Apple claims that their 5 nm A14 chip will outperform the 10 nm Intel i9 or even the i7. They merely claimed that the iPad Pro beat an unspecified MacBook (i9? i7? even i5?) on some internal tests. So keep these 3 things in mind.
1. The MacBook Pro runs tons of heavy duty performance software that the iPad Pro can't run at all rendering that test worthless for people with serious computing needs. 2. The i9 isn't even Intel's most powerful chip. The Xeon, which goes in the Mac Pro, is. 3. Intel won't be at 10 nm forever. AMD is at 7 nm, after all, and is expected to reach 5 nm as early as 2021.
What you aren't considering: "Pro" users whose computing needs tend to the ultra-high performance scale make up a tiny percentage of Mac sales. We already know that Mac is willing to give up the similarly tiny percentage of Windows (bootcamp and virtualization) users. The switch to ARM may mean that Mac is willing to give up workstation crowd too. (Because, er, making workstations will mean that Apple CAN'T put the same chips in workstations that they put in iPhones, ok? There are power/heating/expense constraints that smartphones have to work within. Make custom Apple Silicon to run in workstations? Yeah ... that's a worthwhile expense for the 250-500k Mac Pros that they sell a year. Not to mention it would drive up the cost.)
The purpose of Mac switching to ARM may well to increase convergence with iPad and iPhone users. Not to match Intel i9s and Xeons on computing power.
Not exactly. According to this promo video (around the 1:34 mark) Apple claims that the iPad Pro is faster than 92% of portable PC's.
But please know this: not even Apple claims that their 5 nm A14 chip will outperform the 10 nm Intel i9 or even the i7. They merely claimed that the iPad Pro beat an unspecified MacBook (i9? i7? even i5?) on some internal tests. So keep these 3 things in mind.
Apple hasn't even acknowledged the existence of their A14 chip yet. Do you realize that the A13 chip (single core performance) is within 6% of the fastest Intel i9 based core? Do you really think the A14 won't exceed that? Also, that's just the phone chip. Apple also said they are coming out with a line of chips for the Mac specifically.
1. The MacBook Pro runs tons of heavy duty performance software that the iPad Pro can't run at all rendering that test worthless for people with serious computing needs. 2. The i9 isn't even Intel's most powerful chip. The Xeon, which goes in the Mac Pro, is. 3. Intel won't be at 10 nm forever. AMD is at 7 nm, after all, and is expected to reach 5 nm as early as 2021.
1. That has more to do with memory than processor performance. The fact that we saw Maya running perfectly smoothly under Rosetta 2 emulation throws shade on your argument. 2. In terms of single core performance, the fastest benchmarks recorded are from the Intel Core i9-10900K. Yes, the Intel Xeon W-3175X is faster at multi-core performance. However, the challenge here is to develop a fast core. Sticking a bunch of them on a chip isn't the hard part. 3. TSMC is already coming out with their first 5nm chips with the A14. They will likely be at 3nm by the time Intel is ready to move off of 10nm.
The switch to ARM may mean that Mac is willing to give up workstation crowd too. (Because, er, making workstations will mean that Apple CAN'T put the same chips in workstations that they put in iPhones, ok? There are power/heating/expense constraints that smartphones have to work within. Make custom Apple Silicon to run in workstations? Yeah ... that's a worthwhile expense for the 250-500k Mac Pros that they sell a year. Not to mention it would drive up the cost.)
The purpose of Mac switching to ARM may well to increase convergence with iPad and iPhone users. Not to match Intel i9s and Xeons on computing power.
Why would the thermal constraints of a mobile device dictate the design of a SoC destined for a desktop Mac? Why would you think they'd be using the same processors in both phones and workstations to begin with? You seem to be operating under some assumptions which are hobbling your ability to picture Apple designing chips that are competitive with the higher end Intel chips we've been enjoying. I've seen no evidence to support this, and the fact they intend to complete the transition by the end of next year makes me think they've already got the high end figured out. Mac Pro might be last, but I certainly don't think they spent the years they did on that platform to release it once with Xeons and ditch it, and I certainly don't expect them to continue to release Intel Macs after next year.
The "increase convergence with iPad and iPhone users" thing is, what, regarding cross compatibility with software? 99% of consumers won't know or care anything about any of this. The chip is irrelevant to them.
I think you read that backwards. Linuxplatform is saying the Mac Pro sales volume may not justify the work involved in building a chip only for it. Chips above ~80W TDP wouldn't be suitable for anything but a Mac Pro (and maybe iMac Pro if they keep that around). None of the other current designs can dissipate that much heat well enough.
Of course, it ignores multi-socket systems. There have been two-socket PowerMacs and Mac Pros before, and I suspect there will be again with the switch to ARM. iMac with a single 65W processor, Mac Pro with 2 to 6. It would take a little extra engineering on the inter-processor bus, but that bus is a slower-moving target, so the investment has more time to pay itself off.
As stated in the Platforms State of the Union video at WWDC 2020, they 1) state that they are building chips that are tailored to the unique needs of each Mac, and 2) how they built a scalable architecture that allowed for all the Apple Silicon to date and which will be scalable for the variety of Mac types and uses going forward.
This starts at around 8 minute in with Sri Santhanam:
I think the Mac Pro will likely be the last machine to get this treatment and I hope that it's socketed. I have no idea how Apple will introduce their scalable architecture for high-powered desktops. Maybe they'll design it so that multiple chips that would be found in the iMac or even the MBP are used or maybe the cost and efficiency will dictate having a specific, low-volume chip for the Mac Pro. There are too many unknown factors for use to make any determination.
One thing I look forward to is the potential for Apple to design a Mac Pro that can be updated with a decent regularity (hopefully an annual cycle), and allow for component upgrades. As I recall, the "trashcan" Mac Pro's shortcoming was the lack of available GPU options. I seem to recall that other socketed Macs had CPU limitations. Apple would be able to better control these options going forward, even if only allowing for better accessibility with TB4.
I saw, but I still think the question of whether Apple will bother making a seriously-high-power chip is a valid one. The Mac Pro ships 2-3 orders of magnitude fewer units than anything else Apple builds their own chips for. Chips suitable for it would also need a lot of cores, which means a big die. Reject rate goes up with the square of the area, so a big die means dramatically more rejected chips and a corresponding increase in per-unit cost.
Scale-out is the obvious way to address this, as it lets you use smaller chips while have a lower reject rate. If a flaw falls in the area of a 20mm by 20mm die, it can trash the whole thing. Meanwhile, if you spend the same 20mm by 20mm area on four 10mm by 10mm chips, the same single flaw will only cost you one or two of the four.
I doubt Apple is going to do socketed ARMs, though. Half the security story of using their own chips is integrating the SSD controller into the processor so if the thing is stolen, it can be bricked. I only said "multi-socket" because I'm not sure anybody has done a multi-chip non-socketed system before. My money would be on something NUMA-like where one processor's SSD controller owns the SSD directly (or maybe two processors allowing for even greater storage performance) while the others have to go through the storage processor to get to the storage.
I could see them using a separate T2 just to manage the storage for all of the main processors. That would allow for a similar level of security while also allowing processors to be replaced as needed. They do sell flash cards for the Mac Pro after purchase now.
I will be incredibly surprised if Apple lowers prices on any Mac as a result of this transition, and IF they actually do, it will be a short-lived reduction to inspire transitional purchases.
I quite disagree. Apple, and Tim Cook in particular, are sticklers about profit margins. They pick some number (usually 32%) and price accordingly. .....
Do you have a reputable source for that assertion? I've heard speculation about such things but never anything actually informed.
And, in actuality, the only ones who actually know the real margins are Apple cost accountants and execs. Everybody else is guessing. And their guesses usually reveal a fundamental lack of understanding of cost accounting and margins.
Apple states product line margins in their quarterly financial statements, so its not like anyone is guessing.
They also bury a lot by glombing multiple products together. It's pretty meaningless stuff really.
I will be incredibly surprised if Apple lowers prices on any Mac as a result of this transition, and IF they actually do, it will be a short-lived reduction to inspire transitional purchases.
I quite disagree. Apple, and Tim Cook in particular, are sticklers about profit margins. They pick some number (usually 32%) and price accordingly. .....
Do you have a reputable source for that assertion? I've heard speculation about such things but never anything actually informed.
And, in actuality, the only ones who actually know the real margins are Apple cost accountants and execs. Everybody else is guessing. And their guesses usually reveal a fundamental lack of understanding of cost accounting and margins.
Apple states product line margins in their quarterly financial statements, so its not like anyone is guessing.
They also bury a lot by glombing multiple products together. It's pretty meaningless stuff really.
Your meaning is understood, but I don't think glomb in the word you think it is.
Apple's 10-Q filings to the SEC are not meaningless and clearly used by investors to determine Apple's success in a multitude of product categories.
I will be incredibly surprised if Apple lowers prices on any Mac as a result of this transition, and IF they actually do, it will be a short-lived reduction to inspire transitional purchases.
How do you explain away other Apple decreases in price? Like the regular iPhone 11 over the prior year’s; the iPad, MacBook decreases, etc?
Sorry, I wasn’t thinking about iOS devices. I don’t actually know the pricing of the last several models, though I do remember extremely expensive devices in general (that’s partially because I find it insane to spend more than $500 on a PHONE).
.... Kuo speculates that some models may sell at lower price points than the current range. Kuo doesn't specify whether that is because of a cost savings due to a greater volume being manufactured, but Apple will certainly no longer have to pay a fee to Intel for its processors.
....
Yes, a price reduction might come from decreasing costs due to either (or both) lower production costs or higher volume.
But it is equally as likely that prices might be reduced from a marketing perspective to get the ball rolling. Apple maintains high enough margins on these products that they have room to fiddle with them and lowering selling prices could help stir up excitement in the market.
As for lower production costs (by not paying Intel), that assumption is likely true -- but it may also be eliminated or mitigated by Apple's increase in the fixed costs of R&D, development and support of this new, internal, product. The analogy might be the costs for developers: yes, their cost of deployment is increased by 30% by the Apple Store -- but without the Apple Store their marketing costs would likely increase by that amount or possibly even more.
In fact, as Kuo also implies: selling prices may be reduced simply in order to increase volume in an effort to absorb those increased fixed costs into lower unit costs from the resulting increased volume.
I will be incredibly surprised if Apple lowers prices on any Mac as a result of this transition, and IF they actually do, it will be a short-lived reduction to inspire transitional purchases.
I quite disagree. Apple, and Tim Cook in particular, are sticklers about profit margins. They pick some number (usually 32%) and price accordingly. .....
Do you have a reputable source for that assertion? I've heard speculation about such things but never anything actually informed.
And, in actuality, the only ones who actually know the real margins are Apple cost accountants and execs. Everybody else is guessing. And their guesses usually reveal a fundamental lack of understanding of cost accounting and margins.
With these new processors, if iPad apps are available for Mac, with a long battery life, thinner Macs, with the Mac Os system being much more versatile than iPads, I don't see a future in iPads
Things are going to get more expensive as Apple squeezes its customers through software - and will justify higher hardware prices soon after. Apple user's already pay 30% more for everything - software, ringtones, in-game purchases, subscriptions, books, hardware, cables, hotel rooms, you name it... Once Apple has full control over its users - $700 coasters are going to look like a deal.
Put down the crack pipe, you've had quite enough.
You must be pretty new to the world - definitely Apple's history.
I've been using Apple products since learning LOGO on the Apple II, but nice try.
What do you mean, we pay 30% more for hotel rooms? Books? Ringtones? What the hell are you talking about?
So I wonder at what point Apple decided that they were going to bin Intel?
I thought that perhaps Intel had always been a stop-gap, but a friend of mine thinks that it was when they couldn't fix the thermal problems with the Trashcan Mac Pro.
No, they actually talked about this. The problem with the Mac Pro was that it was designed for two smaller GPUs running in parallel, and that's not the direction the industry was moving.
The small, trash can-shaped Mac Pro — which Apple marketing VP Phil Schiller once touted as evidence that the company could still innovate — was designed to fit two smaller graphics chips, but the industry didn’t move in that direction. “Being able to put larger single GPUs required a different system architecture and more thermal capacity than that system was designed to accommodate,” the exec is reported as saying. “So it became fairly difficult to adjust.”
Skylake was one of the problems, as has also been covered:
I am gonna bet that it will be just iPad Pro running MacOS - same designs. Only that it will have more RAM and more speed for MacOS.
It will absolutely not nee an iPad Pro running macOS. It will be a MacBook Pro, a notebook computer, running macOS. It will definitely have more RAM than an iPad Pro, as we've seen with the developer's transition kit, but hopefully it will be LPDDR4 since they don't have to wait for Intel to get with the program.
That's not the point. My point is that the new MacBook Pro will look like iPad Pro with keyboard. There is a limit to how much you can cram a camera inside a thin display and considering that a camera has not improved whatsoever due to depth limitations in the past decade, it's inevitable that it will have an iPad Pro form factor so it will be able put a better camera on screen and the keyboard is just the same kind you'd get with current iPad Pro Smart Keyboard.
It just seems like a natural evolution. I have been saying this since iPad Pro with Smart Keyboard came out. I just know Apple is going to make MacBook Pros look like them running MacOS on ARM. It will at least finally have a real advantage.
No way. First off, the idea that the FaceTime camera would drive such a massively fucked up design decision like making top-heavy iPad Pro laptops is absurd. Why would they move the guts of the Mac into the display just to accommodate more depth for the camera? That's ridiculous. Not to mention the reduction in battery capacity, thermal envelope for faster chips, trackpads, sound, uh...basically everything that makes a Mac a Mac.
Possibility of macOS on an iPad Pro? Definitely. MacBooks turning into iPads with Smart Cover style keyboards? Not a chance.
Actually he’s right - it already does. Have you seen a picture of a 12” iPad Pro with a Brydge keyboard attached? It looks like a MacBook Pro!
What a keyboard case looks like is completely meaningless to the discussion. You completely ignored everything I pointed out.
And you completely ignored the fact that I was responding so what @netrox said, but whatever.
Uh, I can follow the thread just fine. Your response still makes no sense.
Comments
This starts at around 8 minute in with Sri Santhanam:
https://developer.apple.com/videos/play/wwdc2020/102/
I think the Mac Pro will likely be the last machine to get this treatment and I hope that it's socketed. I have no idea how Apple will introduce their scalable architecture for high-powered desktops. Maybe they'll design it so that multiple chips that would be found in the iMac or even the MBP are used or maybe the cost and efficiency will dictate having a specific, low-volume chip for the Mac Pro. There are too many unknown factors for use to make any determination.
One thing I look forward to is the potential for Apple to design a Mac Pro that can be updated with a decent regularity (hopefully an annual cycle), and allow for component upgrades. As I recall, the "trashcan" Mac Pro's shortcoming was the lack of available GPU options. I seem to recall that other socketed Macs had CPU limitations. Apple would be able to better control these options going forward, even if only allowing for better accessibility with TB4.
You surmise that Apple won't be able to match the priciest Intel i9 and Xeon products lines, but I'm not sure that will be true at the end of Appel's transition.
Either way, defending Intel at this point in time is a hilll I don't value dying on, but make yourself at home.
https://www.youtube.com/watch?v=S8JBXrLDkGs
1. That has more to do with memory than processor performance. The fact that we saw Maya running perfectly smoothly under Rosetta 2 emulation throws shade on your argument.
2. In terms of single core performance, the fastest benchmarks recorded are from the Intel Core i9-10900K. Yes, the Intel Xeon W-3175X is faster at multi-core performance. However, the challenge here is to develop a fast core. Sticking a bunch of them on a chip isn't the hard part.
3. TSMC is already coming out with their first 5nm chips with the A14. They will likely be at 3nm by the time Intel is ready to move off of 10nm.
Scale-out is the obvious way to address this, as it lets you use smaller chips while have a lower reject rate. If a flaw falls in the area of a 20mm by 20mm die, it can trash the whole thing. Meanwhile, if you spend the same 20mm by 20mm area on four 10mm by 10mm chips, the same single flaw will only cost you one or two of the four.
I doubt Apple is going to do socketed ARMs, though. Half the security story of using their own chips is integrating the SSD controller into the processor so if the thing is stolen, it can be bricked. I only said "multi-socket" because I'm not sure anybody has done a multi-chip non-socketed system before. My money would be on something NUMA-like where one processor's SSD controller owns the SSD directly (or maybe two processors allowing for even greater storage performance) while the others have to go through the storage processor to get to the storage.
I could see them using a separate T2 just to manage the storage for all of the main processors. That would allow for a similar level of security while also allowing processors to be replaced as needed. They do sell flash cards for the Mac Pro after purchase now.
They also bury a lot by glombing multiple products together. It's pretty meaningless stuff really.
Apple's 10-Q filings to the SEC are not meaningless and clearly used by investors to determine Apple's success in a multitude of product categories.
As for lower production costs (by not paying Intel), that assumption is likely true -- but it may also be eliminated or mitigated by Apple's increase in the fixed costs of R&D, development and support of this new, internal, product. The analogy might be the costs for developers: yes, their cost of deployment is increased by 30% by the Apple Store -- but without the Apple Store their marketing costs would likely increase by that amount or possibly even more.
And here’s a twenty year graph, basically flat ever since Cook took the reigns. https://www.macrotrends.net/stocks/charts/AAPL/apple/profit-margins
What do you mean, we pay 30% more for hotel rooms? Books? Ringtones? What the hell are you talking about?
No, they actually talked about this. The problem with the Mac Pro was that it was designed for two smaller GPUs running in parallel, and that's not the direction the industry was moving.
Skylake was one of the problems, as has also been covered:
https://appleinsider.com/articles/20/06/24/intel-skylake-chip-issues-reportedly-tipping-point-in-apples-silicon-switch
Uh, I can follow the thread just fine. Your response still makes no sense.