I have been hearing that same line "It appears that silicon is reaching limits, imposed by the laws of physics." since the mid-1990's. Yet, mysteriously year after year; we see performance improvements in the 30-50% range. Those improvements do not sound anything like hitting a limit, improvements in the 2-5% range sound like you are banging against a limit.
What we read in the ‘90s was that we would be reaching the limits when we got processes down to the point in which quantum mechanics would prevent further advances. They knew about where that would happen, though not exactly how, or when. We’ve been able to move further down because of new materials, and clever design solutions. But we’ve pretty much run out of room. Shortly, lines etched on silicon will be not more than about 15-20 atoms wide. At that point, quantum tunneling becomes a major problem. When that happens, there is nothing further we can do.
it’s not that far off. For a number of years chip experts have been saying that below 7nm, things become very iffy. While manufacturers have been talking about 5nm, and in one case, even 3nm. Moving below 7nm is going to be extremely difficult. For one thing, techniques that have worked since 22nm will not work below 7nm, and new techniques will be needed. While 3 have been identified, so far, none of them are working, or working well. They’ve got some time yet, but if they don’t get any them working in the next couple of years, there will be trouble. Also, EUV, which is needed to go down there isn’t working yet, despite being worked on for almost 15 years. What if it doesn’t?
I have been hearing that same line "It appears that silicon is reaching limits, imposed by the laws of physics." since the mid-1990's. Yet, mysteriously year after year; we see performance improvements in the 30-50% range. Those improvements do not sound anything like hitting a limit, improvements in the 2-5% range sound like you are banging against a limit.
You'll be seeing the 30-50%* for years to come on Apple ICs. Two reasons: 1) Everything owned by Apple so can be optimized for 2) a single line of hardware.
* The only reason not to see that is if they decide to work on lowering TDP (by extension - improving battery life) in an iteration.
Don’t be so sure. The performance increase of Apple’s chips have slowing down. From 100% to 60% to 50% to 40%, and now to 25%. While there may simply design choices for that, recently, we don’t know.
Somebody at Appleinsider shoukd have grabbed a pair of pliers to peel back fingernails to get these guys to cough up more information on this SoC. Specifically info on the GPU and Neural engine. The article implies that the Nueral engine is completely separste from the GPU, that is a bit of a surprise. These two features being all Apple are very interesting from a technical standpoint.
Why should Apple have to detail any of that? Companies who sell chips, and have differing lineS, and also competitors, need to detail their offerings so that potential customers can make a decision, and hopefully, buy theirs, whoever they are.
but Apple doesn’t sell its designs. And if we think that they are ahead, why should they detail enough to give competitors more information? As it is, Apple’s SoC has, historically, had between 30-40% of the area on the die not understood by those studying it, neither from microphotographs or analysis. That’s a lot of secret info where almost every square mm of other chips is understood. It’s enough for them to give us broad ideas of what they’ve done, and how they’re using it in actual software and hardware.
i get that we want to know more, but Apple shouldn’t have to reveal it just because we want to know.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
True, but considering Apple is one of the original parents of the ARM lineage, I'm willing to let it slide.
I have been hearing that same line "It appears that silicon is reaching limits, imposed by the laws of physics." since the mid-1990's. Yet, mysteriously year after year; we see performance improvements in the 30-50% range. Those improvements do not sound anything like hitting a limit, improvements in the 2-5% range sound like you are banging against a limit.
You'll be seeing the 30-50%* for years to come on Apple ICs. Two reasons: 1) Everything owned by Apple so can be optimized for 2) a single line of hardware.
* The only reason not to see that is if they decide to work on lowering TDP (by extension - improving battery life) in an iteration.
Don’t be so sure. The performance increase of Apple’s chips have slowing down. From 100% to 60% to 50% to 40%, and now to 25%. While there may simply design choices for that, recently, we don’t know.
Not the numbers I get. Using Geekbench single core results I get:
A6 to A7 - 67% A7 to A8 - 16% A8 to A9 - 62% A9 to A10 - 45% A10 to A11 - 20% (estimate based on leaked scores).
Seems like their increases are large with the new architecture, then small with refinement, then large again with a new architecture. They might be using the same basic architecture from the A9-A11, which is why we see two consecutive drops.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
Do you expect AMD to stop using x86 and send a "letter of intent" stating as much to Intel?
Apple has an ISA (instruction set architecture) license from ARM. They build their own custom micro-architecture (essentially the silicon) to run the ISA. I don't see any reason for Apple to create their own ISA just so they can control 100% in-house. If Apple was going to do this I would have expected it to happen back with the A7 when they went to 100% in-house designed cores. Now they have so much invested into ARMv8 that I doubt they'd try and create their own.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
True, but considering Apple is one of the original parents of the ARM lineage, I'm willing to let it slide.
Not any more, they aren’t. And if controlling their future means relying on technology that only they have, then ARM is holding them back from that. Maybe they should buy ARM.
I have been hearing that same line "It appears that silicon is reaching limits, imposed by the laws of physics." since the mid-1990's. Yet, mysteriously year after year; we see performance improvements in the 30-50% range. Those improvements do not sound anything like hitting a limit, improvements in the 2-5% range sound like you are banging against a limit.
You'll be seeing the 30-50%* for years to come on Apple ICs. Two reasons: 1) Everything owned by Apple so can be optimized for 2) a single line of hardware.
* The only reason not to see that is if they decide to work on lowering TDP (by extension - improving battery life) in an iteration.
Don’t be so sure. The performance increase of Apple’s chips have slowing down. From 100% to 60% to 50% to 40%, and now to 25%. While there may simply design choices for that, recently, we don’t know.
Not the numbers I get. Using Geekbench single core results I get:
A6 to A7 - 67% A7 to A8 - 16% A8 to A9 - 62% A9 to A10 - 45% A10 to A11 - 20% (estimate based on leaked scores).
Seems like their increases are large with the new architecture, then small with refinement, then large again with a new architecture. They might be using the same basic architecture from the A9-A11, which is why we see two consecutive drops.
That’s pretty much what I said. You’re leaving off the 4 to 5 and the 5 To 6. The A7 To A8 was because of consolidation of the chip design after the first 64 bit version the year before.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
Do you expect AMD to stop using x86 and send a "letter of intent" stating as much to Intel?
Apple has an ISA (instruction set architecture) license from ARM. They build their own custom micro-architecture (essentially the silicon) to run the ISA. I don't see any reason for Apple to create their own ISA just so they can control 100% in-house. If Apple was going to do this I would have expected it to happen back with the A7 when they went to 100% in-house designed cores. Now they have so much invested into ARMv8 that I doubt they'd try and create their own.
This has nothing to do with AMD, and I don’t know why you brought it up. AMD sells x86 chips. That’s their business. If they gave up the license, they would be a GPU company. Apple’s business is not ARM chips. That a component in their iOS devices. If Apple wanted to, they could expand on their own designs, and eventually drop ARM, the way they did with Imagination.
the A7 cores were based on ARM’s 64 bit cores intended for low power servers.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
Do you expect AMD to stop using x86 and send a "letter of intent" stating as much to Intel?
Apple has an ISA (instruction set architecture) license from ARM. They build their own custom micro-architecture (essentially the silicon) to run the ISA. I don't see any reason for Apple to create their own ISA just so they can control 100% in-house. If Apple was going to do this I would have expected it to happen back with the A7 when they went to 100% in-house designed cores. Now they have so much invested into ARMv8 that I doubt they'd try and create their own.
This has nothing to do with AMD, and I don’t know why you brought it up. AMD sells x86 chips. That’s their business. If they gave up the license, they would be a GPU company. Apple’s business is not ARM chips. That a component in their iOS devices. If Apple wanted to, they could expand on their own designs, and eventually drop ARM, the way they did with Imagination.
the A7 cores were based on ARM’s 64 bit cores intended for low power servers.
Apple doesn't need to worry about ARM per se; everybody using ARM, and that's almost everyone in mobile, would be in the same boat, but worse off than Apple. More to the point, and you and I have spoken of this, even if the next nodes are difficult or impossible to reach, Apple still has the advantage of designing more function into a larger die, while at the same time, being able to afford to incorporate that die into their SOC's.
I would like to say here that Apple isn’t controlling the entire stack while they’re still using the architectural license from ARM. That’s what they were doing with Imagination. I would say that the ARM license is of greater import.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
True, but considering Apple is one of the original parents of the ARM lineage, I'm willing to let it slide.
Not any more, they aren’t. And if controlling their future means relying on technology that only they have, then ARM is holding them back from that. Maybe they should buy ARM.
"Maybe they should buy ARM."
Now there's a move that would seriously ruffle feathers and send shock waves within the industry. Good thing is, now Apple has they money to do it.
The A11 is the one thing the competition (and the haters) do NOT want to talk about. It's a sore spot with them how far ahead Apple is. Even next years 8990 & 845 (or whatever they call them) won't catch up to the A11.
Hell, Samsung and Qualcomm are barely at 2,000 single core while Apple is at 4,000. It'll be years before they're even close.
Don't dismiss what Samsung or Qualcomm may accomplish with a processor design. The only problem these companies have is that they need to focus on a wide array of processor performance levels unlike Apple which needs only to release one high performance processor a year letting older processors take up the slack at the low end. It amounts to a resource problem as they need to allocate talent to a broad array of customers. I'm pretty much convinced that Qualcomm at least could very well produce a high performance chip if it could dedicate all the resources needed to such a chip, they have a good track record of doing better that ARM in their designs.
Also consider that Apple no longer needs to allocate space on the processor for 32-bit processing.
I have been hearing that same line "It appears that silicon is reaching limits, imposed by the laws of physics." since the mid-1990's. Yet, mysteriously year after year; we see performance improvements in the 30-50% range. Those improvements do not sound anything like hitting a limit, improvements in the 2-5% range sound like you are banging against a limit.
You'll be seeing the 30-50%* for years to come on Apple ICs. Two reasons: 1) Everything owned by Apple so can be optimized for 2) a single line of hardware.
* The only reason not to see that is if they decide to work on lowering TDP (by extension - improving battery life) in an iteration.
Don’t be so sure. The performance increase of Apple’s chips have slowing down. From 100% to 60% to 50% to 40%, and now to 25%. While there may simply design choices for that, recently, we don’t know.
Not the numbers I get. Using Geekbench single core results I get:
A6 to A7 - 67% A7 to A8 - 16% A8 to A9 - 62% A9 to A10 - 45% A10 to A11 - 20% (estimate based on leaked scores).
Seems like their increases are large with the new architecture, then small with refinement, then large again with a new architecture. They might be using the same basic architecture from the A9-A11, which is why we see two consecutive drops.
That’s pretty much what I said. You’re leaving off the 4 to 5 and the 5 To 6. The A7 To A8 was because of consolidation of the chip design after the first 64 bit version the year before.
I think the substantial amount of coprocessing added in this iteration on the SOC (and thermal implications) is why the relatively low IPC jump. Apple by being able/afford to build huge SOC's can really go crazy there in ways nobody can.
Others may eventually come "closer" to Apple's IPC scores but by that time Apple will have a huge multilayed SOCs built in hundreds of millions of devices with tens of coprocessing functions with controllers tightly interwoven with their OS that control the traffic.
That's one hell of a tech moat to bridge for those not vertically integrated, only Samsung has the economic and tech potential to be able to follow but they'll be hampered severely by the fact they don't control the OS (and they kinda suck at software).
I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system. Separate laptops, iPads and Apple TV will be obsolete.
The only thing that could hamper this evolution would be the decrease in number of items people would need to buy, i.e. loss of sales. However, the mobile device will be more premium and worth a higher price, and the secondary devices like stand-alone screens, Apple Watch and wireless earphones, will be the same. Furthermore, I think this evolution is a reason that Apple is investing heavily in content.
So all in all, the future would be:
A phone with processing power and storage space enough to replace not only our current phones, but also our iPads, laptops, "home computers" and Apple TV
Separate screens to connect wirelessly for when we need larger screen estate or to run full-scale OS. In most households one 10-15" to replace the iPad and one 30" and above to replace the home computer and the TV.
An Apple Watch or similar for the most mobile situations.
Another option, of course, is sharing processors and storage space at a centralized location like a cloud service, and all devices consisting of simply a screen and processing and signal transmission enough to communicate with these. However, speed and privacy would be an issue. The benefit, of course, would be sharing processors and storage space so that none is redundant. Cheaper and more efficient. Plus, no need for backups as this would also be automatic and in much safer hands than our homes – if we can trust the service provider.
I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system. Separate laptops, iPads and Apple TV will be obsolete.
The only thing that could hamper this evolution would be the decrease in number of items people would need to buy, i.e. loss of sales. However, the mobile device will be more premium and worth a higher price, and the secondary devices like stand-alone screens, Apple Watch and wireless earphones, will be the same. Furthermore, I think this evolution is a reason that Apple is investing heavily in content.
So all in all, the future would be:
A phone with processing power and storage space enough to replace not only our current phones, but also our iPads, laptops, "home computers" and Apple TV
Separate screens to connect wirelessly for when we need larger screen estate or to run full-scale OS. In most households one 10-15" to replace the iPad and one 30" and above to replace the home computer and the TV.
An Apple Watch or similar for the most mobile situations.
Another option, of course, is sharing processors and storage space at a centralized location like a cloud service, and all devices consisting of simply a screen and processing and signal transmission enough to communicate with these. However, speed and privacy would be an issue. The benefit, of course, would be sharing processors and storage space so that none is redundant. Cheaper and more efficient. Plus, no need for backups as this would also be automatic and in much safer hands than our homes – if we can trust the service provider.
"I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system"
iPhones are already powerful enough to do that. Question is, will Apple implement the technology within iOS to allow that to happen? Samsung has already started going down that path with DEX.
I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system. Separate laptops, iPads and Apple TV will be obsolete.
The only thing that could hamper this evolution would be the decrease in number of items people would need to buy, i.e. loss of sales. However, the mobile device will be more premium and worth a higher price, and the secondary devices like stand-alone screens, Apple Watch and wireless earphones, will be the same. Furthermore, I think this evolution is a reason that Apple is investing heavily in content.
So all in all, the future would be:
A phone with processing power and storage space enough to replace not only our current phones, but also our iPads, laptops, "home computers" and Apple TV
Separate screens to connect wirelessly for when we need larger screen estate or to run full-scale OS. In most households one 10-15" to replace the iPad and one 30" and above to replace the home computer and the TV.
An Apple Watch or similar for the most mobile situations.
Another option, of course, is sharing processors and storage space at a centralized location like a cloud service, and all devices consisting of simply a screen and processing and signal transmission enough to communicate with these. However, speed and privacy would be an issue. The benefit, of course, would be sharing processors and storage space so that none is redundant. Cheaper and more efficient. Plus, no need for backups as this would also be automatic and in much safer hands than our homes – if we can trust the service provider.
"I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system"
iPhones are already powerful enough to do that. Question is, will Apple implement the technology within iOS to allow that to happen? Samsung has already started going down that path with DEX.
With Continuity built deep into macOS, iOS, watchOS and tvOS we already use many devices as ONE. It is oxymoron to demand LTE to be everywhere and being stuck at the age of cabled/docked connection.
The question remains when will MacBook One do away with Intel entirely. 2020 is my rough estimate. More than anything, I’d be most interested in what battery life they could squeeze for web browsing / YouTube video watching.
They might never do this. When you compare the 12" iPad Pro vs 12" Intel Macbook:
Both 41Wh batteries, both rated at 10 hours wifi browsing. Both Intel and Apple have access to advanced chip fabrication so they will both develop over time.
They will both hit limits in what the fabrication will allow. They will likely use 10nm for 3 years and then 7nm for 3 years. I'd expect the industry to be no further than 7nm until 2022. They may have worked out 4nm by then, which will take them to 2025. By this point (8 years), I'd expect chips to be 8x better than they are now. This is overall, they tend to put the improvement into the internal GPU rather than CPU. So let's say 4x GPU, 2x CPU.
What's the upside in switching to their own chip? It's not like people are struggling with the performance, the demand for high performance is diminishing. When it comes to battery life, the biggest advances in the next 8 years should be in the batteries themselves. While solid state batteries may not improve capacity, they can dramatically improve charging speed to minutes.
I think it could go either way for their own CPU in Macs vs Intel. It's easy for the low-end but would Apple manufacture custom 18-core or more chips for the iMac Pro? It would certainly be a lot cheaper on the high-end so great for their pro line. If they wouldn't do this, would they have a split Mac line with both ARM and x86 and software for both? Buying Intel's chips makes the Mac so much easier for Apple to maintain.
The question remains when will MacBook One do away with Intel entirely. 2020 is my rough estimate. More than anything, I’d be most interested in what battery life they could squeeze for web browsing / YouTube video watching.
They might never do this. When you compare the 12" iPad Pro vs 12" Intel Macbook:
Both 41Wh batteries, both rated at 10 hours wifi browsing. Both Intel and Apple have access to advanced chip fabrication so they will both develop over time.
They will both hit limits in what the fabrication will allow. They will likely use 10nm for 3 years and then 7nm for 3 years. I'd expect the industry to be no further than 7nm until 2022. They may have worked out 4nm by then, which will take them to 2025. By this point (8 years), I'd expect chips to be 8x better than they are now. This is overall, they tend to put the improvement into the internal GPU rather than CPU. So let's say 4x GPU, 2x CPU.
What's the upside in switching to their own chip? It's not like people are struggling with the performance, the demand for high performance is diminishing. When it comes to battery life, the biggest advances in the next 8 years should be in the batteries themselves. While solid state batteries may not improve capacity, they can dramatically improve charging speed to minutes.
I think it could go either way for their own CPU in Macs vs Intel. It's easy for the low-end but would Apple manufacture custom 18-core or more chips for the iMac Pro? It would certainly be a lot cheaper on the high-end so great for their pro line. If they wouldn't do this, would they have a split Mac line with both ARM and x86 and software for both? Buying Intel's chips makes the Mac so much easier for Apple to maintain.
Not sure what will happen when and if MS delivers an ARM powered Surface, again; will it even matter? Would it even make sense for Apple to create a boot system for Windows on A series, even if Apple makes an ARM analog to the Mac Book or Mac Book Air?
I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system. Separate laptops, iPads and Apple TV will be obsolete.
The only thing that could hamper this evolution would be the decrease in number of items people would need to buy, i.e. loss of sales. However, the mobile device will be more premium and worth a higher price, and the secondary devices like stand-alone screens, Apple Watch and wireless earphones, will be the same. Furthermore, I think this evolution is a reason that Apple is investing heavily in content.
So all in all, the future would be:
A phone with processing power and storage space enough to replace not only our current phones, but also our iPads, laptops, "home computers" and Apple TV
Separate screens to connect wirelessly for when we need larger screen estate or to run full-scale OS. In most households one 10-15" to replace the iPad and one 30" and above to replace the home computer and the TV.
An Apple Watch or similar for the most mobile situations.
Another option, of course, is sharing processors and storage space at a centralized location like a cloud service, and all devices consisting of simply a screen and processing and signal transmission enough to communicate with these. However, speed and privacy would be an issue. The benefit, of course, would be sharing processors and storage space so that none is redundant. Cheaper and more efficient. Plus, no need for backups as this would also be automatic and in much safer hands than our homes – if we can trust the service provider.
"I guess that in just a few years, cellphones will be powerful enough to be our main computers in life and by simply connecting a stand-alone screen and a keyboard, it converts into running a full-scale OS system"
iPhones are already powerful enough to do that. Question is, will Apple implement the technology within iOS to allow that to happen? Samsung has already started going down that path with DEX.
With Continuity built deep into macOS, iOS, watchOS and tvOS we already use many devices as ONE. It is oxymoron to demand LTE to be everywhere and being stuck at the age of cabled/docked connection.
As much as we live in a mobile-first world, doesn't mean we live in a mobile-only world. Most mainstream users don't need anymore performance and / or features than their smartphone provides. But sometimes when they're at home & their office they want a big-screen experience. I think it's a great idea if you can just dock your smartphone and have a desktop experience when you need it.
Comments
it’s not that far off. For a number of years chip experts have been saying that below 7nm, things become very iffy. While manufacturers have been talking about 5nm, and in one case, even 3nm. Moving below 7nm is going to be extremely difficult. For one thing, techniques that have worked since 22nm will not work below 7nm, and new techniques will be needed. While 3 have been identified, so far, none of them are working, or working well. They’ve got some time yet, but if they don’t get any them working in the next couple of years, there will be trouble. Also, EUV, which is needed to go down there isn’t working yet, despite being worked on for almost 15 years. What if it doesn’t?
Don’t be so sure. The performance increase of Apple’s chips have slowing down. From 100% to 60% to 50% to 40%, and now to 25%. While there may simply design choices for that, recently, we don’t know.
but Apple doesn’t sell its designs. And if we think that they are ahead, why should they detail enough to give competitors more information? As it is, Apple’s SoC has, historically, had between 30-40% of the area on the die not understood by those studying it, neither from microphotographs or analysis. That’s a lot of secret info where almost every square mm of other chips is understood. It’s enough for them to give us broad ideas of what they’ve done, and how they’re using it in actual software and hardware.
i get that we want to know more, but Apple shouldn’t have to reveal it just because we want to know.
so, will we see, at some point, Apple also sending ARM a letter of intent about dropping their license at some time?
because until they do that, they can’t truly say that it’s all theirs.
Not the numbers I get. Using Geekbench single core results I get:
A6 to A7 - 67%
A7 to A8 - 16%
A8 to A9 - 62%
A9 to A10 - 45%
A10 to A11 - 20% (estimate based on leaked scores).
Seems like their increases are large with the new architecture, then small with refinement, then large again with a new architecture. They might be using the same basic architecture from the A9-A11, which is why we see two consecutive drops.
Do you expect AMD to stop using x86 and send a "letter of intent" stating as much to Intel?
Apple has an ISA (instruction set architecture) license from ARM. They build their own custom micro-architecture (essentially the silicon) to run the ISA. I don't see any reason for Apple to create their own ISA just so they can control 100% in-house. If Apple was going to do this I would have expected it to happen back with the A7 when they went to 100% in-house designed cores. Now they have so much invested into ARMv8 that I doubt they'd try and create their own.
That’s pretty much what I said. You’re leaving off the 4 to 5 and the 5 To 6. The A7 To A8 was because of consolidation of the chip design after the first 64 bit version the year before.
This has nothing to do with AMD, and I don’t know why you brought it up. AMD sells x86 chips. That’s their business. If they gave up the license, they would be a GPU company. Apple’s business is not ARM chips. That a component in their iOS devices. If Apple wanted to, they could expand on their own designs, and eventually drop ARM, the way they did with Imagination.
the A7 cores were based on ARM’s 64 bit cores intended for low power servers.
Now there's a move that would seriously ruffle feathers and send shock waves within the industry. Good thing is, now Apple has they money to do it.
Apple by being able/afford to build huge SOC's can really go crazy there in ways nobody can.
Others may eventually come "closer" to Apple's IPC scores but by that time Apple will have a huge multilayed SOCs built in hundreds of millions of devices with tens of coprocessing functions with controllers tightly interwoven with their OS that control the traffic.
That's one hell of a tech moat to bridge for those not vertically integrated, only Samsung has the economic and tech potential to be able to follow but they'll be hampered severely by the fact they don't control the OS (and they kinda suck at software).
The only thing that could hamper this evolution would be the decrease in number of items people would need to buy, i.e. loss of sales. However, the mobile device will be more premium and worth a higher price, and the secondary devices like stand-alone screens, Apple Watch and wireless earphones, will be the same. Furthermore, I think this evolution is a reason that Apple is investing heavily in content.
So all in all, the future would be:
Another option, of course, is sharing processors and storage space at a centralized location like a cloud service, and all devices consisting of simply a screen and processing and signal transmission enough to communicate with these. However, speed and privacy would be an issue. The benefit, of course, would be sharing processors and storage space so that none is redundant. Cheaper and more efficient. Plus, no need for backups as this would also be automatic and in much safer hands than our homes – if we can trust the service provider.
iPhones are already powerful enough to do that. Question is, will Apple implement the technology within iOS to allow that to happen? Samsung has already started going down that path with DEX.
https://www.apple.com/ipad-pro/specs/
https://www.apple.com/macbook/specs/
Both 41Wh batteries, both rated at 10 hours wifi browsing. Both Intel and Apple have access to advanced chip fabrication so they will both develop over time.
They will both hit limits in what the fabrication will allow. They will likely use 10nm for 3 years and then 7nm for 3 years. I'd expect the industry to be no further than 7nm until 2022. They may have worked out 4nm by then, which will take them to 2025. By this point (8 years), I'd expect chips to be 8x better than they are now. This is overall, they tend to put the improvement into the internal GPU rather than CPU. So let's say 4x GPU, 2x CPU.
What's the upside in switching to their own chip? It's not like people are struggling with the performance, the demand for high performance is diminishing. When it comes to battery life, the biggest advances in the next 8 years should be in the batteries themselves. While solid state batteries may not improve capacity, they can dramatically improve charging speed to minutes.
I think it could go either way for their own CPU in Macs vs Intel. It's easy for the low-end but would Apple manufacture custom 18-core or more chips for the iMac Pro? It would certainly be a lot cheaper on the high-end so great for their pro line. If they wouldn't do this, would they have a split Mac line with both ARM and x86 and software for both? Buying Intel's chips makes the Mac so much easier for Apple to maintain.