I wonder how this chip compares to the Core M chip in the MacBook and the processors in the MacBookAir. Probably beats the Core M hands down and runs neck and next with the MBA chips.. Just guessing but I think Apple could put this in at least the MacBook.
I wonder how this chip compares to the Core M chip in the MacBook and the processors in the MacBookAir. Probably beats the Core M hands down and runs neck and next with the MBA chips.. Just guessing but I think Apple could put this in at least the MacBook.
Think bigger. Also consider that these chips aren't used with the same power envelopes and the MBPs have fans for cooling.
Apple's A10X is a 10nm process while Intel is still at 14nm with Kaby Lake in the new MBPs.
Even with the smaller npde process the A-series chips usually have grown in size.
Just imagine what Apple could offer if they designed a chip that had the power envelop and TDP of the Intel chips they use in their notebooks, without all the wasted features of Intel's offerings and at a fraction of the price.
What TSMC calls 10nm and what Intel calls 10nm are not the same thing. TSMC's 10nm is probably somewhere in between Intel's 14nm and Intel's 10nm.
But that's a small point.
Apple's windstorm cores are amazing. I have a new iPad Pro 12.9" and it really is super speedy. But while I really do like it, I wish my Mac could benefit from Apple's CPU design team.
What matters here is the differences between TSMCs 16nm and 10nm process modes, not what Intel, or others are doing. We compare Apple’s line to Apple’s line, after all.
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
It takes a while to download and cache enough 3D map data...
But, but when I got it working... Shit, Oh dear! I haven't felt this way since I saw my first Apple ][ -- 39 years ago.
It isn't what it is, today -- it is the future! *
Dick,
Yours is the opinion I value most on this forum, and your enthusiasm above -- in the context of your lifetime of IT experience -- is very exciting for the future we're all on the cusp of entering.
In my lifetime, I've met more than a few people who acted like the world revolved around them...
Now it does -- for anyone with a newer iPhone or iPad...
I entered Flyover for San Jose... I spun and rotated the world around me to see my old home in Saratoga... then took a few steps to get there (could've swiped)... then, I spun again to rotate Cupertino into view... took a few steps and I was at Apple's new Campus...
Now, I'm in Rome, then Madrid...
Wherever I go -- the world revolves around ME!
Now, how do I take a selfie to prove it???
Well Dick, with Ar you could put yourself into a street view sort of situation. Then just show yourself walking around.
This is certainly a huge advancemment, effectively Apple beat everybody to a sub 14nm process.
There's several mobile devices using a "sub-14nm process", primarily with the Qualcomm 835 10nm chip, like the Samsung S8/S8 Plus, Xiaomi Mi6, HTC U11, OnePlus5 and others. There's also Samsung's 10nm chipset, the Exynos 8895, but that's only used by Sammy themselves at the moment AFAIK.
Apple's A10X is the first 10nm from TSMC but not from the industry. It may of course be the better chip tho, no argument from me.
Yes, this isn’t the first 10nm chip. Whatever a process node means these days anyway. I agree with intel and AMD that process nodes mean less, and that perhaps we should be talking about density instead.
but there is no doubt that the A10x and even the A10 is a better chip than all the rest you mentioned. We already see that with device testing on arstechnica and anandtech. Not even close. The A11 will pull further away.
This is certainly a huge advancemment, effectively Apple beat everybody to a sub 14nm process. Combine that with the demonstrated performance in the new iPads, it becomes obvious that Apple has a huge lead here.
Now we just have to imagine what the new iPhone processor will be like. Im strongly leaning towards Apples new GPU showing up in the chip. We might even get the same CPU cores as the A10X. In otherwords i see A10X fore telling a big jump in iPhone SOC tech.
Yes! Could it be that the A10X Fusion and the A11 are variations on a theme -- maybe a semi-annual tick/tock cycle?
I'm thinking the same basic CPU units as in A10X but with Apples new GPU instead of the imagination derived units. That new GPOU would be highly optimized for handling both graphics and Apples AR and ML initiatives.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
The initiatives Apple has discussed this year at WWDC can all benefit from hardware optimized to excite the algorithms. There is a huge mystery as to what Apples new SOC will offer as a solution to accelerating ML, AR and other advanced concepts. I'm still of the opinion that they will do this all on one chip and most likely as an adjunct to the GPU functionality. One the other hand they could add a dedicated processor for AI type functionality. Blending the functionality with the GPU makes lots of sense though as there is so much overlap in functionality required.
Frankly it is kinda exciting to wait for What Apple has coming. I'm reasonably certain Apples goal is to tailor the designs for mobile use though they might add an AI coprocessor to Macs.
What matters here is the differences between TSMCs 16nm and 10nm process modes, not what Intel, or others are doing. We compare Apple’s line to Apple’s line, after all.
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
The initiatives Apple has discussed this year at WWDC can all benefit from hardware optimized to excite the algorithms. There is a huge mystery as to what Apples new SOC will offer as a solution to accelerating ML, AR and other advanced concepts. I'm still of the opinion that they will do this all on one chip and most likely as an adjunct to the GPU functionality. One the other hand they could add a dedicated processor for AI type functionality. Blending the functionality with the GPU makes lots of sense though as there is so much overlap in functionality required.
Frankly it is kinda exciting to wait for What Apple has coming. I'm reasonably certain Apples goal is to tailor the designs for mobile use though they might add an AI coprocessor to Macs.
I'm not a hardware guy...
I realize the GPU cores are critical for delivering AI, AR, ML...
Though, It seems to me, to make this all work on a mobile device * -- that fast access to iDevice storage/data is even more critical!
Take the Maps AR capabilities... to make this work acceptably, the iDevice would need to intelligently download and cache large amounts of 3D Maps data -- so you could take an AR walk around London, Paris (or Saint Petersburg ) without needing to wait for a constant refresh of data over sloooow cell/WiFi radios. This is just a 21st-century version of planning a trip -- you gather (pre-download and cache) all the maps you need for your planned Grand Tour of Europe. But, if, on a whim, you decide to take lunch at a brasserie near the Paris Bourse -- you do an ad hoc mobile download of the needed additional Maps data. It has to be fast, flexible, seamless and fun! * Each day I come to realize that even an old fart like me is dependent on mobile -- with an occasional need for my iMac 27 5K. Mobile is like sex -- once you've tried it, you're hooked... Once they get FCPX, SketchUP and Xcode running on the iPad, I'm good to go!
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
I used to think the same.
But, why build a better Steam Locomotive when you have the capability to build a Bullet Train or Driverless Car?
In 5 years the X86 and all its Legacy apps will be obsolete baggage -- and those who depend on them will be supplanted by others more flexible. It's happening now!
By year end, users of powerful new iPads will likely have their personal rooms and backyards modeled. So visiting a flea market will be able to immediately see how this antique sideboard looks with this crazy artwork if you repaint the wall. No need to ever touch tape measures-- (or even a paint brush if you convince your spouse to keep it virtual).
Probably a reason Amazon is expanding into big ticket furniture like sofas. With AR, not as much need for big box showrooms.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
The initiatives Apple has discussed this year at WWDC can all benefit from hardware optimized to excite the algorithms. There is a huge mystery as to what Apples new SOC will offer as a solution to accelerating ML, AR and other advanced concepts. I'm still of the opinion that they will do this all on one chip and most likely as an adjunct to the GPU functionality. One the other hand they could add a dedicated processor for AI type functionality. Blending the functionality with the GPU makes lots of sense though as there is so much overlap in functionality required.
Frankly it is kinda exciting to wait for What Apple has coming. I'm reasonably certain Apples goal is to tailor the designs for mobile use though they might add an AI coprocessor to Macs.
The advantage of doing almost all the hardware, along with the software is dramatic. Apple controls their future the way no one else can. It’s interesting on how they can control the secrecy around the chips and the software. So we literally have no idea of what they’re doing, other than what they tell us. I keep looking through the App Store trying to find some new AR apps using the beta API, though I know I won’t find anything yet. There’s so much from older work out there though, and it still looks good.
but whatever Apple is working on in chips will have a great impact. The concept that they are going to add AR and VR to the GPU is just speculation though. They don’t have to go that way, though there are some advantages to it. It’s easier for upgrades in chip design to go with a separate chip. That way they aren’t constrained with silicon area restraints. The chip can be whatever it needs to be without having to compromise with the other elements on chip. There are so many complexities here that I see either way though.
What matters here is the differences between TSMCs 16nm and 10nm process modes, not what Intel, or others are doing. We compare Apple’s line to Apple’s line, after all.
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
Licensing. Back in the mid 1980’s, intel was having a lot of problems. We forget (or weren’t around to know) that intel wasn’t the powerhouse it now is. If IBM had chosen another chip, say the 6800 from Motorola, for example, intel might not be here today, and Motorola would still be the premium technology company it was at the time.
because IBM chose their chip, when intel was having monetary problems, IBM stepped in to help them. But, as a backup, they forced intel to license their chip to other manufacturers - just in case. But as the licenses ended, and intel became what it is today, they didn’t renew those licenses. Only AMD was able to keep their license.
there is no way that Apple could design an x86 compatible chip without a license from intel. And why would intel do that?
but, as I’ve been saying, here and there, about 80% of the slowdown in emulation between one chip family and another resides in just a couple of dozen instructions that need to be emulated in software, because no comparable instruction exists in the other chip family. That why software such as Virtual PC could only run Windows and its software at no better than about 20-30% of the speed of an x86 based system. But, it’s very possible that Apple could add some individual instructions to their ARM chips that duplicate those of x86. That’s because many of these instructions, by themselves, are not subject to license.
soo, if Apple had a chip with a much smaller die, oh, saying, going down from 144mm2 to 96mm2, then they could add those instructions to a chip that had a slightly larger die, say 110mm2. That would allow their chip to run 80% as fast as a comparable x86 chip, using an x86 os running x86 software. But if the ARM chip ran faster anyway, then there might be no slowdown at all, and in succeeeding generations, it could even run faster. So we know that the A10x runs faster than the power comparable M3 used in Apple’s Macbook, and the $799 Surface Pro. It also runs faster than an intel i3 ultra low power chip used in many notebooks, and even comes close to the i5 ultra low power chip used in Apple’s 2016 $1,500 Macbook Pro.
this could be done while running x86 software untouched. And could also run, if Apple allowed it, suitable ARM software (the problem being the lack of a touch Screen)
so what would happen, if having these few instructions, it could run faster than that i5? I would think that it could possibly replace the 6 watt M3 used in the Macbook, with a 5-6 watt Axxx chip - at a lower cost.
and just maybe, if in the future, as this was happening, Apple moves to intel for its chip manufacturing, giving intel an excuse to bless this idea.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
I used to think the same.
But, why build a better Steam Locomotive when you have the capability to build a Bullet Train or Driverless Car?
In 5 years the X86 and all its Legacy apps will be obsolete baggage -- and those who depend on them will be supplanted by others more flexible. It's happening now!
Ah Dick, dreaming again? This could happen, but even if it does, it won’t happen in five years. 10? Not likely. 20, possibly, but not certain. Further out, anything that can be done is possible.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
The initiatives Apple has discussed this year at WWDC can all benefit from hardware optimized to excite the algorithms. There is a huge mystery as to what Apples new SOC will offer as a solution to accelerating ML, AR and other advanced concepts. I'm still of the opinion that they will do this all on one chip and most likely as an adjunct to the GPU functionality. One the other hand they could add a dedicated processor for AI type functionality. Blending the functionality with the GPU makes lots of sense though as there is so much overlap in functionality required.
Frankly it is kinda exciting to wait for What Apple has coming. I'm reasonably certain Apples goal is to tailor the designs for mobile use though they might add an AI coprocessor to Macs.
The advantage of doing almost all the hardware, along with the software is dramatic. Apple controls their future the way no one else can. It’s interesting on how they can control the secrecy around the chips and the software. So we literally have no idea of what they’re doing, other than what they tell us. I keep looking through the App Store trying to find some new AR apps using the beta API, though I know I won’t find anything yet. There’s so much from older work out there though, and it still looks good.
but whatever Apple is working on in chips will have a great impact. The concept that they are going to add AR and VR to the GPU is just speculation though. They don’t have to go that way, though there are some advantages to it. It’s easier for upgrades in chip design to go with a separate chip. That way they aren’t constrained with silicon area restraints. The chip can be whatever it needs to be without having to compromise with the other elements on chip. There are so many complexities here that I see either way though.
My only thought here is that mobile needs power efficiency thus a likely all in one SoC. Given that automotive and desktop are free to go their own way.
What matters here is the differences between TSMCs 16nm and 10nm process modes, not what Intel, or others are doing. We compare Apple’s line to Apple’s line, after all.
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
Licensing. Back in the mid 1980’s, intel was having a lot of problems. We forget (or weren’t around to know) that intel wasn’t the powerhouse it now is. If IBM had chosen another chip, say the 6800 from Motorola, for example, intel might not be here today, and Motorola would still be the premium technology company it was at the time.
because IBM chose their chip, when intel was having monetary problems, IBM stepped in to help them. But, as a backup, they forced intel to license their chip to other manufacturers - just in case. But as the licenses ended, and intel became what it is today, they didn’t renew those licenses. Only AMD was able to keep their license.
there is no way that Apple could design an x86 compatible chip without a license from intel. And why would intel do that?
but, as I’ve been saying, here and there, about 80% of the slowdown in emulation between one chip family and another resides in just a couple of dozen instructions that need to be emulated in software, because no comparable instruction exists in the other chip family. That why software such as Virtual PC could only run Windows and its software at no better than about 20-30% of the speed of an x86 based system. But, it’s very possible that Apple could add some individual instructions to their ARM chips that duplicate those of x86. That’s because many of these instructions, by themselves, are not subject to license.
soo, if Apple had a chip with a much smaller die, oh, saying, going down from 144mm2 to 96mm2, then they could add those instructions to a chip that had a slightly larger die, say 110mm2. That would allow their chip to run 80% as fast as a comparable x86 chip, using an x86 os running x86 software. But if the ARM chip ran faster anyway, then there might be no slowdown at all, and in succeeeding generations, it could even run faster. So we know that the A10x runs faster than the power comparable M3 used in Apple’s Macbook, and the $799 Surface Pro. It also runs faster than an intel i3 ultra low power chip used in many notebooks, and even comes close to the i5 ultra low power chip used in Apple’s 2016 $1,500 Macbook Pro.
this could be done while running x86 software untouched. And could also run, if Apple allowed it, suitable ARM software (the problem being the lack of a touch Screen)
so what would happen, if having these few instructions, it could run faster than that i5? I would think that it could possibly replace the 6 watt M3 used in the Macbook, with a 5-6 watt Axxx chip - at a lower cost.
and just maybe, if in the future, as this was happening, Apple moves to intel for its chip manufacturing, giving intel an excuse to bless this idea.
anybody disagree, or agree?
The big problem with obsessing about i86 compatibility is that iPhone and iPad pretty much prove you dont need it. The high quality of Ipads software library further supports this idea. In a nut shell i86 no longer has a lock on the market.
It is apples portable OS that further sugggests that i86 isnt a big problem anymore. If Apple did go ARM in a Mac you can be pretty sure the OS and supplied apps will be 100% ARM and all 64 bit. Combine this with Apple developer incentives and would see pretty rapid third party support for Mac on ARM. In the end only legacy software would be a problem.
It takes a while to download and cache enough 3D map data...
But, but when I got it working... Shit, Oh dear! I haven't felt this way since I saw my first Apple ][ -- 39 years ago.
It isn't what it is, today -- it is the future! *
Dick,
Yours is the opinion I value most on this forum, and your enthusiasm above -- in the context of your lifetime of IT experience -- is very exciting for the future we're all on the cusp of entering.
In my lifetime, I've met more than a few people who acted like the world revolved around them...
Now it does -- for anyone with a newer iPhone or iPad...
I entered Flyover for San Jose... I spun and rotated the world around me to see my old home in Saratoga... then took a few steps to get there (could've swiped)... then, I spun again to rotate Cupertino into view... took a few steps and I was at Apple's new Campus...
Now, I'm in Rome, then Madrid...
Wherever I go -- the world revolves around ME!
Now, how do I take a selfie to prove it???
I would love to have an AR mode that would allow a temporal shift backwards at a particular location so that a person could look back in time. Not sure how you would obtain the data as there was little in the way of photogrammetry in use in most of the U.S. and it was used sparingly for topological mapping. There might be Landsat data going back a few decades that could be repurposed for AR.
If U.S. would declassify some spy satellite data could possibly extend that back another decade into the 1960s. These have polar orbits and-- though I'm not at all sure-- they possibly imaged some of continental United States when looking at Cuba. When switched to digital with KH-11 in the 1970s, likely more imagery and intelligence services would monitor such things as expected crop harvests.
Yet imagine rendering various cities throughout modern times, based on motion pictures, aerial shots, personal photography, etc. Crazy potential!
What matters here is the differences between TSMCs 16nm and 10nm process modes, not what Intel, or others are doing. We compare Apple’s line to Apple’s line, after all.
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
Licensing. Back in the mid 1980’s, intel was having a lot of problems. We forget (or weren’t around to know) that intel wasn’t the powerhouse it now is. If IBM had chosen another chip, say the 6800 from Motorola, for example, intel might not be here today, and Motorola would still be the premium technology company it was at the time.
because IBM chose their chip, when intel was having monetary problems, IBM stepped in to help them. But, as a backup, they forced intel to license their chip to other manufacturers - just in case. But as the licenses ended, and intel became what it is today, they didn’t renew those licenses. Only AMD was able to keep their license.
there is no way that Apple could design an x86 compatible chip without a license from intel. And why would intel do that?
but, as I’ve been saying, here and there, about 80% of the slowdown in emulation between one chip family and another resides in just a couple of dozen instructions that need to be emulated in software, because no comparable instruction exists in the other chip family. That why software such as Virtual PC could only run Windows and its software at no better than about 20-30% of the speed of an x86 based system. But, it’s very possible that Apple could add some individual instructions to their ARM chips that duplicate those of x86. That’s because many of these instructions, by themselves, are not subject to license.
soo, if Apple had a chip with a much smaller die, oh, saying, going down from 144mm2 to 96mm2, then they could add those instructions to a chip that had a slightly larger die, say 110mm2. That would allow their chip to run 80% as fast as a comparable x86 chip, using an x86 os running x86 software. But if the ARM chip ran faster anyway, then there might be no slowdown at all, and in succeeeding generations, it could even run faster. So we know that the A10x runs faster than the power comparable M3 used in Apple’s Macbook, and the $799 Surface Pro. It also runs faster than an intel i3 ultra low power chip used in many notebooks, and even comes close to the i5 ultra low power chip used in Apple’s 2016 $1,500 Macbook Pro.
this could be done while running x86 software untouched. And could also run, if Apple allowed it, suitable ARM software (the problem being the lack of a touch Screen)
so what would happen, if having these few instructions, it could run faster than that i5? I would think that it could possibly replace the 6 watt M3 used in the Macbook, with a 5-6 watt Axxx chip - at a lower cost.
and just maybe, if in the future, as this was happening, Apple moves to intel for its chip manufacturing, giving intel an excuse to bless this idea.
anybody disagree, or agree?
The big problem with obsessing about i86 compatibility is that iPhone and iPad pretty much prove you dont need it. The high quality of Ipads software library further supports this idea. In a nut shell i86 no longer has a lock on the market.
It is apples portable OS that further sugggests that i86 isnt a big problem anymore. If Apple did go ARM in a Mac you can be pretty sure the OS and supplied apps will be 100% ARM and all 64 bit. Combine this with Apple developer incentives and would see pretty rapid third party support for Mac on ARM. In the end only legacy software would be a problem.
I’m assuming the question he had was about a macOS x86 notebook. Apple has made it pretty clear that they aren’t going to abandon their Mac market. So, if they want to continue advancing that market on their own terms, what I keep proposing makes sense. Moving, over time, their Mac hardware to a modified ARM chip SoC would do that.
now, I can’t agree that if they went macOS on ARM it would be a good bet that all apps would be ARM. I really think that would be a bad idea. Remember how long it took to move large software projects to 64 bit on a Mac before? It took years. Even Apple took a couple of years to move their own software completely over.
i don’t really see Adobe, Microsoft and other large software houses being thrilled to do it once again. If Apple could allow them to leave it as it is, they would be much more successful with the move. I do think that over time, it’s possible that more of the code could move away from those few x86 instructions to ARM instructions. But, as it is right now, x86 is still a richer environment than ARM.
Comments
no other SoC manufacturer is doing what Apple is doing, after all, so it doesn’t matter what processes they are on, or which company’s factories produce their chips. The fact is that the designs of the SoC is more important than the differences between different companies processes.
apple has chosen to go a certain route, while others have chosen a different one. Apple is more performance oriented, while others are more marketing oriented. In the Asian markets, the number of cores is a major marketing push, because they believe that the more cores the better.
but I think that Apple has long range plans, plans we can see unfolding over the years. These plans require a very large amount of processing potential, much more so than what Android devices need. Apple plams years in advance, and are rarely diverted from those long range goals, though they may make a slight change in direction to accommodate competitive forces. They would be stupid if they didn’t.
but we saw derision when the A7 came out with 64 bit processing. But that allowed rapid encryption, up to 50 times the speed of 32 bit chips. That allowed Apple to come out with Touch ID, and the Secure Enclave. That couldn’t have happened with a 32 bit chip. Apple. Onto Jens to pi,e performance on at a much more rapid rate than other ARM manufacturers are doing, with their obsession with 8 cores. While I’m surprised that Apple again went to 3 cores, it has helped multiprocessing immensely.
whats really interesting to me with the significant diminishing of the chip size, is that there are some possible benefits Apple can accumulate for the next generation of chips from that. While we don’t yet know anything about the upcoming (in just another 2.5 months!) A11, we can now say with certainty that it will be a 10nm part, something we suspected.
with the smaller process size, it looks as though Apple has not completely taken advantage of the new process size. That makes sense, as intel doesn’t either with their tick and tock (and tock and tock)). So going from 144 to 96 is a huge shrink. Yet, they added a big and a low power core. This means that they have a lot of space to move up to if they want to. What if they kept the same die size? There would be a lot of room for many more transistors. So what if this machine learning chip we’re reading about won’t be a separate chip, but will be integrated on the SoC die? Maybe we’ll se that with the A11. That could be why a smaller die now.
Yes, this isn’t the first 10nm chip. Whatever a process node means these days anyway. I agree with intel and AMD that process nodes mean less, and that perhaps we should be talking about density instead.
but there is no doubt that the A10x and even the A10 is a better chip than all the rest you mentioned. We already see that with device testing on arstechnica and anandtech. Not even close. The A11 will pull further away.
Frankly it is kinda exciting to wait for What Apple has coming. I'm reasonably certain Apples goal is to tailor the designs for mobile use though they might add an AI coprocessor to Macs.
There's been lots of discussion over the years with the possibility that Apple would bring the Mac to this architecture instead of X86 once the performance was up to snuff, especially with laptops. My question is this....what's keeping Apple from making an x86 chip of their own based on their own technology? I mean, AMD has been doing this for years....heck the 64 bit instruction set is based on their technology, not Intel's. An Apple designed x86 chip might be the interesting game changer in this space if they can do equivalent or better performance with better power consumption.
Best iPad device ever!
I realize the GPU cores are critical for delivering AI, AR, ML...
Though, It seems to me, to make this all work on a mobile device * -- that fast access to iDevice storage/data is even more critical!
Take the Maps AR capabilities... to make this work acceptably, the iDevice would need to intelligently download and cache large amounts of 3D Maps data -- so you could take an AR walk around London, Paris (or Saint Petersburg ) without needing to wait for a constant refresh of data over sloooow cell/WiFi radios.
This is just a 21st-century version of planning a trip -- you gather (pre-download and cache) all the maps you need for your planned Grand Tour of Europe. But, if, on a whim, you decide to take lunch at a brasserie near the Paris Bourse -- you do an ad hoc mobile download of the needed additional Maps data.
It has to be fast, flexible, seamless and fun!
* Each day I come to realize that even an old fart like me is dependent on mobile -- with an occasional need for my iMac 27 5K. Mobile is like sex -- once you've tried it, you're hooked... Once they get FCPX, SketchUP and Xcode running on the iPad, I'm good to go!
I suppose that's better than "MS Windows is like sex — if you've using it, you're likely infected with something."
I used to think the same.
But, why build a better Steam Locomotive when you have the capability to build a Bullet Train or Driverless Car?
In 5 years the X86 and all its Legacy apps will be obsolete baggage -- and those who depend on them will be supplanted by others more flexible. It's happening now!
By year end, users of powerful new iPads will likely have their personal rooms and backyards modeled. So visiting a flea market will be able to immediately see how this antique sideboard looks with this crazy artwork if you repaint the wall. No need to ever touch tape measures-- (or even a paint brush if you convince your spouse to keep it virtual).
Probably a reason Amazon is expanding into big ticket furniture like sofas. With AR, not as much need for big box showrooms.
but whatever Apple is working on in chips will have a great impact. The concept that they are going to add AR and VR to the GPU is just speculation though. They don’t have to go that way, though there are some advantages to it. It’s easier for upgrades in chip design to go with a separate chip. That way they aren’t constrained with silicon area restraints. The chip can be whatever it needs to be without having to compromise with the other elements on chip. There are so many complexities here that I see either way though.
Licensing. Back in the mid 1980’s, intel was having a lot of problems. We forget (or weren’t around to know) that intel wasn’t the powerhouse it now is. If IBM had chosen another chip, say the 6800 from Motorola, for example, intel might not be here today, and Motorola would still be the premium technology company it was at the time.
because IBM chose their chip, when intel was having monetary problems, IBM stepped in to help them. But, as a backup, they forced intel to license their chip to other manufacturers - just in case. But as the licenses ended, and intel became what it is today, they didn’t renew those licenses. Only AMD was able to keep their license.
there is no way that Apple could design an x86 compatible chip without a license from intel. And why would intel do that?
but, as I’ve been saying, here and there, about 80% of the slowdown in emulation between one chip family and another resides in just a couple of dozen instructions that need to be emulated in software, because no comparable instruction exists in the other chip family. That why software such as Virtual PC could only run Windows and its software at no better than about 20-30% of the speed of an x86 based system. But, it’s very possible that Apple could add some individual instructions to their ARM chips that duplicate those of x86. That’s because many of these instructions, by themselves, are not subject to license.
soo, if Apple had a chip with a much smaller die, oh, saying, going down from 144mm2 to 96mm2, then they could add those instructions to a chip that had a slightly larger die, say 110mm2. That would allow their chip to run 80% as fast as a comparable x86 chip, using an x86 os running x86 software. But if the ARM chip ran faster anyway, then there might be no slowdown at all, and in succeeeding generations, it could even run faster. So we know that the A10x runs faster than the power comparable M3 used in Apple’s Macbook, and the $799 Surface Pro. It also runs faster than an intel i3 ultra low power chip used in many notebooks, and even comes close to the i5 ultra low power chip used in Apple’s 2016 $1,500 Macbook Pro.
this could be done while running x86 software untouched. And could also run, if Apple allowed it, suitable ARM software (the problem being the lack of a touch Screen)
so what would happen, if having these few instructions, it could run faster than that i5? I would think that it could possibly replace the 6 watt M3 used in the Macbook, with a 5-6 watt Axxx chip - at a lower cost.
and just maybe, if in the future, as this was happening, Apple moves to intel for its chip manufacturing, giving intel an excuse to bless this idea.
anybody disagree, or agree?
Ah Dick, dreaming again? This could happen, but even if it does, it won’t happen in five years. 10? Not likely. 20, possibly, but not certain. Further out, anything that can be done is possible.
It is apples portable OS that further sugggests that i86 isnt a big problem anymore. If Apple did go ARM in a Mac you can be pretty sure the OS and supplied apps will be 100% ARM and all 64 bit. Combine this with Apple developer incentives and would see pretty rapid third party support for Mac on ARM. In the end only legacy software would be a problem.
Yet imagine rendering various cities throughout modern times, based on motion pictures, aerial shots, personal photography, etc. Crazy potential!
now, I can’t agree that if they went macOS on ARM it would be a good bet that all apps would be ARM. I really think that would be a bad idea. Remember how long it took to move large software projects to 64 bit on a Mac before? It took years. Even Apple took a couple of years to move their own software completely over.
i don’t really see Adobe, Microsoft and other large software houses being thrilled to do it once again. If Apple could allow them to leave it as it is, they would be much more successful with the move. I do think that over time, it’s possible that more of the code could move away from those few x86 instructions to ARM instructions. But, as it is right now, x86 is still a richer environment than ARM.