AppleInsider › Forums › Mac Hardware › Future Apple Hardware › Prediction: Atom based iPad by 2016
New Posts  All Forums:Forum Nav:

Prediction: Atom based iPad by 2016

post #1 of 35
12/5/12 at 8:27am
Thread Starter 

There's been a lot of talk of the Macs going ARM but I'm going argue the opposite:  the iPad moves to x86.  With 22nm Atom next year and 14nm Atoms by 2014 I can see Intel winning the performance per watt battle over time between process node advantages and overall improvement in their designs.

 

If that happens...and Intel gets into the Tick/Tock pattern for mobile then ARM might end up on the wrong end of the power performance curve by 2016. 

 

Apple can move in either direction (Mac to ARM or iPad to X86).  It's just an XCode target and recompile for most app devs. 

 

Intel can certainly provide volume Apple needs and it would once again love to have Apple as a premiere partner.  I don't think that ARM can scale up performance wise as fast as Intel has been scaling down power requirements.

Reply
Reply
post #2 of 35
12/5/12 at 8:37am

No, they're not gonna put X86 chips in the iPad.

PhilBoogie
That's Google alright. For a stupid company they sure do dumb things.
Reply
PhilBoogie
That's Google alright. For a stupid company they sure do dumb things.
Reply
post #3 of 35
12/5/12 at 9:25am
Thread Starter 
Quote:
Originally Posted by Tallest Skil View Post

No, they're not gonna put X86 chips in the iPad.

 

It's more likely than putting ARM in Macs.

Reply
Reply
post #4 of 35
12/5/12 at 9:34am
Originally Posted by nht View Post
It's more likely than putting ARM in Macs.

 

One's going backward. The other is going forward. There's overlap now because of what ARM can't provide* that X86 can.

 

*Yet.

PhilBoogie
That's Google alright. For a stupid company they sure do dumb things.
Reply
PhilBoogie
That's Google alright. For a stupid company they sure do dumb things.
Reply
post #5 of 35
12/5/12 at 9:59am

why would Apple invest hundreds of millions of dollars in chip design for ARM chips, just to give it up to Intel in four years. Not going to happen.

Reply
Reply
post #6 of 35
12/5/12 at 10:37am
Thread Starter 
Quote:
Originally Posted by davion3052 View Post

why would Apple invest hundreds of millions of dollars in chip design for ARM chips, just to give it up to Intel in four years. Not going to happen.

 

Apple invested millions into the AIM alliance and the PowerPC as well.

 

Want to bet that there isn't a Medfield reference design phone running iOS somewhere in Cupertino?

 

The A6/A6X is pretty cool and if Apple goes the IDM route then moving to Intel is far less likely. If they don't then it's not such a big deal.

Reply
Reply
post #7 of 35
12/5/12 at 4:48pm
Quote:
Originally Posted by nht View Post

 

Apple invested millions into the AIM alliance and the PowerPC as well.

 

Want to bet that there isn't a Medfield reference design phone running iOS somewhere in Cupertino?

 

The A6/A6X is pretty cool and if Apple goes the IDM route then moving to Intel is far less likely. If they don't then it's not such a big deal.

Three things:

  1. The AIM Alliance and the PowerPC were nearly 20 years ago.
  2. Apple has many experimental projects in Cupertino. The vast majority of these projects never see the light of day.
  3. Apple processors including the A6/A6X are systems on a chip rather than discrete processors. Is it possible that Apple will revert to discrete components? Certainly. It is likely? Certainly not.
Reply
Reply
post #8 of 35
12/5/12 at 7:52pm
Quote:
Originally Posted by nht View Post

There's been a lot of talk of the Macs going ARM but I'm going argue the opposite:  the iPad moves to x86.  With 22nm Atom next year and 14nm Atoms by 2014 I can see Intel winning the performance per watt battle over time between process node advantages and overall improvement in their designs.
Not a chance. It really has nothing to do with performance, it is more about being the captain of your own ship. The reality is that unless Intel has a change of heart with respect to custom SoC they will have little success in mobile. Why? Pretty simple today's SoC is equivalent to the printed circuit boards back in the day when Wozniak and other engineers made leaps and bounds in clever circuit design. In effect mobile devices are built on SoC technologies, the
Is is where IP goes these days to make unique products.

Frankly I could argue that Intel will have no choice but to support customers with custom chips. It is the way of the future for these small devices.
Quote:
If that happens...and Intel gets into the Tick/Tock pattern for mobile then ARM might end up on the wrong end of the power performance curve by 2016. 
Doesn't matter. Seriously Apples move to Soc technology with ARM cores is all about where engineering is done these days.
Quote:
Apple can move in either direction (Mac to ARM or iPad to X86).  It's just an XCode target and recompile for most app devs. 
Thatis certainly true.
Quote:
Intel can certainly provide volume Apple needs and it would once again love to have Apple as a premiere partner.  I don't think that ARM can scale up performance wise as fast as Intel has been scaling down power requirements.

It isn't a simple matter of measuring processor core performance. In fact the processor core is just a small part of today's chips. You really need to think of that wafer of silicon as a replacement for the fiberglass printed circuit boards of the past. Instead of piecing together a system with a bunch of discrete chips and clever engineering to realize a machine it is rather an effort of merging IP and synthesized logic and application blocks to realize a device.

Honestly I think Intel has missed the boat here. Unless they open up ATOMs cores to custom implementations they will get no where in mobile.
Reply
Reply
post #9 of 35
12/5/12 at 8:44pm
  • hmm
  • Keylogger
Quote:
Originally Posted by Tallest Skil View Post

 

One's going backward. The other is going forward. There's overlap now because of what ARM can't provide* that X86 can.

 

*Yet.


I don't see it as so clear cut. I would suggest that if Apple was going to implement Atom, it would have happened a long time ago. They aren't so likely to change a successful strategy today. With PowerPC -> Intel, they were floundering. Notebooks were the largest potential growth market in the Mac lineup and they were stuck with the G4 process. They don't really have that kind of problem today.

Reply
Reply
post #10 of 35
12/6/12 at 4:42am
Thread Starter 
Quote:
Originally Posted by Mr. Me View Post

  2. Apple processors including the A6/A6X are systems on a chip rather than discrete processors. Is it possible that Apple will revert to discrete components? Certainly. It is likely? Certainly not.

 

The Intel Medfield is a SoC with the CPU/GPU/other stuff on die and DRAM stacked on the package.  Intel's Core chips a pseudo-SOCs these days with the GPU and Northbridge on die.

Reply
Reply
post #11 of 35
12/6/12 at 5:15am
Thread Starter 
Quote:
Originally Posted by wizard69 View Post

Not a chance. It really has nothing to do with performance, it is more about being the captain of your own ship. The reality is that unless Intel has a change of heart with respect to custom SoC they will have little success in mobile. Why? Pretty simple today's SoC is equivalent to the printed circuit boards back in the day when Wozniak and other engineers made leaps and bounds in clever circuit design. In effect mobile devices are built on SoC technologies, that is where IP goes these days to make unique products.

Frankly I could argue that Intel will have no choice but to support customers with custom chips. It is the way of the future for these small devices.

 

Then a reporter asked if Intel customizes parts for its largest customers. “Yes,” Bryant said. “We want to give them a way of differentiating their machines.”

 

Bryant didn’t say much more, but those few words shine a light on another part of the big-time chip business that’s rarely discussed. There are cases where a large chipmaker such as Intel and AMD will provide certain customers with chips that others may not have access to. Sometimes, this merely means that when the chipmaker cranks out a big batch of processors, one customer gets the chips that happen to have the best speed or power ratings. But in other cases, the chipmaker will actually modify processors at the request of a particular customer.

 

http://www.wired.com/wiredenterprise/2012/09/intel-amd-custom-chips/all/

 

 

Intel produced a custom Merom for the first MBA.  If Apple wants some secret sauce in its SoC then Intel probably would do it.  The only exception is that Intel probably really really wants to push it's Gen7 GPU (Ivy Bridge HD Graphics with 4 cores) in their 22nm SOCs over the SGX.

 

That's faster than today's top end SGX554MP4 but we're also looking a year or two from now.

 

Quote:
Doesn't matter. Seriously Apples move to Soc technology with ARM cores is all about where engineering is done these days.

 

There's just as much engineering being done to the Intel Silvermont and Airmont cores and if Apple wants something special it probably has the mojo to get it given the volumes it needs AND the cachet of its branding.  An Apple mobile design win for Intel means a lot more than the desktop/laptop design win did.  That was a nice to have.  This would be a major change in the mobile processor playing field.

 

 

Quote:
It isn't a simple matter of measuring processor core performance. In fact the processor core is just a small part of today's chips. You really need to think of that wafer of silicon as a replacement for the fiberglass printed circuit boards of the past. Instead of piecing together a system with a bunch of discrete chips and clever engineering to realize a machine it is rather an effort of merging IP and synthesized logic and application blocks to realize a device.
Honestly I think Intel has missed the boat here. Unless they open up ATOMs cores to custom implementations they will get no where in mobile.

 

You guys don't seem to realize that Medfield and all the follow on smartphone/tablet Atoms are SOCs.  In this generation with SGX544MP2 GPUs on the tablets and SGX 545 on netbooks and SGX540 on phones.  If Apple had said "we'll use the Saltwell processor for the iPad 4 but we need SGX554MP4 instead and these three things removed from the die and these two added you really don't think that Intel would have said "Hell yeah!".

 

Frankly Saltwell isn't fast enough for it to be a compelling move for Apple.  I do see 14nm Airmont in 2014 being compelling.  If the cost is moving to Gen8/9 Intel GPUs (the Haswell ones on 14nm) that's probably okay.

Reply
Reply
post #12 of 35
12/6/12 at 5:19am
Thread Starter 
Quote:
Originally Posted by hmm View Post


I don't see it as so clear cut. I would suggest that if Apple was going to implement Atom, it would have happened a long time ago. They aren't so likely to change a successful strategy today. With PowerPC -> Intel, they were floundering. Notebooks were the largest potential growth market in the Mac lineup and they were stuck with the G4 process. They don't really have that kind of problem today.

 

Intel Atom has sucked until Medfield.  It simply was never a viable option for Apple.

 

With 2012 Medfield they have parity with A9s so they are a half step behind.  (Why do that?)

With 2013 22nm (Valleyview, etc) I expect parity with A15s.  (Okay,  but the A6 is already here since Apple is slightly ahead of the pack so they are still a half step behind Apple)

with 2014 14nm I expect them to be ahead of ARM on both power and performance in 2014. (Now you have a reason to think about switching horses)

 

Which is why I hedged and said 2016 iPads.

Reply
Reply
post #13 of 35
12/6/12 at 5:28am
Thread Starter 

Regardless of whether you think I'm a crackpot to suggest an Atom powered iPad by 2016, I hope I've shown that it's hugely unlikely for Macs to move to ARM in the same timeframe.

 

There's no convincing power story to move the MBA to ARM.  Not with Haswell having 10W parts and Atom showing up with either parity or exceeding ARM performance per watt AND x86 compatibility in the next couple years.  The ARM in the Chromebook is a ~4-5W TDP part.

Reply
Reply
post #14 of 35
12/6/12 at 10:29am

Intel can say whatever it wants too but the reality is they aren't setup to do custom SoC the way ARM is.   could they get that way, possibly but it would take a major mindset change at Intel.   As to future Intel cores of course there is lots of engineering going into them, that is because their current solutions are crap.   However you missed the most important point or glossed over it, Apple can't engineer the chip to suit their needs with Intels current approach to this market.   Further it is in Apples strategic interest to engineer their own solutions.

 

AS to ATOM being a SoC, it might be that if you become very liberal with the term but current implementations lack the integration to really compete with anything that can be done with ARM hardware in this sector.   As for Intel say hell yeah apparently they already turned down the possibility of becoming a foundry for Apple.   That pretty much says it all.

Quote:
Originally Posted by nht View Post

Quote:
Originally Posted by wizard69 View Post

Not a chance. It really has nothing to do with performance, it is more about being the captain of your own ship. The reality is that unless Intel has a change of heart with respect to custom SoC they will have little success in mobile. Why? Pretty simple today's SoC is equivalent to the printed circuit boards back in the day when Wozniak and other engineers made leaps and bounds in clever circuit design. In effect mobile devices are built on SoC technologies, that is where IP goes these days to make unique products.

Frankly I could argue that Intel will have no choice but to support customers with custom chips. It is the way of the future for these small devices.

 

Then a reporter asked if Intel customizes parts for its largest customers. “Yes,” Bryant said. “We want to give them a way of differentiating their machines.”

 

Bryant didn’t say much more, but those few words shine a light on another part of the big-time chip business that’s rarely discussed. There are cases where a large chipmaker such as Intel and AMD will provide certain customers with chips that others may not have access to. Sometimes, this merely means that when the chipmaker cranks out a big batch of processors, one customer gets the chips that happen to have the best speed or power ratings. But in other cases, the chipmaker will actually modify processors at the request of a particular customer.

 

http://www.wired.com/wiredenterprise/2012/09/intel-amd-custom-chips/all/

 

 

Intel produced a custom Merom for the first MBA.  If Apple wants some secret sauce in its SoC then Intel probably would do it.  The only exception is that Intel probably really really wants to push it's Gen7 GPU (Ivy Bridge HD Graphics with 4 cores) in their 22nm SOCs over the SGX.

 

That's faster than today's top end SGX554MP4 but we're also looking a year or two from now.

 

Quote:
Doesn't matter. Seriously Apples move to Soc technology with ARM cores is all about where engineering is done these days.

 

There's just as much engineering being done to the Intel Silvermont and Airmont cores and if Apple wants something special it probably has the mojo to get it given the volumes it needs AND the cachet of its branding.  An Apple mobile design win for Intel means a lot more than the desktop/laptop design win did.  That was a nice to have.  This would be a major change in the mobile processor playing field.

 

 

Quote:
It isn't a simple matter of measuring processor core performance. In fact the processor core is just a small part of today's chips. You really need to think of that wafer of silicon as a replacement for the fiberglass printed circuit boards of the past. Instead of piecing together a system with a bunch of discrete chips and clever engineering to realize a machine it is rather an effort of merging IP and synthesized logic and application blocks to realize a device.
Honestly I think Intel has missed the boat here. Unless they open up ATOMs cores to custom implementations they will get no where in mobile.

 

You guys don't seem to realize that Medfield and all the follow on smartphone/tablet Atoms are SOCs.  In this generation with SGX544MP2 GPUs on the tablets and SGX 545 on netbooks and SGX540 on phones.  If Apple had said "we'll use the Saltwell processor for the iPad 4 but we need SGX554MP4 instead and these three things removed from the die and these two added you really don't think that Intel would have said "Hell yeah!".

 

Frankly Saltwell isn't fast enough for it to be a compelling move for Apple.  I do see 14nm Airmont in 2014 being compelling.  If the cost is moving to Gen8/9 Intel GPUs (the Haswell ones on 14nm) that's probably okay.

Reply
Reply
post #15 of 35
12/6/12 at 10:33am

ARM benefits from process shrinks just like Intel does.   

Quote:
Originally Posted by nht View Post

Quote:
Originally Posted by hmm View Post


I don't see it as so clear cut. I would suggest that if Apple was going to implement Atom, it would have happened a long time ago. They aren't so likely to change a successful strategy today. With PowerPC -> Intel, they were floundering. Notebooks were the largest potential growth market in the Mac lineup and they were stuck with the G4 process. They don't really have that kind of problem today.

 

Intel Atom has sucked until Medfield.  It simply was never a viable option for Apple.

 

With 2012 Medfield they have parity with A9s so they are a half step behind.  (Why do that?)

With 2013 22nm (Valleyview, etc) I expect parity with A15s.  (Okay,  but the A6 is already here since Apple is slightly ahead of the pack so they are still a half step behind Apple)

with 2014 14nm I expect them to be ahead of ARM on both power and performance in 2014. (Now you have a reason to think about switching horses)

 

Which is why I hedged and said 2016 iPads.

Reply
Reply
post #16 of 35
12/6/12 at 10:40am

Crackpot, not really.   I just think you are missing the point with Apple going to custom ARM SoCs.

 

As to "Mac" with ARM processors I can't see any good reason for that to happen my self.   It would just confuse the family of devices.   However I could see Apple offering a different product with ARM as the processor of choice; it would be marketed as a separate product, probably iOS derived, to prevent product confusion.    An iPad with keyboard if you will.   

 

The reason they would do an ARM based laptop is the same reason they will stay with ARM in iPads.   It allows them to command their own ship.   That is they could tailor a SoC for the exact functionality they would want to put into the device.   It is no different than Apple TV where the A series chip was tailored for that machine.

Quote:
Originally Posted by nht View Post

Regardless of whether you think I'm a crackpot to suggest an Atom powered iPad by 2016, I hope I've shown that it's hugely unlikely for Macs to move to ARM in the same timeframe.

 

There's no convincing power story to move the MBA to ARM.  Not with Haswell having 10W parts and Atom showing up with either parity or exceeding ARM performance per watt AND x86 compatibility in the next couple years.  The ARM in the Chromebook is a ~4-5W TDP part.

Reply
Reply
post #17 of 35
12/6/12 at 12:19pm
Thread Starter 
Quote:
Originally Posted by wizard69 View Post

ARM benefits from process shrinks just like Intel does.   

 

Yes, but Intel is getting to 14nm mobile before anyone else.  Instead of competing a process node behind (32nm vs 28nm in 2012) they'll be a process node ahead with Atom and that's a big difference.

 

Quote:
AS to ATOM being a SoC, it might be that if you become very liberal with the term but current implementations lack the integration to really compete with anything that can be done with ARM hardware in this sector.  

 

 

The Medfield SoC used in the K800 and Xolo are SoCs just like the A6. In what way is it any less "integrated" and not a SoC.

 

Given that the Xolo outperformed the Galaxy Nexus (Dual Core OMAP Cortex A9) and on par with the HTX One X (Quad Core Tegra 3 Cortex-A9) and has decent real world battery life despite the small battery in the Xolo (5.4 Whr vs 6.85 Whr in the Nexus) is a clear indicator that it IS competitive with ARM A9 designs.  

 

Since the Xolo X900 is an early-2012 phone that's not that shabby for Intel.  Intel is not trying for "not shabby" but industry leading performance.  It might get there next year at 22nm but I suspect it won't be until 2014 where it beats ARM performance in all respects for what AnandTech calls a "Conroe moment".

Reply
Reply
post #18 of 35
1/19/13 at 7:20am
Thread Starter 

Taking honor bets.  Loser agrees pays charity of winner's choice, no verification...just do it or not on your own honor.  This is just for fun and bragging rights.

 

My charity choice is Dean Kaman's First http://www.usfirst.org/

 

I'm going to say the odds are only 10-1 against Intel iPad design win by end of 2016 but if I like your charity I'll do 1-1.  

Reply
Reply
post #19 of 35
1/19/13 at 2:01pm
Quote:
Originally Posted by nht View Post

Taking honor bets.  Loser agrees pays charity of winner's choice, no verification...just do it or not on your own honor.  This is just for fun and bragging rights.

My charity choice is Dean Kaman's First http://www.usfirst.org/

I'm going to say the odds are only 10-1 against Intel iPad design win by end of 2016 but if I like your charity I'll do 1-1.  

This isn't even worth gambling on.

Besides that I thought your position was to support the idea.
Edited by wizard69 - 1/19/13 at 2:07pm
Reply
Reply
post #20 of 35
1/19/13 at 2:45pm
Thread Starter 
Quote:
Originally Posted by wizard69 View Post

This isn't even worth gambling on.

Besides that I thought your position was to support the idea.

 

I do.  I figure 1-10 odds are pretty good for something most folks this has a one in a million chance (or even zero chance) of happening.

 

/shrug

 

Folks are pretty adamant about their positions around here so why not put up or shut up?

Reply
Reply
post #21 of 35
1/19/13 at 9:01pm

It's funny you said it's less like for Macs to move to ARM. Well I think the people at Microsoft think ARM is the future since they're trying to run Windows 8 (and I would assume future releases) on ARM based machines. Apple is probably experimenting with the same.

 

It's a matter of time before we see ARM based systems that can run Pro Softwares and create media and professional content.

Reply
Reply
post #22 of 35
1/20/13 at 1:59am
Quote:
Originally Posted by blackbook 
It's a matter of time before we see ARM based systems that can run Pro Softwares and create media and professional content.

They are already powerful enough (a laptop/desktop system would scale up the power) but the software compatibility will be trickier.

Apple has a bit of a predicament here because the future of computer chips is an SoC. They already know how to build their own but if they help Intel, they're just going to sell it everyone.

It's ok for Apple to help Intel on the desktop because Apple's prices mean they won't get a huge marketshare anyway and it means they retain compatibility with all the existing software.

However, Intel is moving into mobile anyway so any improvements they make on the desktop side will filter into the mobile products where they will compete with Apple.

On the mobile side, the price of the chips, the production volume, production schedule, quality control, chip features, power consumption, IO controllers, GPU performance and so on are very important - so important that it would be difficult to hand over to Intel. If it means Apple has to sell iPhones and iPads at $50 higher price, that could seriously affect their products.

On the Mac side, switching to ARM would create a problem for software developers and if Apple couldn't make a chip that was significantly faster or cheaper then there would be little point. If they can make chips that are $200 cheaper that are 1.5-2x as fast, there's a point, especially on the high-end. They might be able to put 12-core CPUs in machines that cost $2,000.

Obviously ARM on both means binary compatible software but touch on the desktop has a lot of things to overcome. It has to multi-task in the sense of having apps side by side. Something as simple as a calculator has to show at the same time as a spreadsheet. This means you have to allow the UI to split - they can do this in columns using swipe from left or right to add a new blank column and then tap the space to select an app to load into it and move the divider to close. But there's only one software keyboard to manage multiple contexts and touch apps generally aren't built to scale and desktop apps are designed around a multi-window system.

Tim Cook has said they won't converge the two sides, which is a sensible choice for the foreseeable future. Given that fact, there doesn't seem to be any reason to switch the mobile side to Intel unless they can do a better job than Apple and not increase the price of Apple's devices. I doubt that's the case. Switching Macs to ARM would cause problems with Bootcamp and given that Apple developed the solution, they obviously saw a need for it. Emulation would be ok for most things but VMWare can refuse to load graphical apps at any time so it's not reliable enough.

I think the safest bet for Apple is to stick with ARM for mobile and Intel for the desktop but it would be nice to see them plug an ARM chip into the desktops/laptops somewhere if it wasn't too expensive to use as a co-processor and for low power modes. It could even have clocking ability to check how iOS apps run on slower CPUs.
Reply
Reply
post #23 of 35
1/20/13 at 8:50pm
I don't see Apple moving to ARM ever for the Macs. The progress in i86 world is still strong and frankly Atom don't hold a candle to equivalent AMD chips. Here is a recent press release: http://www.amd.com/us/press-releases/Pages/amd_unveils_new_apus.aspx for new AMD hardware. AMD will be shipping true SoC i86 processors this year. Intel is behind the eight ball with comparable low power processors so maybe this will finally light a fire at Intel.
Reply
Reply
post #24 of 35
1/20/13 at 10:19pm
  • hmm
  • Keylogger
Quote:
Originally Posted by nht View Post

Taking honor bets.  Loser agrees pays charity of winner's choice, no verification...just do it or not on your own honor.  This is just for fun and bragging rights.

 

My charity choice is Dean Kaman's First http://www.usfirst.org/

 

I'm going to say the odds are only 10-1 against Intel iPad design win by end of 2016 but if I like your charity I'll do 1-1.  


I don't think Apple is going to change designs unless their current path falls significantly behind for more than one cycle. They're an extremely large company, and their current R&D has worked out pretty well. I don't see them changing directions unless they foresee hitting a wall.

Reply
Reply
post #25 of 35
1/21/13 at 12:35am
Quote:
Originally Posted by hmm 
I don't think Apple is going to change designs unless their current path falls significantly behind for more than one cycle.

That's really it, there would need to be a problem that requires a solution. I don't see any problem with the ARM setup. Switching to x86 would just mean developers have to recompile and resubmit 700,000+ apps.

Although there's some desktop compatibility with the x86 architecture, it doesn't mean the Adobe CS Suite can suddenly run on it. iOS doesn't have a Java runtime and desktop apps have to have a new UI for touch.

There would need to be clear benefits to x86 for mobile and I don't think there are.
Reply
Reply
post #26 of 35
1/21/13 at 4:17am
Thread Starter 
Quote:
Originally Posted by blackbook View Post

It's funny you said it's less like for Macs to move to ARM. Well I think the people at Microsoft think ARM is the future since they're trying to run Windows 8 (and I would assume future releases) on ARM based machines. Apple is probably experimenting with the same.

 

It's a matter of time before we see ARM based systems that can run Pro Softwares and create media and professional content.

 

I think Surface RT is a bust and that's not a very bold opinion.  By this time next year it'll be Surface Pro all the way.

Reply
Reply
post #27 of 35
1/21/13 at 4:34am
Thread Starter 
Quote:
Originally Posted by Marvin View Post

They are already powerful enough (a laptop/desktop system would scale up the power) but the software compatibility will be trickier.

 

The A15 is powerful for an arm but an old Core i3 crushes the A15 in every benchmark by the proverbial country mile.

 

http://www.phoronix.com/scan.php?page=article&item=samsung_exynos5_dual&num=4

 

Yes, that's a 35W TDP part but I have no clue why Phoronix choose an old Core i3 over benchmarking the 22nm Core i7-3667U w/HD4000 with a 17W TDP.

 

 

Quote:
Apple has a bit of a predicament here because the future of computer chips is an SoC. They already know how to build their own but if they help Intel, they're just going to sell it everyone.

It's ok for Apple to help Intel on the desktop because Apple's prices mean they won't get a huge marketshare anyway and it means they retain compatibility with all the existing software.

However, Intel is moving into mobile anyway so any improvements they make on the desktop side will filter into the mobile products where they will compete with Apple.

On the mobile side, the price of the chips, the production volume, production schedule, quality control, chip features, power consumption, IO controllers, GPU performance and so on are very important - so important that it would be difficult to hand over to Intel. If it means Apple has to sell iPhones and iPads at $50 higher price, that could seriously affect their products.

 

Intel has stated they don't intend on losing a design win over costs.  At least not the first one.  And ARM is going to get more expensive (and more power hungry) as it gets bigger.

 

As far as IP transfer there's going to be leakage anyway.  Hence the move from Samsung.  Helping TSMC will help also help all other TSMC customers.

 

The advantage of Intel over TSMC IS production volume, production schedule and quality control.  Just like Intel's advantage over Freescale/Moto/IBM.

 

Custom IP is covered under patent agreements.  Custom design covered under contract provisions.

 

 

Quote:
I think the safest bet for Apple is to stick with ARM for mobile and Intel for the desktop but it would be nice to see them plug an ARM chip into the desktops/laptops somewhere if it wasn't too expensive to use as a co-processor and for low power modes. It could even have clocking ability to check how iOS apps run on slower CPUs.

 

Funny you should mention the word bet...thus far no one seems firm enough in their convictions to make one.

Reply
Reply
post #28 of 35
1/21/13 at 4:55am
Thread Starter 
Quote:
Originally Posted by hmm View Post

I don't think Apple is going to change designs unless their current path falls significantly behind for more than one cycle. They're an extremely large company, and their current R&D has worked out pretty well. I don't see them changing directions unless they foresee hitting a wall.

 

You mean like in 2005?

 

The wall is they don't want to do business with Samsung and TSMC/GF/UMC/etc have problems getting to 16nm/14nm FinFET in volume production while Intel has already has 3 fabs doing 14nm in volume for a year and knocking on the doors of 10nm with 450mm wafers only a couple more year out. 

Reply
Reply
post #29 of 35
1/21/13 at 5:43am
Thread Starter 
Quote:
Originally Posted by Marvin View Post

That's really it, there would need to be a problem that requires a solution. I don't see any problem with the ARM setup. Switching to x86 would just mean developers have to recompile and resubmit 700,000+ apps.

Although there's some desktop compatibility with the x86 architecture, it doesn't mean the Adobe CS Suite can suddenly run on it. iOS doesn't have a Java runtime and desktop apps have to have a new UI for touch.

There would need to be clear benefits to x86 for mobile and I don't think there are.

 

The Cortex A15 is more power efficient than the A9 for the same work but at load it's a lot more power hungry at the top end.  When both the CPU and GPU are ramped the Exynos 5 Dual peak TDP is closer to 8W.  

 

http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown/14

 

Want to bet the A50 won't continue this trend of higher peak TDP? Apps, especially games, are more than willing to use all the CPU and GPU power available.

 

As Anand points out, Intel showed Haswell running the Unigen Heaven benchmark at IDF pulling less than 8W.

 

By 14nm the clear benefit of Intel vs ARM could be performance per watt at the 4W+ TDP range AND a more reliable source of processors in the volumes that Apple wants.

 

If Intel gets to 450mm wafers before the other big boys then cost could also be an Intel advantage.  If that's true then Intel will beat TSMC on price, performance per watt and reliability as a supplier.  Trifecta.

 

There's a few "if's" in there which is why this is a bet rather than a proclamation of a sure thing.

Reply
Reply
post #30 of 35
1/21/13 at 7:33am
Quote:
Originally Posted by nht 
Yes, that's a 35W TDP part but I have no clue why Phoronix choose an old Core i3 over benchmarking the 22nm Core i7-3667U w/HD4000 with a 17W TDP.

Because it's still a crazy comparison. Laptops have fans to cool them down, the Chromebook is passively cooled so the power draw is adjusted to keep it going.
Quote:
Originally Posted by nht 
ARM is going to get more expensive (and more power hungry) as it gets bigger.

As what gets bigger?
Quote:
Originally Posted by nht 
The advantage of Intel over TSMC IS production volume, production schedule and quality control.  Just like Intel's advantage over Freescale/Moto/IBM.

This is the same company that delayed the rollout of Ivy Bridge:

http://www.theverge.com/2012/2/27/2827175/ivy-bridge-delay-confirmed-intel-executive-manufacturing-process

The same company that messed up their chipsets:

http://newsroom.intel.com/community/intel_newsroom/blog/2011/01/31/intel-identifies-chipset-design-error-implementing-solution

The same company that is now an entire year behind with their Xeon architecture, didn't have USB 3 support for 4 years, that aims to match the graphics capabilities of their rivals' products from the previous year and falls short.
Quote:
Originally Posted by nht 
When both the CPU and GPU are ramped the Exynos 5 Dual peak TDP is closer to 8W.

"Intel actually tipped me off to this test to find out what happens if we try and force both the CPU and GPU to run at max performance at the same time"

http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown/13

The device still aims for a 4W power draw. Intel's laptop chips are quite happy to sit at 17W as Intel demonstrated with the Ultrabook chips and 8W was for the GPU test. You have to compare them in the same target platform:

http://www.androidauthority.com/exynos-5-dual-benchmarks-125134/

Intel will be able to improve with 22nm Medfield but their rivals will improve too:

http://www.extremetech.com/extreme/142926-intel-unveils-22nm-soc-transistors-while-tsmc-and-globalfoundries-plan-risky-process-jumps
http://www.eetimes.com/electronics-news/4398727/TSMC-taps-ARM-V8-in-road-to-16-nm-FinFET

If TSMC can't deliver what Apple needs, they could turn to Intel but they don't need to use Atom. They can get Intel to manufacturer their own chip designs.
Quote:
Originally Posted by nht 
Want to bet the A50 won't continue this trend of higher peak TDP?

What trend? They have to put roughly the same battery packs in so they just can't sustain higher wattage without affecting battery life as well as heat dissipation.
Quote:
Originally Posted by nht 
As Anand points out, Intel showed Haswell running the Unigen Heaven benchmark at IDF pulling less than 8W.

Yes it was a good demo but again you are talking about 8W sustained with cooling on 22nm, not 4W sustained passively cooled on 28nm.

Intel still has to play catch-up here. When they are in front, then it will be worth Apple considering it otherwise they'd just end up behind Samsung.
Reply
Reply
post #31 of 35
1/21/13 at 8:59am
Thread Starter 
Quote:
Originally Posted by Marvin View Post

As what gets bigger?
 

 

core die size

 

 

Quote:
This is the same company that delayed the rollout of Ivy Bridge:

http://www.theverge.com/2012/2/27/2827175/ivy-bridge-delay-confirmed-intel-executive-manufacturing-process

The same company that messed up their chipsets:

http://newsroom.intel.com/community/intel_newsroom/blog/2011/01/31/intel-identifies-chipset-design-error-implementing-solution

The same company that is now an entire year behind with their Xeon architecture, didn't have USB 3 support for 4 years, that aims to match the graphics capabilities of their rivals' products from the previous year and falls short.
"Intel actually tipped me off to this test to find out what happens if we try and force both the CPU and GPU to run at max performance at the same time"

http://www.anandtech.com/show/6536/arm-vs-x86-the-real-showdown/13

The device still aims for a 4W power draw. Intel's laptop chips are quite happy to sit at 17W as Intel demonstrated with the Ultrabook chips and 8W was for the GPU test. You have to compare them in the same target platform:

http://www.androidauthority.com/exynos-5-dual-benchmarks-125134/

Intel will be able to improve with 22nm Medfield but their rivals will improve too:

http://www.extremetech.com/extreme/142926-intel-unveils-22nm-soc-transistors-while-tsmc-and-globalfoundries-plan-risky-process-jumps
http://www.eetimes.com/electronics-news/4398727/TSMC-taps-ARM-V8-in-road-to-16-nm-FinFET

If TSMC can't deliver what Apple needs, they could turn to Intel but they don't need to use Atom. They can get Intel to manufacturer their own chip designs.
What trend? They have to put roughly the same battery packs in so they just can't sustain higher wattage without affecting battery life as well as heat dissipation.
Yes it was a good demo but again you are talking about 8W sustained with cooling on 22nm, not 4W sustained passively cooled on 28nm.

Intel still has to play catch-up here. When they are in front, then it will be worth Apple considering it otherwise they'd just end up behind Samsung.

 

So I'm guessing you like the odds?

 

Of course Intel is playing catch up on power consumption.  Hence the 2016 date rather than a 2013 date.  To me it looks a little like 2003 with Intel looking like it has a great future roadmap even though AMD has been kicking the crap out of them.

Reply
Reply
post #32 of 35
1/21/13 at 10:53am
Quote:
Originally Posted by nht 
core die size

The A6X (123mm2) is smaller than the A5X (163mm2).

http://www.chipworks.com/blog/recentteardowns/2012/03/16/the-new-ipad-a-closer-look-inside/
http://www.macrumors.com/2012/11/02/teardown-of-fourth-generation-ipad-reveals-lg-display-a6x-chip-lightning-connector/

The A5 was 122m2 and the A6 was 97mm2.

http://www.chipworks.com/blog/recentteardowns/2012/09/21/apple-iphone-5-the-a6-application-processor/

It seems to go up and down depending on what stage they are at with the design. I don't see the TDP going up much.
Quote:
Originally Posted by nht 
Of course Intel is playing catch up on power consumption.  Hence the 2016 date rather than a 2013 date.  To me it looks a little like 2003 with Intel looking like it has a great future roadmap even though AMD has been kicking the crap out of them.

I don't remember AMD having over 95% marketshare.

Apple was originally going to use Intel processors in the iPhone but Tony Fadell convinced them not to. By 2016, Intel might have something more compelling but they might not.

You have to also take into consideration graphics performance. Intel's Medfield scored 41fps in Egypt offscreen:

http://www.engadget.com/2012/09/18/motorolas-razr-i-benchmarks-intel-2ghz-medfield/

The iPhone 5 scored 147fps:

http://www.macworld.com/article/2010360/lab-tests-iphone-5-leaves-old-iphones-in-the-dust.html

In that regard, it is more like the comparison with AMD. With Intel chips, Apple won't have the control to make their graphics performance that high.
Reply
Reply
post #33 of 35
1/21/13 at 11:45am
  • hmm
  • Keylogger
Quote:
Originally Posted by nht View Post

 

You mean like in 2005?

 

The wall is they don't want to do business with Samsung and TSMC/GF/UMC/etc have problems getting to 16nm/14nm FinFET in volume production while Intel has already has 3 fabs doing 14nm in volume for a year and knocking on the doors of 10nm with 450mm wafers only a couple more year out. 

 

This comes down to whether they foresee such things as a problem. Intel's shipping dates slipped on ivy bridge. Sandy faced an initial recall due to problems with their SATA implementation. Its reverberations set their EP release even further back. They are very good. It doesn't make them infallible. In 2005 the problems were ongoing. The notebooks were stuck with G4s. It seemed like everyone I knew with a G5 had at least one major hardware problem. They had trouble at both ends of the spectrum. Are there signs that will happen repeat itself here?

 

Quote:
Originally Posted by Marvin View Post


That's really it, there would need to be a problem that requires a solution. I don't see any problem with the ARM setup. Switching to x86 would just mean developers have to recompile and resubmit 700,000+ apps.

Although there's some desktop compatibility with the x86 architecture, it doesn't mean the Adobe CS Suite can suddenly run on it. iOS doesn't have a Java runtime and desktop apps have to have a new UI for touch.

There would need to be clear benefits to x86 for mobile and I don't think there are.

 

iOS lacks many things. I wouldn't mind a real file system as I require organization. Admittedly I haven't investigated if this functionality can be duplicated via apps. As of right now these markets are a bit different. Companies don't offer the same functionality on their iOS apps, but they typically charge much less. I think a huge amount of potential exists there. I'm also tempted to buy an ipad + stylus just for sketching.

Reply
Reply
post #34 of 35
1/21/13 at 1:23pm
Thread Starter 
Quote:
Originally Posted by Marvin View Post

The A5 was 122m2 and the A6 was 97mm2.

 

 

Ah right, I must have transposed those sizes in my memory.  My bad.

 

 

Quote:

You have to also take into consideration graphics performance. Intel's Medfield scored 41fps in Egypt offscreen:

http://www.engadget.com/2012/09/18/motorolas-razr-i-benchmarks-intel-2ghz-medfield/

The iPhone 5 scored 147fps:

http://www.macworld.com/article/2010360/lab-tests-iphone-5-leaves-old-iphones-in-the-dust.html

In that regard, it is more like the comparison with AMD. With Intel chips, Apple won't have the control to make their graphics performance that high.

 

The Medfield has a SGX 540 at 400MHz which is slow.

 

I would expect that Apple would insist on being able to spec the GPU in any agreements...especially given Intel is currently using PowerVR SGX in their own designs at the moment.  

 

An Apple design win puts Intel on top and Intel knows this.  This combined with the volume means that Apple has a lot of leverage in negotiations.

 

Again, there's a lot of negativity toward this prediction and it is warranted but thus far no one seem willing to accept a bet that nobody but them will know if they honor for a charity.  

Reply
Reply
post #35 of 35
1/21/13 at 3:50pm
This discussion is interesting but I think people mis the most important point here. System on Chips are effectively the printed circuit boards of the past. As such it allows for customized engineering to fit specific needs of the products being built. Apple can't go with Intel until Intel has addressed their inability to do custom chips. Intel is seemingly unwilling to do so.

In a nut shell it isn't the CPU nor the GPU that apple is tied too, but rather the ability to lash up all the required functionality they need on one die. Frankly that is why ARM is so hot in this market right now, itis the ability to build tailored devices that is so attractive to manufactures like Apple. It really doesn't matter if Intel is one process node ahead because they don't support the infrastructure for the likes of Apple to take advantage of it.

Beyond that I can see Intel coming on hard times if they don't address the continued success of ARM, low power devices and innovation. Even AMD has seen the writing on the wall as they Agressively try to incorporate their IP with ARM and have offered up custom and fully baked i86 SoC chips.
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Future Apple Hardware
AppleInsider › Forums › Mac Hardware › Future Apple Hardware › Prediction: Atom based iPad by 2016