Intel looks to build ultra-efficient mobile chips Apple 'can't ignore'
Computing giant Intel is once again looking to gain ground in the mobile device market currently dominated by low-power ARM-based processors, saying that it plans to ply its substantial faculties toward currying favor with major player Apple.
During Intel's annual investor day on Thursday, CEO Paul Otellini outlined the company's plan to leverage its multi-billion-dollar chip fabrication plants, thousands of developers and industry sway to catch up in the lucrative mobile device sector, reports Forbes.
?Our job is to ensure our silicon is so compelling, in terms of running the Mac better or being a better iPad device, that as they make those decisions they can?t ignore us,? Otellini said.
Mobile devices like Apple's market-leading iPad are quickly devouring the PC market, which saw a nearly 6 percent drop in worldwide shipments in the fourth quarter of 2011. Apple Chief Executive Tim Cook said in January that he expects the tablet market to one day outnumber PCs, and it seems that Intel tends to agree given its realigned efforts.
Currently, Apple relies on Intel processors to power its Mac line of computers following a switch away from IBM's PowerPC platform in 2005. For its iDevice products, which far outsell Macs, the Cupertino-based tech monolith uses proprietary ARM-based chips fabricated by Samsung.
Intel has been noticeably absent from the mobile processor space, though the recent acquisition of wireless chip maker Infineon could give the Santa Clara-based company a leg-up in offering integrated mobile platform solutions. Since the purchase, Intel has shipped 500 million wireless chipsets. Also of note are the thousands of programmers employed by Intel who can help tune mobile device software to run efficiently on x86 processors. It was this army of software engineers that helped Apple abandon PowerPC.
Intel's Atom "Medfield" chipset is bound for use Motorola and Lenovo smartphones.
Intel reportedly inked agreements to supply Motorola and Lenovo with Atom "Medfield" processors for use in upcoming products, and more OEMs are supposedly in line to implement the new chipsets. According to Co-Manager of Intel's Mobile Communications Group Mike Bell, the new manufacturing process used to build the chips is superior to offerings from the competition.
?What the process technology does is gives us better performance, at better power at better size,? Bell said. ?We think this is a fundamental advantage that we have.?
While Apple's newest A5 and A5X processor family is built on the 32nm and 45nm processes, respectively, Intel is developing a more efficient 22nm component slated for release in 2013. The firm expects to be building parts using a 14nm process by 2014.
Not surprisingly, Intel was said to have been interested in producing Apple's A-series silicon last year in an attempt to marginalize Samsung's share of the fabrication business.
The two companies remain strange bedfellows as the chip maker provides parts for Mac computers while pushing its own competing Ultrabook thin-and-light platform. An upcoming MacBook refresh is expected to feature Intel's "Ivy Bridge" chipset with new processors entering the Core i5 and Core i7 families. Rumors circulated earlier this year that said Apple was looking to use its own ARM processors in upcoming iterations of the MacBook, especially in power-critical applications like the thin-and-light MacBook Air, but those whispered promises have yet to materialize.
It remains to be seen what effects Intel's renewed efforts in mobile will yield, though it is unlikely that Apple will turn away from its own mobile chip-making business which by some estimates could become the world's largest by the end of 2012.
During Intel's annual investor day on Thursday, CEO Paul Otellini outlined the company's plan to leverage its multi-billion-dollar chip fabrication plants, thousands of developers and industry sway to catch up in the lucrative mobile device sector, reports Forbes.
?Our job is to ensure our silicon is so compelling, in terms of running the Mac better or being a better iPad device, that as they make those decisions they can?t ignore us,? Otellini said.
Mobile devices like Apple's market-leading iPad are quickly devouring the PC market, which saw a nearly 6 percent drop in worldwide shipments in the fourth quarter of 2011. Apple Chief Executive Tim Cook said in January that he expects the tablet market to one day outnumber PCs, and it seems that Intel tends to agree given its realigned efforts.
Currently, Apple relies on Intel processors to power its Mac line of computers following a switch away from IBM's PowerPC platform in 2005. For its iDevice products, which far outsell Macs, the Cupertino-based tech monolith uses proprietary ARM-based chips fabricated by Samsung.
Intel has been noticeably absent from the mobile processor space, though the recent acquisition of wireless chip maker Infineon could give the Santa Clara-based company a leg-up in offering integrated mobile platform solutions. Since the purchase, Intel has shipped 500 million wireless chipsets. Also of note are the thousands of programmers employed by Intel who can help tune mobile device software to run efficiently on x86 processors. It was this army of software engineers that helped Apple abandon PowerPC.
Intel's Atom "Medfield" chipset is bound for use Motorola and Lenovo smartphones.
Intel reportedly inked agreements to supply Motorola and Lenovo with Atom "Medfield" processors for use in upcoming products, and more OEMs are supposedly in line to implement the new chipsets. According to Co-Manager of Intel's Mobile Communications Group Mike Bell, the new manufacturing process used to build the chips is superior to offerings from the competition.
?What the process technology does is gives us better performance, at better power at better size,? Bell said. ?We think this is a fundamental advantage that we have.?
While Apple's newest A5 and A5X processor family is built on the 32nm and 45nm processes, respectively, Intel is developing a more efficient 22nm component slated for release in 2013. The firm expects to be building parts using a 14nm process by 2014.
Not surprisingly, Intel was said to have been interested in producing Apple's A-series silicon last year in an attempt to marginalize Samsung's share of the fabrication business.
The two companies remain strange bedfellows as the chip maker provides parts for Mac computers while pushing its own competing Ultrabook thin-and-light platform. An upcoming MacBook refresh is expected to feature Intel's "Ivy Bridge" chipset with new processors entering the Core i5 and Core i7 families. Rumors circulated earlier this year that said Apple was looking to use its own ARM processors in upcoming iterations of the MacBook, especially in power-critical applications like the thin-and-light MacBook Air, but those whispered promises have yet to materialize.
It remains to be seen what effects Intel's renewed efforts in mobile will yield, though it is unlikely that Apple will turn away from its own mobile chip-making business which by some estimates could become the world's largest by the end of 2012.
Comments
Well, Apple can ignore it for a few years at least, then we'll see what happens. That said, this could be a very costly mistake for Intel if they can't compete effectively with ARM by 2015. This is like steering a big ship 180 degrees, so if Intel pulls it off by 2015 and is clobbering ARM it will be one of the biggest tech successes of the early part of this century.
It could have been better for Intel to fab ARM chips because Intel's fabs are kickass, mobile designs less so. So Intel could have started fabbing ARM then gradually build up its mobile expertise too, once they have a better grasp of the mobile space.
It could be a case of "Not Invented Here" for Intel as well as Intel worried all along about cannibalising the already-eclipsed PC processor market.
We have got to the stage where PCs are so powerful but so ghastly to use, let's face it. 2012 is indeed the turning point in our modern era.
Apple and Intel remain strange bedfellows as the chip maker provides parts for Mac computers while pushing its own competing Ultrabook thin-and-light platform. An upcoming MacBook refresh is expected to feature Inte's "Ivy Bridge" chipset with new processors entering the Core i5 and Core i7 families. Interestingly, Intel was said to have been interested in producing Apple's A-series silicon in an attempt to marginalize Samsung's share of the fabrication business. It remains to be seen what effects Intel's efforts will yield, though it is unlikely that Apple will turn away from its own mobile chip-making business which a recent study found to be capable of becoming a world leader by the end of 2012.
Tim Cook made a compelling case that Mac is Mac and iOS is iOS, he doesn't see a convergence too soon, so Intel for MacBook Air and Macs have a good few years still. That said, it doesn't mean by 2015 a Mac that's a Mac and not iOS can't be running a fast ARM processor. So, again, 2012-2015 is critical for Intel. Either it can capitalise on mobile and post-PC processors or become another lumbering behemoth like Microsoft (though Intel arguably makes the most superb PC-class CPUs, and has decimated the discrete-GPU industry by brute force and unethical practices (eg. locking out Nvidia)).
Apple may well set up its own fabs the way it is going. It's costly, of course, but possible. Designing killer chips is one thing. Fabbing them at less than 10nm is another whole bag of cats.
In Xcode, when developing an iOS App and you run it in the simulator, it is actually compiled in to an Intel binary (that is why it's only a simulator not an emulator). This is for performance reasons of course, but one side-effect is that all iOS apps are basically already tested on Intel. They would still have to be recompiled and re-uploaded to the App Store of course, but shouldn't need any source changes should an x86 iDevice come along.
@Nvidia2008
If you read the review on Anandtech, regarding Intel's latest Medfield mobile SOC, it performed very well in terms of CPU performance, battery life and camera performance. The only area where it lagged was GPU and that's only because Medfield is single-core design. Later this year, Intel is releasing a dual-core CPU / GPU variant of Medfield plus the fact they will be fabbing 22nm mobile SOC's next year and 14nm version in 2014, so in terms of manufacturing they're well ahead of everyone in the ARM camp. With the introduction of Medfield, Intel is finally back in the game and as long as they execute their mobile roadmap the way they have their PC roadmap, they will be a major player in the mobile device sector within two years.
@Nvidia2008
If you read the review on Anandtech, regarding Intel's latest Medfield mobile SOC, it performed very well in terms of CPU performance, battery life and camera performance. The only area where it lagged was GPU and that's only because Medfield is single-core design. Later this year, Intel is releasing a dual-core CPU / GPU variant of Medfield plus the fact they will be fabbing 22nm mobile SOC's next year and 14nm version in 2014, so in terms of manufacturing they're well ahead of everyone in the ARM camp. With the introduction of Medfield, Intel is finally back in the game and as long as they execute their mobile roadmap the way they have their PC roadmap, they will be a major player in the mobile device sector within two years.
But will Apple be able to customise the chips for optimisation with its software, like they currently do with ARM? IF so, why not if it has advantages? I think we are a ways of that though.
Apple needs to keep with Samsung and others by using quad core cpu in iPhones and iPads.
Intel, I have been telling you this for years! Intel should go to 4 nm, nanotechnology and even beyond right now. And now means now. Now you see the writing on the wall. Good riddance!
Quote:
Originally Posted by peter236
Apple needs to keep with Samsung and others by using quad core cpu in iPhones and iPads.
Correction, Apple "needs" quad core ARM chips when they don't kill the battery life and hence reduce the user experience with their devices. Much like 4G, apple wouldn't go there unless the tech was ready for general use. The early 4G droid adopters were running amok ranting about how Apple didn't have 4G, all the while draining their batteries in 2 hours.
It won't happen. Apple is full-steam ahead with ARM.
Quote:
Originally Posted by Shameer Mulji
@Nvidia2008
If you read the review on Anandtech, regarding Intel's latest Medfield mobile SOC, it performed very well in terms of CPU performance, battery life and camera performance. The only area where it lagged was GPU and that's only because Medfield is single-core design. Later this year, Intel is releasing a dual-core CPU / GPU variant of Medfield plus the fact they will be fabbing 22nm mobile SOC's next year and 14nm version in 2014, so in terms of manufacturing they're well ahead of everyone in the ARM camp. With the introduction of Medfield, Intel is finally back in the game and as long as they execute their mobile roadmap the way they have their PC roadmap, they will be a major player in the mobile device sector within two years.
Apple tests their CPU choices against an internal test harness that taxes GCD, OpenCL and OpenGL. None of these reviews ever do that. They circle jerk with 3DMark and other useless tests.
Quote:
Originally Posted by nvidia2008
Well, Apple can ignore it for a few years at least, then we'll see what happens. That said, this could be a very costly mistake for Intel if they can't compete effectively with ARM by 2015. This is like steering a big ship 180 degrees, so if Intel pulls it off by 2015 and is clobbering ARM it will be one of the biggest tech successes of the early part of this century.
It could have been better for Intel to fab ARM chips because Intel's fabs are kickass, mobile designs less so. So Intel could have started fabbing ARM then gradually build up its mobile expertise too, once they have a better grasp of the mobile space.
It could be a case of "Not Invented Here" for Intel as well as Intel worried all along about cannibalising the already-eclipsed PC processor market.
We have got to the stage where PCs are so powerful but so ghastly to use, let's face it. 2012 is indeed the turning point in our modern era.
In a few years ARM Multicore CPUs will be at 14nm or lower on 300mm wafers. This late April announcement from ARM has a lot of valuable information on the A-15. http://www.arm.com/about/newsroom/arm-announces-cortex-a15-quad-core-hard-macro.php
With hindsight, Intel made a strategic mistake in 2006 when it sold off its ARM XScale division. They've been trying to steer the big x86 mobile devices ship ever since then.
I presume at the time it made sense for them to streamline and focus on their core knowledge x86 architecture.
So maybe, yes the 22nm process will finally make x86 mobile chips attractive for OEMs. The question is how much, if any advantage will Intel have before ARM fabs catch up with 22nm production?
Quote:
Originally Posted by nvidia2008
That said, this could be a very costly mistake for Intel if they can't compete effectively with ARM by 2015. This is like steering a big ship 180 degrees...
It could have been better for Intel to fab ARM chips because Intel's fabs are kickass, mobile designs less so. So Intel could have started fabbing ARM then gradually build up its mobile expertise too, once they have a better grasp of the mobile space.
It could be a case of "Not Invented Here" for Intel as well as Intel worried all along about cannibalising the already-eclipsed PC processor market.
Quote:
Originally Posted by Shameer Mulji
...they will be fabbing 22nm mobile SOC's next year and 14nm version in 2014, so in terms of manufacturing they're well ahead of everyone in the ARM camp. With the introduction of Medfield, Intel is finally back in the game and as long as they execute their mobile roadmap the way they have their PC roadmap, they will be a major player in the mobile device sector within two years.
Quote:
Originally Posted by zunx
Intel, I have been telling you this for years! Intel should go to 4 nm, nanotechnology and even beyond right now.
Well Star Trek has been telling us for years that humanity should travel around the galaxy in faster than light spacecraft,
it's not going to happen until we know how.
So, do you, Zunx, know how to make a 4nm processor with nanotechnology for $50 to $400 per piece,
so it can go into a device a consumer can afford?
If so, Intel will only be too happy to meet you.
Performance-wise, including battery life, Medfield is already competitive with the best ARM chips. It seems that the major obstacle before a wider adoption of Intel's mobile chips is price. This was discussed with more details in the <a href = "http://www.anandtech.com/show/5770/lava-xolo-x900-review-the-first-intel-medfield-phone">Lava Xolo thread on AnandTech</a>.
<img src = "http://images.anandtech.com/graphs/graph5770/45974.png">
2012 Medfiled, is performance / watt just about comparable with current ARM design. Although Intel Wins in performance, ARM still wins with lower power usage. But then it is bigger, and hotter.
2013 with 22nm ULP, it should be a lot better for idle performance but then you have 28nm LP version of Cortex A15 big.LITTLE SoC, which could means performance and Idle power wise should keep the advantage to ARM.
2014 Intel will be releasing their Mobile Atom on 14nm first before their Mainstream CPU parts. Which means Intel will again played the game with Nodes advantage, and this time it could well be two nodes ahead as TSMC aren't even reading 20nm parts til late 2014. Although Samsung might be capable of holding up.
So if all things goes well 2014 will truly be the year that Intel is truly equal or better then ARM camp in terms of performance and power.
But then how much better? According to Intel Roadmap, the consequence of quickly shifting Atom through nodes is that there aren't much changes planned to its architecture. Which means its performance / Mhz will pretty much be the same. For higher performance it will have to scale to higher clock speed. The Cortex A15 is, from a high level Point of view,a faster and more powerful Out of Order architecture then Atom. But Atom has Hyperthreading, 64 Bit ( performance advantage on x86 side only ), and better software optimization. So Lets give Intel a 20% advantage here for Smaller Nodes, Higher Frequency and may be other tricks they have.
Idle Power will still be hard to beat, the LITTLE Cortex A7 was designed with Ultra Low Power in Mind. I would doubt Intel could win, But lets call it equal with a node advantage from Intel.
Cost -
So Nvidia would have to add R&D per Unit , Unit Price Per Wafer and their Margin ~ 20% to their price of SoC.
Intel owns their Fab, so they get BOTH the Fabs Margin and their Final Sale Margin. Intel could properly make a SoC that has same performance and priced the same as Nvidia but still gets 3x% profits margin.
Now , Unlike other players Apple only make SoC for themselves. Apple makes NO profits on the SoC. Just for the simple Numbers, Apple could make a same cost SoC as Nvidia but with 20% more transistors. And Hence Faster performance.
Now for Intel to grab Apple's SoC business, they must provide equal or better performance x86 SoC for the cheaper price. Intel would then have to make chips that is good enough ( Post 2014 ) and with less then 15% Profits Margin. Much lower then what they are used to get with Desktop CPU 50%.
And then even so, the cost of the current Apple SoC is less then $30, even intel could provide something equal or better at $25. Would Apple even be bothered with the $5 dollars on the iPhone per unit sold but to lose control of their own SoC?
Apple is slowly getting there process wise but what is funny here is that Intel absolutely needs these process shrinks to compete at all. Medville is still wanting in many ways when looked at as a tablet engine.
Another thing here is that Apple has a distinct advantage going ARM, in that they can obtain a higher level of integration and customization that Intel will never support.
The tests are rigged to show Intel in a positive light. It isn't that he is dishonest as all the info is there, it is more a question of being ethical.
In any event I think Intel is pushing so damn hard here because they know they are loosing the battle. I86 is just so bloated it doesn't have a competitive chance without being two nodes ahead of ARM.
Quote:
Originally Posted by ascii
In Xcode, when developing an iOS App and you run it in the simulator, it is actually compiled in to an Intel binary (that is why it's only a simulator not an emulator). This is for performance reasons of course, but one side-effect is that all iOS apps are basically already tested on Intel. They would still have to be recompiled and re-uploaded to the App Store of course, but shouldn't need any source changes should an x86 iDevice come along.
And presumably the reverse would happen should an ARM-based Mac come along. I guess Apple would use the same fat-binary approach they used when switching from PowerPC to Intel - an application would have both Intel and ARM executables within it.
Isn't it strange to think that in ten years time we might have users on here complaining that Apple is dropping support for their x86 Macs?
Quote:
Originally Posted by ascii
In Xcode, when developing an iOS App and you run it in the simulator, it is actually compiled in to an Intel binary (that is why it's only a simulator not an emulator). This is for performance reasons of course, but one side-effect is that all iOS apps are basically already tested on Intel. They would still have to be recompiled and re-uploaded to the App Store of course, but shouldn't need any source changes should an x86 iDevice come along.
Correct me if I am wrong. With XCode and Apple OS being Unix based that when an app is developed with XCode that they can put the OS and apps on another processor by simply recompiling for the other processor and that's basically all that has to be done. So, if for some strange reason, another processor platform came out that was better that all Apple and the developers would have to do is recompile the OS and the Apps and install with the other processor?
In terms of what's going on with ARM vs Intel processors, ARM got a head start in mobile devices. If Apple can keep up against Intel in the processor arena, then Apple just has to choose fabricators for their processors. Right now, they use Samsung and Samsung is the largest chip mfg., and they have the 22 nanometer process they are playing with. Obviously, Intel could mfg Apple's chips for them if Intel would get off their ego based behavior. Apple has bought out several chip design companies since they want to design various chips if no one else has something better. It's going to get interesting with regards to the SSD controllers, since Anobit has the best SSD technology.
I mean, Apple could transfer over to the Intel mobile chips down the road if they have to, but I don't know if they have to. With mobile devices they aren't so concerned as to who's chip because it is a new platform. With the computers, CIOs got brainwashed into thinking that everything had to be X86 based, which seems kind of shortsighted.
What's funny is people are so concerned with the chip rather than the OS. The OS is what people interface with and the processor is kind of immaterial as long as programs run fast, and gets decent battery life.
Personally, I could give a rip as to what processor the device is using. We, as consumers, care about speed, battery life (in mobile devices), ease of use and reliability and how good the apps and app development platform is and of course the overall design of the product. Many just got a little brainwashed by Intel.
What was Apple supposed to do back when they started developing the iPhone and iPad initially? Wait for Intel to do something? Intel didn't begin development of their mobile chip designs early enough because they didn't see the potential growth of those markets until they started selling more mobile devices than computers.
Personally, I see mobile devices taking over a portion of the PC industry, but overall, the iPads, iPhones are devices which displace other eletronic devices or simply create a device which can easily do things that were normally done manually.