Multi-core ARM chips bound for Apple's next-gen iPhones

[AppleInsider]

Apple's new iPhone 3G S hasn't hit store shelves yet but already there's talk of next year's models, which are expected to leverage ARM's multi-core processor designs in giving the touchscreen handset a significant performance boost while simultaneously increasing battery life once again.

Like Palm's new Pre device, the iPhone 3G S is believed to be powered by a single-core ARM Cortex-A8 running at roughly 600MHz. It may be the last of its breed, however. Future embedded processors from the world's leading mobile chip designer are expected to follow a path similar to those for conventional PCs, by which performance gains are achieved by scaling up the number of available processor cores.

Reference designs for ARM's next-generation Cortex-A9 call for versions with either two or four cores, with dual-core versions "definitely" slated to turn up in smartphones that will ship sometime in 2010, James Bruce, wireless segment manager for ARM, told CNet News.com in a phone interview earlier this week.

"What we've done on the A9 is actually make it more power efficient than the A8," he said. "The dual-core A9 will be coming out on 45-nanometer rather than the (current) 65-nanometer process."

The smaller, more precise design of the A9 will signal a faster, more power-efficient chip. And although power consumption will increase some 10 to 20 percent at peak performance when compared to the A8, Bruce noted that smartphones based off the new design will actually see better battery life in real-world usage situations.

That's because, like the A8 in the iPhone 3G S, the A9 will process instructions in larger batches more efficiently than its predecessor, which means that computations will happen more rapidly and processes will take less time, leaving the battery in future iPhones with more juice for browsing the web, sending email, and uploading video files. The A9 also sports some out-of-order processing characteristics, letting the chip make use of processor cycles that would otherwise be wasted by current chip designs.

"You're getting a 2X increase (over the previous ARM design)," Brunce said of the A8 in the upcoming iPhone 3G S. "And actually the A9 takes that even further. It's a superscalar design but it's also an out-of-order design as well. There is some out-of-order aspects with the A8 but the A9 is a very aggressive out-of-order processor."



In addition to these enhancements, Apple is expected to build some of its own proprietary technologies into future iPhone chips based off ARM's A9 reference designs with the help of resources acquired last year in the purchase of fabless chip designer P.A. Semi. Last July, it secured "a long-term architecture license to ARM's current and future technology" that will allow it to do so.

By returning to its roots and developing its own brand of ARM chips in-house, Apple stands to keep a tighter lid on its future product plans while possibly saving on costs at the same time. It will also allow the electronics maker to innovate in a way going forward that will differentiate its handheld products from a growing array of competitive devices that will be left to rely on ARM's broadly available reference designs.

[ View this article at AppleInsider.com ]

[backtomac]

A9 + LTE = upgrade for me.

Hopefully next summer.

[mazzy]

See! All you who got pissy about the upgrade price, can now happily wait until your time comes and by then a faster iPhone will be out. But then everyone who purchased the 3GS now will get pissy because THEY won't get the new phone cheaper next year. Oh well . Round and round we spin.

[BuffyzDead]

Due to the two year contract thang, I'll be skipping the 3G S, even though I'd qualify for the discount price ~12/17/09.

This will be my next iPhone, the 3G SS OR the 4G.

[teckstud]

Does this mean next year's iPhone will get multi tasking with multi Apps?

[wizard69]

First, just because we suspect that Apple via PA Semi is going to put it's own IP on the SoC it doesn't mean everybody else is stuck with reference designs. TI, Samsung, Nvidia, and a host of others are fully capable of adding their own IP. In fact considering the commodity nature of cell phone processors and world wide demand I suspect Apple will have a hard time remaing competitive going it alone. There is huge incentive for ARM licensees to produce unique and competitive new variants.

As to ARM A9 I suspect that it will be awhile before it shows up in IPhone. Rather it will show up first in follow on IPod Touches and other tablets. These chips will make excellent game machines and tablets. I'm not even sure we will have to wait till 2010. Apple could steal Christmas with an A9 based Touch with a fat battery and a slightly larger screen.


Dave

[DanaCameron]

Originally Posted by BuffyzDead

Due to the two year contract thang, I'll be skipping the 3G S, even though I'd qualify for the discount price ~12/17/09.

This will be my next iPhone, the 3G SS OR the 4G.

Me too! Someone mentioned that the 3G S was most likely a stop gap to tide consumers over until the truly next-gen iPhone came out. Seems like they may have been right.

[Booga]

Originally Posted by mazzy

See! All you who got pissy about the upgrade price, can now happily wait until your time comes and by then a faster iPhone will be out. But then everyone who purchased the 3GS now will get pissy because THEY won't get the new phone cheaper next year. Oh well . Round and round we spin.

My wife and I are playing leapfrog with generations. So her phone upgrades this year, mine next... between the two of us we'll always have one of the latest in the household.

[chronster]

Ugh...

I understand the next iPhone stands to benefit from this announcement, but doesn't it seem weird to focus so much attention on it? The real news here is the processor. It's fair to mention the fact that the iPhone could use this, but at the same time, will there be an article about what the next iphone might have anytime news about the latest and greatest technology comes along? Plenty of other devices will benefit from this, not just the iPhone.

Then again, this is APPLE insider, and I guess it makes sense to keep people looking toward the future and how it will effect products they have an eye on.

So in conclusion, I got nothin :/

[hmurchison]

Well it looks like OpenCL and Grand Central Dispatch will be coming to iPhone SDK 3.5 next June. Huah.

[SpamSandwich]

This story is hilarious. Speculating about something at least a year off... at Apple?

[Crowley]

Apple may not be the only manufacturer that benefits from these chips, but suddenly Grand Central seems like the hole in one that pushes the iPhone's OS to new heights. TI, Samsung, nVidia and a host of others will struggle to catch up if that gem of technology proves its worth to even half its potential.

[chronster]

Originally Posted by Crowley

TI, Samsung, nVidia and a host of others will struggle to catch up if that gem of technology proves its worth to even half its potential.

you're talking about google voice? (formerly known as grand central.)

[ghostface147]

iPhone LTE?

[addabox]

Originally Posted by wizard69

First, just because we suspect that Apple via PA Semi is going to put it's own IP on the SoC it doesn't mean everybody else is stuck with reference designs. TI, Samsung, Nvidia, and a host of others are fully capable of adding their own IP. In fact considering the commodity nature of cell phone processors and world wide demand I suspect Apple will have a hard time remaing competitive going it alone. There is huge incentive for ARM licensees to produce unique and competitive new variants.

As to ARM A9 I suspect that it will be awhile before it shows up in IPhone. Rather it will show up first in follow on IPod Touches and other tablets. These chips will make excellent game machines and tablets. I'm not even sure we will have to wait till 2010. Apple could steal Christmas with an A9 based Touch with a fat battery and a slightly larger screen.


Dave

I understand what you're saying, but I wonder if that's actually true.

My impression is that the larger handset market operates much like the larger commodity PC market-- there's a race to get bragging rights for the latest off-the-shelf hotness (faster CPUs, cooler screen tech, higher MP cameras, etc) and, as of late, a need to get some good looking animations going with the UI as well as some ad worthy home screens, but very little in the way of basic innovation.

I mean, I take your point that someone could do some work in this area, but it doesn't appear to be considered a point of competitive advantage. And why would it? Unless your hardware innovations are driven by specific software implementations, what's the point? Some hard to explain proprietary tech that makes your still kinda cumbersome OS be cumbersome slightly faster? When you could just be trumpeting CPU speed and screen resolution and OLED and 12 MP cameras, without out all the bother?

Apple is in a unique position, because whatever they do in hardware is a means to an end-- they have ambitions for their OS and UI that may not be achievable with off-the-shelf hardware.
By using an OS X derived mobile OS, they have a desktop's worth of computing power that they can shoehorn into handsets as the hardware becomes available. They have very good motivations for doing everything possible to get more performance into the iPhone, because they have the software that can take full advantage of it.

More specifically, they have very good motivations for doing fundamental work on efficiency and power, because that will allow them to continue to innovate and differentiate with software, as opposed to just boosting their spec lists.

[solipsism]

Originally Posted by DanaCameron

Me too! Someone mentioned that the 3G S was most likely a stop gap to tide consumers over until the truly next-gen iPhone came out. Seems like they may have been right.

I’m not sure what part of this update is just a stopgap. The CPU, GPU, camera, RAM, and 3G have all have been upgraded. We’ll have to wait for iSuppli and iFixit to see more of the internal goodness. Considering that last year mainly updated the capacity and added GPS and 3G it’s extremely substantial. Hopefully it’ll be enough to get 3rd-party apps to feasibly run in the background.

Originally Posted by hmurchison

Well it looks like OpenCL and Grand Central Dispatch will be coming to iPhone SDK 3.5 next June. Huah.

It would be nice if the iPhone OS X can substantially benefit from GC and OpenCL.

Originally Posted by ghostface147

iPhone LTE?

Not unless you jump several years into the future. We’re only getting a 7.2Mbps HSDPA radio. There is no talk of HSUPA (which surprised me) and there is still Evolved HSPA still to go in the 3G range before we get to LTE. On top of that, there are no radios that will work for phones at this point. First you need the infrastructure tested and the USB/ EC/34 cards for notebooks as power and size is less important before you can even consider trying to get these radios in a svelte phone. Progress happens in steps.

[AdamIIGS]

Originally Posted by teckstud

Does this mean next year's iPhone will get multi tasking with multi Apps?

Sure so we can end up like a doofus at work that bought a Pre and got 4 1/2 hours out of his battery after it was fully charged, that would be .... awesome?

Oh and the keyboard is a joke, iPhone killer indeed.

[hmurchison]

Originally Posted by addabox

Apple is in a unique position, because whatever they do in hardware is a means to an end-- they have ambitions for their OS and UI that may not be achievable with off-the-shelf hardware.
By using an OS X derived mobile OS, they have a desktop's worth of computing power that they can shoehorn into handsets as the hardware becomes available. They have very good motivations for doing everything possible to get more performance into the iPhone, because they have the software that can take full advantage of it.

More specifically, they have very good motivations for doing fundamental work on efficiency and power, because that will allow them to continue to innovate and differentiate with software, as opposed to just boosting their spec lists.

+1

A lot of companies would love to do custom designs but it ain't cheap. Apple should be able to leverage economy of scale for their custom chips by using them in the future products where they are using off the shelf parts now.

[solipsism]

Originally Posted by AdamIIGS

Oh and the keyboard is a joke, iPhone killer indeed.

It’s certainly no anything-killer but it does fill a niche that the iPhone doesn’t address at this time. That is the best move for Palm if they were going to make any headway at this stage in the game.

[PXT]

Only 365 days to go !

[melgross]

Originally Posted by wizard69

First, just because we suspect that Apple via PA Semi is going to put it's own IP on the SoC it doesn't mean everybody else is stuck with reference designs. TI, Samsung, Nvidia, and a host of others are fully capable of adding their own IP. In fact considering the commodity nature of cell phone processors and world wide demand I suspect Apple will have a hard time remaing competitive going it alone. There is huge incentive for ARM licensees to produce unique and competitive new variants.

As to ARM A9 I suspect that it will be awhile before it shows up in IPhone. Rather it will show up first in follow on IPod Touches and other tablets. These chips will make excellent game machines and tablets. I'm not even sure we will have to wait till 2010. Apple could steal Christmas with an A9 based Touch with a fat battery and a slightly larger screen.


Dave

I think the big thing with Apple is that they want to do some things that the designs that are sold by others won't allow them to do.

It's possible, but not likely that Palm, for example, will get TI to make special mods to their designs just for them, even if Palm knew what, and how to spec it.

By Apple modding its own chips, they an do whatever they want to, and patent the designs so that others can't copy them.

It's like Intel and AMD. AMD has the license to produce x86 compatible chips, but not Intel's latest designs. That gives Intel an advantage.

[melgross]

Originally Posted by DanaCameron

Me too! Someone mentioned that the 3G S was most likely a stop gap to tide consumers over until the truly next-gen iPhone came out. Seems like they may have been right.

It was the 3G that was the stopgap. The 3Gs is the first truly upgraded model in the sense of hardware. The next one will be another upgrade.

[Bageljoey]

Originally Posted by DanaCameron

Me too! Someone mentioned that the 3G S was most likely a stop gap to tide consumers over until the truly next-gen iPhone came out. Seems like they may have been right.

This seems a like a stretch. Things are changing VERY fast in the mobile processor world right now so you can assume at this point that any revision on a year to year basis is going to be significant. If all you do is look forward to what is coming next year, I don't think it will be a good idea to buy a phone for several years...

In my mind, a "stop gap" type move would be something like next WWDC bumping the storage to 32 and 64 and improving the battery because the fabled Cortex A-9 was not ready for a June or July roll out.

The 3G S is best available and should be for many months. It may not be a necessary upgrade for 3G owners, but it is certainly not stop gap...

[jarroyo1031]

I'm getting all the phones no matter what so I guess I'll enjoy that one next year.

[SpamSandwich]

Originally Posted by pxt

only 365 days to go !

lmao!

[Dick Applebaum]

1) How well does the ARM Cortex 8 compare to the CPUs used in most netbooks... more/less/equivalent?

2) How feasible would it be to build an ARM multi-core device using multiple A 8s instead of waiting for an A 9 or a custom SOC from PAS?

[wizard69]

Originally Posted by addabox

I understand what you're saying, but I wonder if that's actually true.

Well obviously this is how I see the ARM business these days with respect to smart phones. There is a lot demand right now for high performance low power SoC.

My impression is that the larger handset market operates much like the larger commodity PC market-- there's a race to get bragging rights for the latest off-the-shelf hotness (faster CPUs, cooler screen tech, higher MP cameras, etc) and, as of late, a need to get some good looking animations going with the UI as well as some ad worthy home screens, but very little in the way of basic innovation.

If nvidia adds a GPU to an ARM Processor that is innovation. Same for what ever TI adds. Either is trying to configure the SoC in such a way that engineers can't resist.

I mean, I take your point that someone could do some work in this area, but it doesn't appear to be considered a point of competitive advantage. And why would it? Unless your hardware innovations are driven by specific software implementations, what's the point? Some hard to explain proprietary tech that makes your still kinda cumbersome OS be cumbersome slightly faster? When you could just be trumpeting CPU speed and screen resolution and OLED and 12 MP cameras, without out all the bother?

Apple is in a unique position, because whatever they do in hardware is a means to an end-- they have ambitions for their OS and UI that may not be achievable with off-the-shelf hardware.
By using an OS X derived mobile OS, they have a desktop's worth of computing power that they can shoehorn into handsets as the hardware becomes available. They have very good motivations for doing everything possible to get more performance into the iPhone, because they have the software that can take full advantage of it.

The point is design wins. You are looking at this as a consummer issue when it is really an issue of engineering / design.

As to software iPhone OS 2.x and 3.x there is still a lot of hardware that isn't fully utilized on the current SoC. So I don't see the point in your arguement. Even when Apple goes fully custom I still expect excess I/O. That due to designing a processor to support multiple devices.

More specifically, they have very good motivations for doing fundamental work on efficiency and power, because that will allow them to continue to innovate and differentiate with software, as opposed to just boosting their spec lists.

You don't think TI or Nvidia is working hard on performance per watt? Plus ARM has a whole catalog of IP optimized for low power. Frankly the only way Apple will be sucessful is if they do implement features to add to that spec list of yours. What those features will be I don't know but they have to be there to justify the engineering.




Dave

[addabox]

Originally Posted by wizard69

Well obviously this is how I see the ARM business these days with respect to smart phones. There is a lot demand right now for high performance low power SoC.

Yeah.

If nvidia adds a GPU to an ARM Processor that is innovation. Same for what ever TI adds. Either is trying to configure the SoC in such a way that engineers can't resist.

The point is design wins. You are looking at this as a consummer issue when it is really an issue of engineering / design.

OK, but I thought we were talking about Apple vs. other cell phone manufacturers.

As to software iPhone OS 2.x and 3.x there is still a lot of hardware that isn't fully utilized on the current SoC. So I don't see the point in your arguement. Even when Apple goes fully custom I still expect excess I/O. That due to designing a processor to support multiple devices.

But it's not just I/O, it's everything happening within a constrained size/power/heat envelope. If Apple goes custom they can tailor hardware to precisely meet whatever form factor requirements their particular lineup require. It doesn't seem to me the current iPhone suffers from a surfeit of processing power, at any rate.

You don't think TI or Nvidia is working hard on performance per watt? Plus ARM has a whole catalog of IP optimized for low power. Frankly the only way Apple will be sucessful is if they do implement features to add to that spec list of yours. What those features will be I don't know but they have to be there to justify the engineering.

Sure, but again I thought we were talking about rival handset makers. The point of a custom Apple solution would be to tailor silicon to their specific hardware/software needs, in a way not easily replicated by their rivals. Whatever TI or Nvidia put on the market will be available to one and all.

So I guess the actual point of contention is "can Apple with their chip guru acquisitions/hires produce something that material outstrips the incumbent big players", to which my answer would be "they don't have to if they can produce something that's a better match for their hardware/software model, as opposed to using off-the-shelf components that are of necessity designed to serve the broadest possible market."

Apple has different priorities from the typical handset makers (as far as I can tell), but the Nvidias and TIs are obliged to design toward the typical handset makers because that's where the market is. Hence, Apple might do very well for themselves by having more control over chips they put in their devices.

[801]

This could be the CPU for a "netbook" tablet it seems to me.

[brucep]

Originally Posted by Booga

My wife and I are playing leapfrog .

Nice.

[ltcommander.data]

I too find it kind of funny how the title draws the direct conclusion that the next iPhone will use A9 while the first sentence of the article indicates that we are still not sure if the announced iPhone 3G S even uses the A8.

In any case, I'm of the opinion that the A9 won't show up in iPhone 2010. A major pillar of Apple's iPhone development marketing is that the iPhone/iPod Touch is a consistent platform allowing a broad user base and simplifying development. This has been true in the iPhone/iPod Touch platforms of 2007 and 2008. 2 years of relative stability and 40 million devices with ARM11+MBX is good enough, so iPhone 2009 having ARM8+SGX is a timely split in the ecosystem to allow more speed.

Similarly, it makes the most sense for iPhone 2010 to also maintain ARM8+SGX to avoid splitting the ecosystem again so soon. Seeing the iPhone 2009 didn't receive a major external redesign, it isn't hard to assume that that will be would coming to iPhone 2010 so that can serve as the main selling point rather than ARM9. As well, for Apple's first foray into in-house iPhone processor development it makes sense for them to base it on an existing design than go for the ARM9 which no one has experience with. I agree that iPhone 2010 will be the time for a Apple SoC, but I think it'll be ARM8 based so that they can use the presumably Samsung ARM8 in the iPhone 3G S as a template. And if 45nm can bring such amazing power savings to ARM9, quoted at 10-20% over ARM8, then the savings will be even greater if ARM8 were shrunk. There's nothing saying that Apple couldn't find someone to fab their own ARM8 SoC at 45nm. Apple could use the shrink to upclock the ARM8 a bit over that of the iPhone 3G S to offer more speed just like the 2nd gen iPod Touch, without splitting the ecosystem with ARM9. An ARM9 Apple SoC can come in iPhone 2011 using the experience picked up from an ARM8 SoC.

[DanaCameron]

Originally Posted by Bageljoey

This seems a like a stretch. In my mind, a "stop gap" type move would be something like next WWDC bumping the storage to 32 and 64 and improving the battery because the fabled Cortex A-9 was not ready for a June or July roll out.

"Stop gap" may not be the right term to describe the jump from the 3G to the 3G S. I don't recall who originally referred to it as that (or if was even in this thread), but I thought I'd read somewhere that Apple had intended to make more significant hardware changes under the iPhone's hood for this year's upgrade.

Regardless, my point in referencing the sentiment was that, as others (including you) have mentioned, the iPhone 3G S doesn't seem to represent a significant-enough improvement over the iPhone 3G to warrant many current iPhone 3G owners to upgrade right away, particularly when newer ("better") technologies (e.g., 4G, ARM multi-core Cortex A9, etc.) already seem likely to work their way into the iPhone (and other devices) a year from now.

Plus, for many of us current iPhone 3G owners, our 2-year AT&T contracts expire about a year from now. As does, unless I'm mistaken, Apple's exclusivity contract with AT&T. If sometime in 2010 Apple expands the iPhone to Verizon, or the threat of doing so prompts AT&T to devise more enticing service plans (pricing and features), I'd hate to already be locked in the middle of a new AT&T contract, or in a worse position to upgrade to an even more significant "new and improved" iPhone because I sprang for the 3G S as a knee-jerk reaction to have the latest and greatest (which, for me, is what it would be at this point).

Originally Posted by Bageljoey

If all you do is look forward to what is coming next year, I don't think it will be a good idea to buy a phone for several years...

Amen to that, brotha!! And not just for phones. But I try to do my due diligence and time my purchases (or upgrades) as well as I can to avoid buyer's remorse as much as possible. For instance, I consider the promise of the Nehalem processors to be a significant enough hardware upgrade to forgo all of the incremental updates (tempting though they are) to Apple's computers coming out this year with the current architecture. I could use a new Mac, but for Nehalem… I'll wait. My feeling is that the iPhone 3G is good enough that I'm willing to wait for the "Nehalem-equivalent" iPhone hardware upgrade possibly coming in a year or so.

Originally Posted by Bageljoey

Things are changing VERY fast in the mobile processor world right now so you can assume at this point that any revision on a year to year basis is going to be significant. The 3G S is best available and should be for many months. It may not be a necessary upgrade for 3G owners, but it is certainly not stop gap...

I think you're right that the rate at which mobile processor technology is progressing lends itself to rapid (at least yearly) improvements in device offerings from many manufacturers. Apple, Inc. typically does a great job of improving their product lines. So it's almost a given that Apple had to release a hardware update to the iPhone this year.

And while the hardware improvements are impressive and tempting (I'd love to have an iPhone 3G S), I think the real update for the iPhone is the iPhone OS 3.0 Software Update coming out tomorrow. And while I'm sure it won't be rock solid until its .1, or even .2 release, I think it (and the flood of new 3.x-compliant apps on the way) will hold me over for a while—if not a full year, hopefully until December 12th when I qualify for standard upgrade pricing.

[JeffDM]

Originally Posted by BuffyzDead

Due to the two year contract thang, I'll be skipping the 3G S, even though I'd qualify for the discount price ~12/17/09.

This will be my next iPhone, the 3G SS OR the 4G.

You would be halfway between cycles anyway. Something that seems to be lost on people is that no one says a device has to be replaced every year. 3GS is a sturdy upgrade, but nothing that's worth losing sleep over, or getting huffy.

Originally Posted by solipsism

Not unless you jump several years into the future. We’re only getting a 7.2Mbps HSDPA radio. There is no talk of HSUPA (which surprised me) and there is still Evolved HSPA still to go in the 3G range before we get to LTE. On top of that, there are no radios that will work for phones at this point. First you need the infrastructure tested and the USB/ EC/34 cards for notebooks as power and size is less important before you can even consider trying to get these radios in a svelte phone. Progress happens in steps.

7.2 Mbps is pretty good too, hardly anything to scoff at.

[hmurchison]

I don't think we should assume that the Cortex A8 and Cortex A9 are significantly different designs. For instance back in 2007

http://www.eetindia.co.in/ART_880049...P_45ff4430.HTM

A8 refinement
The A9 core itself is a refinement of the Cortex-A8. "There are a couple of extra instructions in support of multiprocessing, but it is backward compatible," he said. Like the A8, the A9 it is superscalar, with a multi-issue eight-stage pipeline. Early branch resolution is evaluated asynchronously to instruction fetch, with continuous fetch and decode of two instructions allowed per clock cycle.

But the A9 pipeline goes further by supporting out-of-order instruction dispatch and completion. The new architecture adds to ARM's established multiprocessor capability with an accelerator coherence port supporting hardware accelerators and DMA units, support for TrustZone technology, with interrupt virtualisation, and a generalised interrupt controller.

My guess is that the improved "smarts" of A9 will lead to a design that can be clocked lower yet still be more efficient (out of order execution) saving battery power. Also I love what microprocessors (multicore) are doing with power management in shutting cores off or ramping down.

[Dick Applebaum]

Quote:
Originally Posted by Booga
My wife and I are playing leapfrog .
Nice.

Originally Posted by brucep

Nice.

Mmm... reminds me of the first rule of "Nudist Camp"

Rule 1: When playing leap-frog, one must complete the leap

[melgross]

Originally Posted by ltcommander.data

I too find it kind of funny how the title draws the direct conclusion that the next iPhone will use A9 while the first sentence of the article indicates that we are still not sure if the announced iPhone 3G S even uses the A8.

In any case, I'm of the opinion that the A9 won't show up in iPhone 2010. A major pillar of Apple's iPhone development marketing is that the iPhone/iPod Touch is a consistent platform allowing a broad user base and simplifying development. This has been true in the iPhone/iPod Touch platforms of 2007 and 2008. 2 years of relative stability and 40 million devices with ARM11+MBX is good enough, so iPhone 2009 having ARM8+SGX is a timely split in the ecosystem to allow more speed.

Similarly, it makes the most sense for iPhone 2010 to also maintain ARM8+SGX to avoid splitting the ecosystem again so soon. Seeing the iPhone 2009 didn't receive a major external redesign, it isn't hard to assume that that will be would coming to iPhone 2010 so that can serve as the main selling point rather than ARM9. As well, for Apple's first foray into in-house iPhone processor development it makes sense for them to base it on an existing design than go for the ARM9 which no one has experience with. I agree that iPhone 2010 will be the time for a Apple SoC, but I think it'll be ARM8 based so that they can use the presumably Samsung ARM8 in the iPhone 3G S as a template. And if 45nm can bring such amazing power savings to ARM9, quoted at 10-20% over ARM8, then the savings will be even greater if ARM8 were shrunk. There's nothing saying that Apple couldn't find someone to fab their own ARM8 SoC at 45nm. Apple could use the shrink to upclock the ARM8 a bit over that of the iPhone 3G S to offer more speed just like the 2nd gen iPod Touch, without splitting the ecosystem with ARM9. An ARM9 Apple SoC can come in iPhone 2011 using the experience picked up from an ARM8 SoC.

They have to advance this as much as possible. They can't wait until 2011. This is not longer in phone time, it's on computer time.

If Apple won't use the fastest, then others will. This will limit the programs, which is something Apple won't allow. It's the software after all.

And if Apple is pushing the platform as a gaming platform, which they certainly seem to be doing, then they need all the power they can get.

[melgross]

Originally Posted by hmurchison

I don't think we should assume that the Cortex A8 and Cortex A9 are significantly different designs. For instance back in 2007

http://www.eetindia.co.in/ART_880049...P_45ff4430.HTM



My guess is that the improved "smarts" of A9 will lead to a design that can be clocked lower yet still be more efficient (out of order execution) saving battery power. Also I love what microprocessors (multicore) are doing with power management in shutting cores off or ramping down.

That's possible, but don't forget that a longer pipeline allows for a higher clock.

[beamer8912]

Originally Posted by ltcommander.data

I too find it kind of funny how the title draws the direct conclusion that the next iPhone will use A9 while the first sentence of the article indicates that we are still not sure if the announced iPhone 3G S even uses the A8.

In any case, I'm of the opinion that the A9 won't show up in iPhone 2010. A major pillar of Apple's iPhone development marketing is that the iPhone/iPod Touch is a consistent platform allowing a broad user base and simplifying development. This has been true in the iPhone/iPod Touch platforms of 2007 and 2008. 2 years of relative stability and 40 million devices with ARM11+MBX is good enough, so iPhone 2009 having ARM8+SGX is a timely split in the ecosystem to allow more speed.

Similarly, it makes the most sense for iPhone 2010 to also maintain ARM8+SGX to avoid splitting the ecosystem again so soon. Seeing the iPhone 2009 didn't receive a major external redesign, it isn't hard to assume that that will be would coming to iPhone 2010 so that can serve as the main selling point rather than ARM9. As well, for Apple's first foray into in-house iPhone processor development it makes sense for them to base it on an existing design than go for the ARM9 which no one has experience with. I agree that iPhone 2010 will be the time for a Apple SoC, but I think it'll be ARM8 based so that they can use the presumably Samsung ARM8 in the iPhone 3G S as a template. And if 45nm can bring such amazing power savings to ARM9, quoted at 10-20% over ARM8, then the savings will be even greater if ARM8 were shrunk. There's nothing saying that Apple couldn't find someone to fab their own ARM8 SoC at 45nm. Apple could use the shrink to upclock the ARM8 a bit over that of the iPhone 3G S to offer more speed just like the 2nd gen iPod Touch, without splitting the ecosystem with ARM9. An ARM9 Apple SoC can come in iPhone 2011 using the experience picked up from an ARM8 SoC.

The evidence we've seen doesn't indicate that Apple will stay with the A8 processor. There are a few indicators we've seen mixed in with the 3G S phone as well as 3.0 that lead me to believe they'll change to the A9 next year.

First off, is the job listing for a programmer familiar with the ARM structure for use with the iPhone. Sure this could mean they'll be used only for the A8 but I would imagine (since they were just hired in late May) that Apple intends to stick with the ARM processors and move on down the line to the A9 come next year.

Second, and probably the most significant sign, is the appearance of the iPhone3,1 in the 3.0 firmware. If you recall, this is how we found out that the 3GS was getting a processor upgrade since it was labeled 2,1. So this would indicate a change of processors next year. (Very little else would be a call for such a major increment. Note, the 2G iPhone was 1.1 and the 3G iPhone was 1.2)

This makes me believe Apple knows exactly what the want in the next iPhone, and it is my belief that the A9 will be one of the upgrades next year.

[ZenTigerpaw]

Originally Posted by DanaCameron

Me too! Someone mentioned that the 3G S was most likely a stop gap to tide consumers over until the truly next-gen iPhone came out. Seems like they may have been right.

Reminds me of Mac OS 9.

[wizard69]

Originally Posted by addabox

Yeah.

Yeah

OK, but I thought we were talking about Apple vs. other cell phone manufacturers.

Exactly.

The point is manufactures outside of Apple will have a very large stable of processors to choose from. Nothing says Apple will have the better solution.

But it's not just I/O, it's everything happening within a constrained size/power/heat envelope. If Apple goes custom they can tailor hardware to precisely meet whatever form factor requirements their particular lineup require.

Which is I/O! In any event I don't believe for a minute that Apple will tailor make a processor for every device it sells. At best they will have a small hand full of SoC. No matter how you look at it some elements of those SoC will go unused depending on the platform.

It doesn't seem to me the current iPhone suffers from a surfeit of processing power, at any rate.



Sure, but again I thought we were talking about rival handset makers. The point of a custom Apple solution would be to tailor silicon to their specific hardware/software needs, in a way not easily replicated by their rivals. Whatever TI or Nvidia put on the market will be available to one and all.

The point I'm trying to make is that Apple can't win at that game. Nokia or whomever will always have more processors to choose from than Apple. Apple puts themselves at a disadvantage with custom solutions if they don't intend to implement custom IP. Even custom IP is nothing special if it doesn't significantly enhance the platform.

So my question is this: what IP could Apple implement that would justify the trouble of a custom A9 solution? It is a good question and honestly I have trouble coming up with the an answer.

When I do think about it I think about possibilities like these:
1.
Custom I/O processing. Most likely a small ARM to drive the GSM radio.
2.
Bringing Bluetooth and WiFi on chip. Apparently PA Semi has done really well in the past with this type of I/O.
3.
An Apple designed vector processor. Frankly I'd be surprised as they have GPU hardware in the pipeline to take care of this.
4.
A fast RAM area on chip. This would be used to support the GPU like with dedicated RAM on PCI video cards. Done right this could have a huge impact on video performance as chip to chip comms would be very fast. Also ideally this would not conflict with normal RAM access.


The above are just ideas that come up real quick. Now Apple has it's own goals of course, that likely look like none of the above, but can they do the silicon better than all of the competition? I'm not convinced they can.

So I guess the actual point of contention is "can Apple with their chip guru acquisitions/hires produce something that material outstrips the incumbent big players", to which my answer would be "they don't have to if they can produce something that's a better match for their hardware/software model, as opposed to using off-the-shelf components that are of necessity designed to serve the broadest possible market."

At the iPhone level I'm not convinced their model is that much different than the competition. In the end they will still design a processor to fit a number of applications.

Apple has different priorities from the typical handset makers (as far as I can tell), but the Nvidias and TIs are obliged to design toward the typical handset makers because that's where the market is. Hence, Apple might do very well for themselves by having more control over chips they put in their devices.

As noted above I'm still wondering what IP Apple could put into these SoC that would justify the engineering expense. I still haven't latched onto something that says this is it. As to the handful of suppliers to the cell phone market I think you are simply wrong about what will be available. Simply put there will be more variety than Apple could afford to build.

Now that is if PA Semi is even working on cell phone chips. I sometimes get the feeling they are really working on Mac chipsets. Or tablet chips, I just see more value for Apple that way. This "we bought PA to make iPhone chips" could be smoke and mirrors. Of course there is little evidence other than the Papermaster thing.



Dave