Next-Generation iPhone to Use A6 System-on-a-Chip?

[Geckotek]

no it's just proof that people can still be crappy photographers

Yeah, reminds me of "Oh, such a nice picture. You must have a really nice camera!"

 

[Mago]

You are right, not all of us could be so lucky. [...]
the A5X under load consumes more power than an entire iPhone 4S.

UNDER LOAD, and such load end as soon as the work ends, and with 1/4 of the pixels measn that the A5X uses 1/4 of the total iPhone 4S energy to complete the same task.

More, from going to a 32nm feature the chip will burn 50-67% of the power that the 45nm feature use to burn.

also a more powerful CPU means it can lower the clock and voltage saving more power...

For all this reasons I'm confident the next iPhone wil feature a 32nm A5X with 768MB Ram on SOC, and save a lot of power regard the 4s' A5..

 

[Geckotek]

UNDER LOAD, and such load end as soon as the work ends, and with 1/4 of the pixels measn that the A5X uses 1/4 of the total iPhone 4S energy to complete the same task.

More, from going to a 32nm feature the chip will burn 50-67% of the power that the 45nm feature use to burn.

also a more powerful CPU means it can lower the clock and voltage saving more power...

For all this reasons I'm confident the next iPhone wil feature a 32nm A5X with 768MB Ram on SOC, and save a lot of power regard the 4s' A5..

But why the A5x? Don't you think the graphics cores are overkill for an iPhone? That's why I'm saying a 32nm A5. Or an A6?

 

[pgiguere1]

I'm going to guess if it were *just* a different clock speed and die size, they would not call it an A6... Because we already have different clock speed and die size A5s, and they are all just called "A5."

Of course, there is more to the SoC than just CPU (and GPU), so it could be similar CPU/GPU wise to the A5 but have some other additions/features aside from die size and clock.

The different die size A5 in the newer iPad 2 isn't clocker higher and doesn't have more RAM though.

It's unclear what minimal spec boost qualifies to jump a number in the AX naming convention (it's just marketing after all), but notice they called the iPhone 4's SoC "A4" while they did not call the 3GS's anything, even though it was basically a 65nm A4 with less RAM and a lower clock speed, but the same architecture nonetheless. Both were single-core Cortex A8, just like both the A5 and A6 could both be dual-core Cortex A9, but still offer significantly different performance.

 

[Mago]

But why the A5x? Don't you think the graphics cores are overkill for an iPhone? That's why I'm saying a 32nm A5. Or an A6?

A5X is only needed for both the next iPhone and the revised new iPad, this saves a LOT of money on R&D.

Having multiple cores unused ? who cares? Intel & AMD uses to sell variants with disabled cores of the same chip just to fill the market despite the same producction cost, its more expensive to develope an entirely new chip.

 

[NT1440]

A5X is only needed for both the next iPhone and the revised new iPad, this saves a LOT of money on R&D.

Having multiple cores unused ? who cares? Intel & AMD uses to sell variants with disabled cores of the same chip just to fill the market despite the same producction cost, its more expensive to develope an entirely new chip.

Apple's plans in the future clearly display the need for more oomph with less power consumption.

Do you really think now of all times apple is going to save money on R&D in their chips after years of massive buyouts and acquisitions?

 

[cldeweese7]

What's funny is I can recall the same photoshop claims about the A5.

Hey, I'm all for the A6, just don't want to get my hopes up, haha.

 

[Geckotek]

A5X is only needed for both the next iPhone and the revised new iPad, this saves a LOT of money on R&D.

Having multiple cores unused ? who cares? Intel & AMD uses to sell variants with disabled cores of the same chip just to fill the market despite the same producction cost, its more expensive to develope an entirely new chip.

Multiple GPU cores unused. An A5 would be even cheaper to produce. This is why I think we'll see an die-shrinked A5, not an A5x. It'll be A5 or A6.

 

[patohi]

If it is an A5X then perhaps it would have 4 CPU cores instead of 4 gpu cores? Unless apple is making a harder push on the gaming front....

 

[Mago]

Multiple GPU cores unused. An A5 would be even cheaper to produce. This is why I think we'll see an die-shrinked A5, not an A5x. It'll be A5 or A6.

No, coz the new iPhone needs more graphic power for A5 achitecture, if going to A5 they will need to raise the Clock at least 33% losing the power savings of the 32nm feature, but more, the next iPhone will be restricted to incorporate new features (or to run smothly just anounced ones as the Fly Mode 3D map)....

 

[33man]

Why so many post and threads about the new ipod touch ???

The next iphone will come on 12th Sept.... So relax and be ready to be surprised, first by the invitation card, a 4.3'' size having a NFC tag on it...

Well i hope A6 for the following reasons, less consumption (if battery is still < 2000 mAh), more power and less heat...

 

[Geckotek]

No, coz the new iPhone needs more graphic power for A5 achitecture, if going to A5 they will need to raise the Clock at least 33% losing the power savings of the 32nm feature, but more, the next iPhone will be restricted to incorporate new features (or to run smothly just anounced ones as the Fly Mode 3D map)....

Why would it need that much more graphics power?? I agree, it will need a clock bump, but the A5 is already capable of that. Everything your arguing is even more true of the A5x.....

 

[NumberNine]

You're first sentence clashes horribly with the rest of your post....

Also, "better" is completely up to the user and use case. I personally dislike how busy everything in the Android UI is, including their notifications in the top bar.

Cool

 

[irDigital0l]

lol at people getting their hopes up

 

[chrmjenkins]

Cool picture, but A6 is just a marketing name, even if true we still don't know much about what the CPU will actually be, other than it won't be identical to the A5/A5X.

It could be, in order of likeliness:

1) Dual 32nm Cortex A9s clocked higher than in the A5.
2) Quad 32nm Cortex A9s.
3) Dual 45nm Cortex A9s clocked higher than in the A5.
4) New core architecture (A15s in advance?).

I doubt 3. Samsung has shown their 32nm fabs are ready for high volume with the Galaxy S3.

Given that the leaked battery only has a marginally higher capacity than the 4S's, it would seem unlikely that the CPU would draw more than the A5 given that the bigger screen and LTE chip would also draw more, resulting in a decrease of battery life which Apple would try to avoid. That's why I don't think they will throw in 4 cores or overclock the current 45nm cores.

There are a lot more unknowns than you give credit for. First, the battery is a new 3.8V chemistry that is more efficient. What that means, and how much more efficient, we really don't know. I've not seen an in-depth analysis of these new batteries yet. LG's new battery chemistry claims 60% more efficient, but we don't know if that's the same chemistry or the battery the iPhone is getting.

We also don't know the power draw of the cellular radio and the screen. How much will the LTE at 28nm draw compared to 45nm HSPA+ radios? Also, is apple prepared to take a lower LTE battery life if the 3G battery life is still the same or a little better? Because the Gobi chip is bound to be better at 3G battery life by virtue of 28nm.

For the screen, if it is indeed an in-cell display, the backlighting will likely go through some unknown adjustment. Additionally, if the screen is somehow IGZO too, that would mean a large reduction in screen energy usage.

The last one is not likely at all because even if Cortex A15 cores were ready to be produced right now, I doubt they would be in the kind of yield Apple needs with a product as successful as the iPhone.

Yields mainly follow the process evolution, not the design. A bad design is a bad design and will yield low. Masks costs millions of dollars so I don't imagine they iterate them much, if at all, for large ASIC designs, although I admit my knowledge of the fab side is limited.

I think Anantech's analysis is right, and it wouldn't be surprising to see a new marketing name (A6) even if neither the core architecture or number of cores have changed. After all, that's exactly what happened between the iPhone 3GS and iPhone 4 and they changed their marketing approach by naming the iPhone 4's SoC A4 while not giving a name to the 3GS's SoC.

I was going to say something, but then you did....

The different die size A5 in the newer iPad 2 isn't clocker higher and doesn't have more RAM though.

It's unclear what minimal spec boost qualifies to jump a number in the AX naming convention (it's just marketing after all), but notice they called the iPhone 4's SoC "A4" while they did not call the 3GS's anything, even though it was basically a 65nm A4 with less RAM and a lower clock speed, but the same architecture nonetheless. Both were single-core Cortex A8, just like both the A5 and A6 could both be dual-core Cortex A9, but still offer significantly different performance.

I think it's pretty clear what doesn't constitute a name change. Process switch and increase in GPU cores (and memory architecture to support the GPU cores) do not cause a numerical iteration. So theoretically that leaves number of cores and CPU architecture as the remaining candidates. This chip being A6 would then mean we see quad core A9 or dual core A15 most likely. However, I wouldn't be surprised to see a tweaked A9 core at a higher clock as the A6, mucking up the naming convention all together.

We also don't know the purpose of the 32nm A5 design. The 65nm A4 which is still in the iPhone 4 and the previous AppleTV never got a 45nm shrink, so why did the A5? Test the 32nm node?

For reference, the switch to a 32nm SoC in the new iPad 2 gave a ~25% battery life increase. Given that the iPad's screen draws more energy than the iPhone's proportionally to the rest of the components, the SoC's efficiency is a bit more important in the iPhone than in the iPad, and we could expect maybe a 35% battery life increase by switching to 32nm A9 cores in the next iPhone, at the same clocks. If we cancel this energy saving by overclocking by around 35%, we would get roughly the Integer/Floating point performance gains shown in the above iPhone 3GS/4 comparison.

That assumes you keep the same core voltage in both cases.

if it is in fact an A6, this will be Apple's most ambitious, aggressive iPhone update EVER. very exciting stuff!

That presumes we know something about the A6 design.

Also looks legit if you analyze the error level. The JPEG compression artefacts are the same everywhere. But of course the image could have been manipulated before it has been resized and resaved with a higher compression level.

I would argue resizing and saving almost certainly happened, given the article picture is not likely the native resolution of any camera that would have been used.

But why the A5x? Don't you think the graphics cores are overkill for an iPhone? That's why I'm saying a 32nm A5. Or an A6?

They are overkill. The A5X was a compromise design to make the retina iPad work. I don't think apple wants to use it again if they don't have to. Next iPad will have rogue graphics I'd bet.

 

[deconstruct60]

A5X is only needed for both the next iPhone and the revised new iPad, this saves a LOT of money on R&D.

Not really. The quad GPU configuration of the 5X primarily just sucks up more die space with the functionality that is already largely present in the A5. Assigning different sub GPU function units to different parts of the screen isn't a big R&D project. It isn't zero but is also isn't a significant increment.

Having multiple cores unused ? who cares? Intel & AMD uses to sell variants with disabled cores of the same chip just to fill the market despite the same producction cost, its more expensive to develope an entirely new chip.

x86 mobile and desktop chips cost $50-600. These A5 A5X chips cost $20-30. They generate much higher revenues so it is much easier to "blow" extra die space on deabled functionality. For sub $50 units there is far more pressure to maximize the number of working dies on a wafer. That means smaller dies equal lower costs and substantially higher profits.

Qurd ARM core and Quad GPUs in a phone are largely a waste. The economies of scale allow for differentiated units for larger tablets and phones.

An debate with a bit more merit would be A5X for the iPad mini even if it doesn't have a "retina" display. In that one it is overkill but the volume would help both iPad and iPad mini get shared costs down. In that context, probably would want to put the iPhone on a different, smaller, process that the iPad2 and AppleTV have already flushed out. The R&D for a smaller A5 has already been spent. There is no savings in merging everything into A5X. That opens up capacity on the old A5 line for the larger model.

A5X is problematical in that it runs too hot to put the RAM in the SoC package. The iPhone doesn't have space to blow like that (e.g, external RAM outside the SoC package).

 

[Voidness]

You don't change the best possible design. And Steve Jobs MOST definitely worked on this iPhone, along with a few years worth of products that will be introduced in the future. Hint: Apple doesn't think up a product and release it in a month, it takes a long time, and they plan it out years ahead.

I know That was sarcasm. I think it's absurd when people see this and say it's "just a taller 4S".

 

[patohi]

So it will be called the A6.
Why? Because it has to..
It will be a 4 core A9 like the galaxy s3 or maybe a 2 core A15
32nm
Sounds right

----------

Now on to yes or no NFC chip?

----------

The iPhone is too important and big for apple to be stingy on.... I'm sure they have no problem spending more on R & D for a new chip just for the new iPhone.... We could be looking at a refreshing of the whole iPhone, iPad and iPod line up with new chips for the holidays that are apples bread and butter....

 

[chrmjenkins]

Not really. The quad GPU configuration of the 5X primarily just sucks up more die space with the functionality that is already largely present in the A5. Assigning different sub GPU function units to different parts of the screen isn't a big R&D project. It isn't zero but is also isn't a significant increment.




x86 mobile and desktop chips cost $50-600. These A5 A5X chips cost $20-30. They generate much higher revenues so it is much easier to "blow" extra die space on deabled functionality. For sub $50 units there is far more pressure to maximize the number of working dies on a wafer. That means smaller dies equal lower costs and substantially higher profits.

Qurd ARM core and Quad GPUs in a phone are largely a waste. The economies of scale allow for differentiated units for larger tablets and phones.

An debate with a bit more merit would be A5X for the iPad mini even if it doesn't have a "retina" display. In that one it is overkill but the volume would help both iPad and iPad mini get shared costs down. In that context, probably would want to put the iPhone on a different, smaller, process that the iPad2 and AppleTV have already flushed out. The R&D for a smaller A5 has already been spent. There is no savings in merging everything into A5X. That opens up capacity on the old A5 line for the larger model.

A5X is problematical in that it runs too hot to put the RAM in the SoC package. The iPhone doesn't have space to blow like that (e.g, external RAM outside the SoC package).

Good points. Many people forget the heat issues that forced the RAM off package in the iPad 3. Even if a 32nm retired that issue, it's still a design that is too large by compromise. They'd much rather have had a rogue part I'm sure.

The iPhone is too important and big for apple to be stingy on.... I'm sure they have no problem spending more on R & D for a new chip just for the new iPhone.... We could be looking at a refreshing of the whole iPhone, iPad and iPod line up with new chips for the holidays that are apples bread and butter....

It's possible they could do cycled risk. 32nm A5 proved the process. Now A15 based design proves the core. Then, iPad design proves the new GPU. I doubt they view that as necessary, though.

 

[deconstruct60]

Yields mainly follow the process evolution, not the design. A bad design is a bad design and will yield low. Masks costs millions of dollars so I don't imagine they iterate them much, if at all, for large ASIC designs, although I admit my knowledge of the fab side is limited.

A design can be "bad" for yields if it is larger. A die that is 5x5mm versus 10x10mm will have different yields from respective different single wafers given the same number and location of defects appear on each wafer involved.

If the processors are smaller the defects impacts fewer processors.

The other issue here is that typically contract fabs charge per wafer. If there are two or 200 working dies on the wafer you pony up the same price to have it processed.

They are overkill. The A5X was a compromise design to make the retina iPad work. I don't think apple wants to use it again if they don't have to. Next iPad will have rogue graphics I'd bet.

Rogue graphics? How about just fewer, but higher throughput GPUs ? That doesn't make them "rogue" just 'next generation'.

 

[DPazdanISU]

I'm hoping more cell phone carriers offer the iPhone 5...

 

[chrmjenkins]

A design can be "bad" for yields if it is larger. A die that is 5x5mm versus 10x10mm will have different yields from respective different single wafers given the same number and location of defects appear on each wafer involved.

If the processors are smaller the defects impacts fewer processors.

The other issue here is that typically contract fabs charge per wafer. If there are two or 200 working dies on the wafer you pony up the same price to have it processed.

Actually, the same number of processors would be affected each defect can only affect 1 processor.

Rogue graphics? How about just fewer, but higher throughput GPUs ? That doesn't make them "rogue" just 'next generation'.

I think you misunderstood. 'Rogue' is ImgTec's codename for their next generation GPU core. They are much higher throughput than the current SGX5xx cores.

I'm hoping more cell phone carriers offer the iPhone 5...

T-mobile's training note pretty much confirmed they aren't, so it will just be pre-paid and regional carriers if any are added in the US.

 

[deconstruct60]

Actually, the same number of processors would be affected each defect can only affect 1 processor.

5x5mm and 10x10mm coming off the same size wafer there will be twice as many of the first kind. So the percentage impact will be much smaller. It can't be the same because the number of units get off when there are no defects is different.

One of the significant problems here is that Apple is standing in line with everyone else to get wafers through. Pragmatically there is a maximum number of wafers they can put through the system. If the demand for their product outstrips the number of wafers they can put there then they are going to leave alot money on the table. Android phone now outsell iPhones. If Windows Phone comes back in a significant way... that too will reduce the empty wafer queue slots.

I think you misunderstood. 'Rogue' is ImgTec's codename for their next generation GPU core. They are much higher throughput than the current SGX5xx cores.

OK. Power VR Rogue . It is a bit early for those. For the iPad 2013 sure, but the window between when ImgTec released that 600 series (Janurary 2012 ) and when the next iPhone was in the design freeze stage was pretty small. Apple tends to be a bit conservative with technology adoption.

I wouldn't be surprised if there was a prototype A6 (if the change is graphics is what is driving the number bump) floating around in a prototype iPhone but it may not be the what's going to be released.

 

[chrmjenkins]

5x5mm and 10x10mm coming off the same size wafer there will be twice as many of the first kind. So the percentage impact will be much smaller. It can't be the same because the number of units get off when there are no defects is different.

Yes, percentage is smaller, but raw number affected is unchanged.

One of the significant problems here is that Apple is standing in line with everyone else to get wafers through. Pragmatically there is a maximum number of wafers they can put through the system. If the demand for their product outstrips the number of wafers they can put there then they are going to leave alot money on the table. Android phone now outsell iPhones. If Windows Phone comes back in a significant way... that too will reduce the empty wafer queue slots.

Less true with Samsung essentially giving them a dedicated fab. Also, relative growth of android and WP doesn't matter from a supply perspective provided the overall demand for phones is the same. It would only be a leverage/negotiation issue.

OK. Power VR Rogue . It is a bit early for those. For the iPad 2013 sure, but the window between when ImgTec released that 600 series (Janurary 2012 ) and when the next iPhone was in the design freeze stage was pretty small. Apple tends to be a bit conservative with technology adoption.

I wouldn't be surprised if there was a prototype A6 (if the change is graphics is what is driving the number bump) floating around in a prototype iPhone but it may not be the what's going to be released.

Yes, I'm talking next iPad for Rogue.

 

[patohi]

It's possible they could do cycled risk. 32nm A5 proved the process. Now A15 based design proves the core. Then, iPad design proves the new GPU. I doubt they view that as necessary, though.

And remember, Apple will be putting that new 9 pin iOS connection on the iPhone and iPad mini..... So they could refresh the new iPad and drop the iPad 2.... A refresh for chips is possible!

An A6x in the iPad retina

A6 in the iPhone

The 32nm A5 in the iPad mini and iPod

All new 9 pin connectors