Next-Gen iPhone & iPhone Nano Speculation

[mczak]

Cortex-A15 is very unlikely. Rather a quad core A9 or, more probable, a higher frequency dual A9.

Yeah with the claim of "2xCPU, 2xGPU" this sounds very much like dual-core Cortex-A9, SGX543MP4. Dual-core A9 because A15 isn't ready yet, and apple doesn't brag about core count hence a dual-core Cortex-A9 at around 1.5-1.6Ghz is way more useful than a quad core at below 1Ghz would be. Samsung can do Cortex-A9 (even quads) clocked that high on 32nm (in the Exynos chips), and TSMC claimed they can do dual-core Cortex-A9 at 3.1Ghz (ok stated suitable for mobile "only" up to 2Ghz), I don't know if A6 is 32nm Samsung or 28nm TSMC if that rumor ever cleared up but I don't see any reason why the A6 couldn't clock it that high.
My guess is they've also increased memory bandwidth by a factor of 2 (maybe doubling frequency IIRC the A5 ran the lpddr2 at 266Mhz but the standard says up to 533Mhz is possible).
As for graphics it seems it's too early for Rogue. I'm not sure if a sgx543mp2 at 400Mhz wouldn't be possible (intel clocks their sgx parts that high after all) but as long as a sgx543mp4 at 200Mhz uses less power this seems more likely. After all apple claims the chip is 22% smaller than the original (presumably) A5, which would put it right around 100mm². That's way bigger than the shrunk A5, so somewhere that space needs to go .

 

[Mike11]

If the A6 is just Cortex-A9, then what's Apple gonna call the SoC in next year's iPad, presumably Cortex-A15 based? A6X? That would be confusing. A7? From A6 to A7 in 6 months?

 

[tangey]

It's the gpu that seems incredible...2x a5 iPhone is stupidly fast....must be clocking that mp2 @ 500mhz...

You are confusing ipad2 and iphone4s. Iphone4s clocks the graphics @200mhz, so new one will be running at 400mhz (assuming its mp2, which I assume it is).

With regard to CPU, I think it's 1.6 dual A9, on the basis that apple would definitely make a big thing out of either quad core or a move to A15.

 

[french toast]

Are you sure one phone contains support for all those bands? My understanding was that for LTE, it wasn't just the bands but amplifiers and antennas.

Anyways, iPhone supports all the DC-HSPA carriers in Europe, apparently.

Which baseband is Nokia using? Apple is obviously using Qualcomm's.

Qualcomm msm8960.

 

[french toast]

You are confusing ipad2 and iphone4s. Iphone4s clocks the graphics @200mhz, so new one will be running at 400mhz (assuming its mp2, which I assume it is).

With regard to CPU, I think it's 1.6 dual A9, on the basis that apple would definitely make a big thing out of either quad core or a move to A15.

I wssnt getting mixed up..just got the frequency wrong
Yea I think we might be seeing current gen technology at work here, which isn't a bad thing as apple optimises the hell out of it for best performance.

Looking forward to seeing that safari web browser in ios 6 take ad advantage of it..could well be seeing some sunspider records broken

 

[Kaotik]

Are you sure one phone contains support for all those bands? My understanding was that for LTE, it wasn't just the bands but amplifiers and antennas.

Anyways, iPhone supports all the DC-HSPA carriers in Europe, apparently.

Which baseband is Nokia using? Apple is obviously using Qualcomm's.

Yes, all Lumia 920's sold anywhere in the world support GSM 800/900/1800/1900, WCDMA 850/900/1900/2100 and LTE 800/900/1800/2100/2600

 

[anexanhume]

From @nerdtalker (Brian Klug of Anandtech):

Everyone keeps asking - the A15 claim isn’t based on the 2x number but other sources as well, 2x claim just backs it up.

https://twitter.com/nerdtalker/status/245977097162592256

It all depends on how Apple came up with 2X CPU. In order for it to be a higher clocked A5, the CPUs would have to be clocked at 1,6 GHz wich is pretty high even with a dieshrink. Now add in LTE on essentially the same old size battery and iPhone might no longer be known for its battery life

2X CPU could also be quad A9s (marketing wise)

Or its dual 800 MHz Cortex A15s

They also provided app specific speed-ups:

Those don't seem well-suited to a huge quad core speedup, so it's gotta be juiced dual core A9 or a dual core A15 in the 1 to 1.2 GHz range. (assuming A15 core is 1.4 times faster than similarly clocked A9 coming from ARM. In fact, 1.2/.8 * 1.4 = 2.1 for lols)

edit: Anand is reiterating their A15 claims:

Can't provide more hints on the A15 stuff. Think about voltage/frequency curves, also think about perf gains promised as to why not A9.

https://twitter.com/anandshimpi/status/245994652531036160

 

[french toast]

From @nerdtalker (Brian Klug of Anandtech):

https://twitter.com/nerdtalker/status/245977097162592256

If it is cortex a15..which I think is a 50/50..then it will be running at 1.2ghz I think....web browsing will be lightning fast on this device.

 

[mczak]

From @nerdtalker (Brian Klug of Anandtech):

https://twitter.com/nerdtalker/status/245977097162592256

Hmm ok. Well in my book 2x CPU claim doesn't really back this up as it would mean it's clocked so low (around 1Ghz?) that you could easily achieve the same performance with dual-core A9 so why even bother with A15?
Though I guess if it's low-clocked dual-core A15 and low-clocked sgx543mp4 it would make a very good chip for a new iPad (just jack up the clocks by 50%...) but the trouble is a new iPad is still far away...

 

[ams]

To get 2x cpu and double gpu performance in the same or better power envelope with lte ALL in a crazy thin 7.6mm chassis is quite frankly ridiculous....so much so I'm not sure I can believe it's possible on current technology.

What is so surprising about this? The CPU performance of iphone 4s was quite ordinary, with that particular [A5] SoC using dual core A9 processors clocked at only 800MHz. So doubling that CPU performance (by using dual core A9 processors clocked at 1.6GHz) while keeping power consumption under control would be very easy to do with a die shrink. Many (including myself) incorrectly assumed that the CPU performance would be no better than ipad 3 because it really didn't need to be any better for the iphone 5 to be a hit with consumers. But there is much stronger competition this year with dual core Krait processors (and quad core A9 processors) used already in many phones, so that may have been reason enough for Apple to enhance the CPU performance (without needing to increase die size devoted to the CPU). Now, with respect to GPU performance, most people correctly anticipated that iphone 5 would have approximately 2x performance improvement vs. iphone 4s. That means that GPU performance on A6 with iphone 5 is no better (if not just a bit slower) than on A5X with ipad 3. The A6 iphone 5 GPU (which could be a die shrunk SGX543MP2 running at approximately 2x operating frequency vs. A5 iphone 4s GPU, or could be a die shrunk SGX543MP4 running at slightly reduced operating frequency vs. A5X ipad 3 GPU) should be similar in overall performance to Adreno 320, which will be shipping in an LG phone within the next 1-2 months. To get a general idea of how close the GPU performance will be, see here: http://images.anandtech.com/graphs/graph6185/49438.png , http://images.anandtech.com/graphs/graph6185/49440.png . Now, with respect to the very thin chassis of the iphone 5, that is impressive, but this is largely due to the use of in-cell display technology from LG/Sharp/etc.

 

[anexanhume]

Hmm ok. Well in my book 2x CPU claim doesn't really back this up as it would mean it's clocked so low (around 1Ghz?) that you could easily achieve the same performance with dual-core A9 so why even bother with A15?
Though I guess if it's low-clocked dual-core A15 and low-clocked sgx543mp4 it would make a very good chip for a new iPad (just jack up the clocks by 50%...) but the trouble is a new iPad is still far away...

Based on crude math of 1.4 times the DMIPS/MHz for A15 versus A9 you'd need a 1.2 core clock versus 0.8 to get to 2.1x. I think 1.2 is ok, even though it's on the low end of the high end SoCs out there.

 

[liolio]

Apple stated thatthe new chip is 22% tinier than the previous A5 (I guess @45nm), how does that compare to the the A5 @32nm and the A5x?
Sorry I'm moving to work can do the searches and math now and my self

 

[french toast]

What is so surprising about this? The CPU performance of iphone 4s was quite ordinary, with that particular [A5] SoC using dual core A9 processors clocked at only 800MHz. So doubling that CPU performance (by using dual core A9 processors clocked at 1.6GHz) while keeping power consumption under control would be very easy to do with a die shrink. Many (including myself) incorrectly assumed that the CPU performance would be no better than ipad 3 because it really didn't need to be any better for the iphone 5 to be a hit with consumers. But there is much stronger competition this year with dual core Krait processors (and quad core A9 processors) used already in many phones, so that may have been reason enough for Apple to enhance the CPU performance (without needing to increase die size devoted to the CPU). Now, with respect to GPU performance, most people correctly anticipated that iphone 5 would have approximately 2x performance improvement vs. iphone 4s. That means that GPU performance on A6 with iphone 5 is no better (if not just a bit slower) than on A5X with ipad 3. The A6 iphone 5 GPU (which could be a die shrunk SGX543MP2 running at approximately 2x operating frequency vs. A5 iphone 4s GPU, or could be a die shrunk SGX543MP4 running at slightly reduced operating frequency vs. A5X ipad 3 GPU) should be similar in overall performance to Adreno 320, which will be shipping in an LG phone within the next 1-2 months. To get a general idea of how close the GPU performance will be, see here: http://images.anandtech.com/graphs/graph6185/49438.png , http://images.anandtech.com/graphs/graph6185/49440.png . Now, with respect to the very thin chassis of the iphone 5, that is impressive, but this is largely due to the use of in-cell display technology from LG/Sharp/etc.

Yes I can see the performance and battery life is doable on a 4s chassis and baseband...if juggled about on 32nm...

What makes it crazy good is those performance improvements..combined with a larger more colourfull higher res screen, lte connectivity and claimed/reported BETTER battery life...in a human hair like 7.6mm....you have to admit that is ridiculous advancement with out using next gen battery/cpu/gpu components..even with 32nm and in cell.

 

[Jubei]

Hmm ok. Well in my book 2x CPU claim doesn't really back this up as it would mean it's clocked so low (around 1Ghz?) that you could easily achieve the same performance with dual-core A9 so why even bother with A15?
Though I guess if it's low-clocked dual-core A15 and low-clocked sgx543mp4 it would make a very good chip for a new iPad (just jack up the clocks by 50%...) but the trouble is a new iPad is still far away...

Wich do you think draws more power? 800 MHz A15 or 1,2-1,6 GHz Cortex A9. The answer is probably the same one Apple picked

Also Apple probably gets other benefits from using next gen architecture. Samsung claimed Exynos 5 improved memory bandwidth from 6 GB to 12.8 GB/s

 

[mczak]

Wich do you think draws more power? 800 MHz A15 or 1,2-1,6 GHz Cortex A9. The answer is probably the same one Apple picked.

I've got some doubts about this. The A15 is said to be significantly more complex than the A9, and such complexity doesn't come for free.

Also Apple probably gets other benefits from using next gen architecture. Samsung claimed Exynos 5 improved memory bandwidth from 6 GB to 12.8 GB/s

I don't think that's got much to do with it. You need the memory bandwidth for the gpu mostly, and I can't see why you couldn't do that with Cortex-A9 (the memory controller is external to the cpu cores anyway).

edit: there's also other reasons why I don't like the idea of low-clocked A15. The risk of using a not yet proven cpu design just doesn't seem worth it if it cannot outperform the proven design anyway. Which actually would also be a reason why it could end up that way - maybe the idea would have been to clock it higher but it just couldn't meet the required power target for some reason (because of design errors or whatnot) - ok that is very speculative but I'll note the only A15 which was shown in action as of yet in some cpu prototype (was that omap5?) also couldn't reach reasonable clocks.

 

[anexanhume]

I've got some doubts about this. The A15 is said to be significantly more complex than the A9, and such complexity doesn't come for free.


I don't think that's got much to do with it. You need the memory bandwidth for the gpu mostly, and I can't see why you couldn't do that with Cortex-A9 (the memory controller is external to the cpu cores anyway).

Yeah, I wonder if they are using LPDDR3 to feed the GPU faster.

 

[wco81]

I wonder if the 16:9 aspect ratio will be pushed to iPads eventually. The current 4:3 ratio is kind of a stand out now, with most Android tablets -- and W8 tablets IIRC -- being 16:9.

Also, thinner is nice but that constraint seems to limit how good the camera or other components could be. I think they would have gotten away with same thickness, resulting in greater instead of less volume than the 4S.

 

[anexanhume]

I wonder if the 16:9 aspect ratio will be pushed to iPads eventually. The current 4:3 ratio is kind of a stand out now, with most Android tablets -- and W8 tablets IIRC -- being 16:9.

Also, thinner is nice but that constraint seems to limit how good the camera or other components could be. I think they would have gotten away with same thickness, resulting in greater instead of less volume than the 4S.

I would have preferred that with more battery, but I think as long as the performance and battery life metrics stay the same, they want to keep making it thinner. I suppose I don't mind as that means a case on it is even less thick comparatively than before.

As for 16:9, I don't think it's happening for the iPad. At least not anytime soon. Doesn't seem to mesh with the usage model they push.

 

[Laurent06]

I'll note the only A15 which was shown in action as of yet in some cpu prototype (was that omap5?) also couldn't reach reasonable clocks.

Yes, but IIRC Exynos5 was also recently demo'ed.

 

[mczak]

Yes, but IIRC Exynos5 was also recently demo'ed.

That is true though as far as I can tell there was no confirmation of clock speed on the demo unit (ultimately the chip should run at up to 1.7Ghz).
But anyway I guess we should find out soon enough about the A6 chip.