Next-Gen iPhone & iPhone Nano Speculation

[anexanhume]

Apple may feel the need to bump to quad CPU on 32nm, but I don't see marketing driving any of their decisions. They've always been focused on core functionality and aesthetics. Specs have always been a consequence of the functions they want, not the other way around.

I also have a hard time believing they need quad given iOS's restrictive multitasking APIs.

 

[Pressure]

Pretty sure the marketing people are not the ones running Apple.

They could simply use a shrinked A5X and proclaim it to be the fastest phone for gaming etc.

Nintendo is not the driving force any longer for portable gaming.

It's a phone, it will be fast enough until Cortex-A15 or a custom silicon can replace the current dual-core Cortex-A9.

 

[tangey]

Pretty sure the marketing people are not the ones running Apple.

I'm pretty sure marketability plays a big part in all of apple's products. I'll say one word. Siri.

They could simply use a shrinked A5X and proclaim it to be the fastest phone for gaming etc.

In graphics terms alone, that would make no sense at all to me, Its highly likely they went with SGX543MP4 in the A5X because they could not run an MP2@500Mhz @45nm.

They could replace the 543MP4 with 543MP2, increase clock to 500Mhz and save a significant area of die, and maybe run a bit quicker. SGX54xMP2 @ >=500Mhz has been announced by other SGX licencees @32nm.

Although Apple is clearly not terribly concerned with die area, to be able to reduce the current design by arond 20% must be important when you are planning a chip that might number well north of 150M units.

 

[dagamer]

Well there had been several whispers in the past that Microsoft evaluated (amongst other solutions as always) also an ARM based SoC; unfortunately no one apart from true insiders knows what they've really gone for in the end. So far happy go merry rumor mongerers point repeatedly in IBM's direction for the CPU and frankly I'm not so sure any console developer would be all that happy having to deal with any API overhead suddenly for consoles.

In any case even for the above theory and unless I'm having another blond moment I don't see any direct benefits for IMG if they haven't landed a GPU IP deal for any of the upcoming consoles (most likely and I mean Microsoft and SONY and not any indirect possible Apple future plans).

I'm not saying that an ARM SoC was the only chip in there, but in addition to a PowerPC CPU and Radeon GPU. Considering the chips only cost $25 (and even less in bulk), it gives the next Xbox easy access to all of the Windows 8 apps without having to try and port Windows 8 to PowerPC. And it's far more power efficient than other chips, so it can handle non-Xbox 720 game tasks with ease. Then Microsoft can pivot and say that all Xbox 720 indie games run on Windows 8, buy the game once, play it anywhere, and boom instant pay day.

It's the freaking developer holy grail. Make the app once, adapt super easily for different platforms (Win8 tablets, Windows Phone 8 devices, xbox 720), and developers will flock like no other.

 

[wco81]

It's the freaking developer holy grail. Make the app once, adapt super easily for different platforms (Win8 tablets, Windows Phone 8 devices, xbox 720), and developers will flock like no other.

Why would this be easier than developing for Android, where developers have to test their software on dozens of popular devices?

W8 devices will span a broader range of HW features, profiles, etc.

More likely they will try to ensure compatibility or target a handful or two of the most popular devices or SOC platforms.

 

[ltcommander.data]

In graphics terms alone, that would make no sense at all to me, Its highly likely they went with SGX543MP4 in the A5X because they could not run an MP2@500Mhz @45nm.

They could replace the 543MP4 with 543MP2, increase clock to 500Mhz and save a significant area of die, and maybe run a bit quicker. SGX54xMP2 @ >=500Mhz has been announced by other SGX licencees @32nm.

Although Apple is clearly not terribly concerned with die area, to be able to reduce the current design by arond 20% must be important when you are planning a chip that might number well north of 150M units.

I wonder does SGX554MP2 offer advantages? Since the next iPhone isn't expected to increase pixel count appreciably, going SGX554MP2 instead of SGX543MP4 allows them to double ALU performance in a more space efficient manner while avoiding aggressive clock speeds needed if sticking with a SGX543MP2. Apple's primary concern probably remains battery life, so it'd be interesting to know whether a low clock speed SGX554MP2 solution would be more power efficient than a high clock speed SGX543MP2. Then if Apple wants to use a lower risk iterative approach to SoC improvement they could potentially go dual Cortex A9 + SGX554MP2 for iPhone 5, dual Cortex A15 + SGX554MP4 for iPad 4, and dual Cortex A15 + Rogue for iPhone 6.

 

[Pressure]

Scratch that, misread.

 

[tangey]

I wonder does SGX554MP2 offer advantages? Since the next iPhone isn't expected to increase pixel count appreciably, going SGX554MP2 instead of SGX543MP4 allows them to double ALU performance in a more space efficient manner while avoiding aggressive clock speeds needed if sticking with a SGX543MP2.

Frankly way above my abilities to make a guess about that. All I know is that to date no one has publicly licensed 554, whereas every other announced SGX ip has been licensed multiple times. One assumes its been developed with a licencee in mind, whether it's apple, intel, or someone else, who knows. All I will say is that it contains some extra circuitry for dx9 compliance that apple have seen no need to require to date, so from purely that point of view, it might not be as efficient.

 

[ltcommander.data]

Frankly way above my abilities to make a guess about that. All I know is that to date no one has publicly licensed 554, whereas every other announced SGX ip has been licensed multiple times. One assumes its been developed with a licencee in mind, whether it's apple, intel, or someone else, who knows. All I will say is that it contains some extra circuitry for dx9 compliance that apple have seen no need to require to date, so from purely that point of view, it might not be as efficient.

That is true. I didn't think about the unneeded DX9 compliance.

http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested/1

Anand has tested the 32nm A5 shrink. 16% battery life increase in web browsing, 29% in gaming, and 19% in video decode. Very impressive considering all the other components are presumably unchanged. There is definitely room to increase clock speeds if sticking with SGX543MP2 for iPhone 5, but I guess there isn't an easy way to tell if ~2x GPU clock is doable. With the GPU performance of the Exynos 4412, if Apple wants to definitively regain the GPU performance crown, they probably need to aim for at least a 2x performance increase.

 

[Ailuros]

That is true. I didn't think about the unneeded DX9 compliance.

http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested/1

Anand has tested the 32nm A5 shrink. 16% battery life increase in web browsing, 29% in gaming, and 19% in video decode. Very impressive considering all the other components are presumably unchanged. There is definitely room to increase clock speeds if sticking with SGX543MP2 for iPhone 5, but I guess there isn't an easy way to tell if ~2x GPU clock is doable. With the GPU performance of the Exynos 4412, if Apple wants to definitively regain the GPU performance crown, they probably need to aim for at least a 2x performance increase.

Will the 4412 be more than 10-20% faster from the GPU side than the iPhone4S? Wouldn't it rather come down to as to what kind of performance difference Apple is really aiming for in the end?

No I don't know what Apple's plans are, but if they won't go for a A5X shrink under 32nm for the next iPhone and they'll increase frequency of a MP2 by say 50%, I'd still suggest that it'll be comfortably faster than a 4412 for the GPU. In fact it could bank right in between iPad2 and iPad3 performance for Egypt GLBenchmark2.1 and we'll see how GPUs will fair in future synthetic benchmarks with higher shader complexity like GLBenchmark2.5.

 

[ltcommander.data]

Will the 4412 be more than 10-20% faster from the GPU side than the iPhone4S? Wouldn't it rather come down to as to what kind of performance difference Apple is really aiming for in the end?

No I don't know what Apple's plans are, but if they won't go for a A5X shrink under 32nm for the next iPhone and they'll increase frequency of a MP2 by say 50%, I'd still suggest that it'll be comfortably faster than a 4412 for the GPU. In fact it could bank right in between iPad2 and iPad3 performance for Egypt GLBenchmark2.1 and we'll see how GPUs will fair in future synthetic benchmarks with higher shader complexity like GLBenchmark2.5.

I was thinking that given the iPhone 5's late September/October expected release, Apple would want to pad their GPU performance increase to meet potential challengers in the Exynos 5 series Mali-T604, Qualcomm S4 Adreno 320, and similar SGX544MP2 in OMAP5 which will all presumably be available later this year. Assuming Apple continues to aim to have top GPU performance on release of course. They may well be aiming for a more incremental performance increase and focusing their attention on the case redesign as the key product differentiator this year.

 

[Pressure]

You should also have in mind that the added graphic capabilities would just be to enhance eye-candy, as iOS is v-synced at 60Hz.

But yeah, console quality games on your phone is fast approaching, although for my own personal needs that have been totally replaced by the iPad so I am much more interested in that SoC.

Not sure I need much more from my phone at this point until iOS 6 or a new version of Windows and Android needs it.

Right now it basically just feels like a bulletin point you have to check off that you have quad-core etc. Doesn't really enhance the user experience.

 

[Arwin]

You should also have in mind that the added graphic capabilities would just be to enhance eye-candy, as iOS is v-synced at 60Hz.

But yeah, console quality games on your phone is fast approaching, although for my own personal needs that have been totally replaced by the iPad so I am much more interested in that SoC.

Not sure I need much more from my phone at this point until iOS 6 or a new version of Windows and Android needs it.

Right now it basically just feels like a bulletin point you have to check off that you have quad-core etc. Doesn't really enhance the user experience.

Same here, and it will stay that way for iOS devices for me at the very least until they introduce iPhones with bigger screens.

Though personally, having spent the first few weeks with my iPad 3 now (basically current fastest iOS device right?), I'd say that console quality games on phone are still a ways off. Crytek's thingy for instance looks nice, but has a fixed camera and everything and not a very smooth framerate. Will be a while yet before they reach Uncharted level, even the Vita version, and by the time they do, I expect next-gen consoles to be released.

 

[sebbbi]

Though personally, having spent the first few weeks with my iPad 3 now (basically current fastest iOS device right?), I'd say that console quality games on phone are still a ways off. Crytek's thingy for instance looks nice, but has a fixed camera and everything and not a very smooth framerate. Will be a while yet before they reach Uncharted level, even the Vita version, and by the time they do, I expect next-gen consoles to be released.

Have you tried Modern Combat 3? It looks pretty much on par to most PS Vita games at least, and is butter smooth on iPad 2 at least.

 

[ltcommander.data]

Have you tried Modern Combat 3? It looks pretty much on par to most PS Vita games at least, and is butter smooth on iPad 2 at least.

The upcoming NOVA 3 uses an improved Modern Combat 3 engine with more dynamic lighting and dynamic shadowing, but they are also specifically pointing out that it was designed with dual cores in mind. Anand's review of the iPad 3 found that some iPhoto tasks can already max out the dual 1GHz cores which is the fastest CPU (along with the iPad 2) on an iOS device. If we are thinking that GPU performance is already sufficient, I guess this begs the question of whether Apple's next push should be improved CPU performance? Keeping CPU performance stagnant for another generation/year could stall future cutting-edge app development now that the latest apps are already starting to make good use of current dual cores. A 50% CPU clock speed bump from 800MHz to 1.2GHz for the next iPhone 5 sounds significant, but if apps are beginning to peg 1GHz dual cores, an additional 200MHz may not be sufficient. Even more aggressive clock speed bumps or going quad core are options, but it might be worthwhile just to go to dual Cortex A15 this year, even if Rogue is not ready, and sticking with a Series 5XT GPU similar to the OMAP5.

 

[wco81]

Isn't the assumption that Apple couldn't do much with the CPU on the A5X because they were still on the larger die and most of their cost went to supporting the new display?

They could have waited a quarter or so and had access to improved lithography (and possibly A15) but then it would mess up their scheduling (iPad in the first quarter, iPhone later in the year).

They are pushing the screen on this iPad as the tentpole feature, as Siri was for the 4S.

 

[Arwin]

Have you tried Modern Combat 3? It looks pretty much on par to most PS Vita games at least, and is butter smooth on iPad 2 at least.

I haven't, but partly because I hate CoD and couldn't get through Battlefield 3 either. Anyway, even if this looks decent and smooth, there's no shadows or lighting happening at all. Nothing. While plaudible for being smooth and clean, and to some extent making the baked lighting work well for them. If they're planning on adding some of that for the next game they're working on, let's see where that goes. So it's still catching up to Vita as far as I am concerned, which means it's still a ways away from even current gen, which was my point. Mind you, it will eventually become viable even for bigger games no doubt, even if the controls currently still hold it back badly in some genres (which is where Android has a big advantage).

But yeah, one or two gens further and we'll probably see some cool stuff. I agree that iPad 3's priority was the screen, and having got it, it was worth the priority, even simply because I bought it for 2D stuff, not 3D. The screen is spectacular, and I love reading stuff on it.

 

[dagamer]

Why would this be easier than developing for Android, where developers have to test their software on dozens of popular devices?

W8 devices will span a broader range of HW features, profiles, etc.

More likely they will try to ensure compatibility or target a handful or two of the most popular devices or SOC platforms.

The problem with Android is that with games, the different SoCs require different texture types. As far as I know with Windows, there's a standard DirectX texture compression format which simplifies a lot work.

 

[Lazy8s]

While I haven't played Modern Combat 3 in a while, I'm pretty sure it was doing some shader and lighting effects.

iPhoto for iOS is more of a highlight of the need for GPGPU utilization than for CPU enhancement, which is a shame considering how well, in general, Apple has accelerated the rest of the OS/apps and how they proposed the OpenCL spec in the first place. Though I expect OpenCL for iOS in the near future, iPhoto could've used some extra polishing.

The very same goes for the compositing of the HDR photo in the camera app.

 

[tangey]

BBC online today posted a piece about IMG, including an interview with the CEO, Hossein Yassaie.
http://www.bbc.co.uk/news/business-17939014

There's an interesting one-liner commentary by the author towards the end of the piece.

"For example, Apple's popular voice command system, Siri, is driven by the GPU."

Now I have no way of knowing if that is factually correct, but I can say that everything else in the article is spot on, and unlike many media outlets correctly identify IMG as licensing IP where many media outlets still write that IMG sells chips.

Assuming it is correct, it likely makes this the first widely deployed GPU compute application. I wonder do Apple have internal openCL in ios 5, and are just not exposing it at this time ?