Next-Gen iPhone & iPhone Nano Speculation

[wco81]

Browsing is more often than not limited by single threaded performance (Javascript).

More RAM could reduce reloading of previously loaded pages too.

 

[Grall]

At least you have finally recognise that Apple will go this route (irrespective of the reasons behind it).

I haven't "admitted" anything. My stance has been entirely consistent the whole time.

Any advantages with quadcores in a phone are largely non-existant, due to the lack of actual need for it. I've asked you for specifics regarding the need for quad-cores and you simply refuse to respond, meaning I take it you don't actually know any answers. Do you even know of one single "killer app" for quad cores in a phone...?

There's corner cases where you benefit from more than two, but they're corner cases due to the fact you just don't do heavy processing on a phone because that is largely pointless and cumbersome, and you burn out the battery fast which means then you can't take or make calls, which is dumb since that's the primary function of a phone...

Apple will do what apple does, all in its own good time. Most, if not all analysts expect the next iphone to again be a dual-core device, which very likely means apple will still be dual-core for the entire next following year, if they keep up their current release schedule of only one update annually. This despite your percieved "market pressure".

Like I said in my edit, expect quad cores in ipad well before iphone, because a tablet is vastly more suited to content generation and processing. It's also easier to fit a big SoC in a tablet, obviously. Maybe it'll happen this coming update, I don't know - and neither do you. I still wouldn't bet any money I couldn't afford to lose on it happening this year, though.

Now, unless you actually can mention some concrete examples of obvious advantages for quad-cores in a phone other than your silly claims of "games" (what phone games precisely need, or even use quad-cores? Would you seriously even notice the difference ON A PHONE SCREEN?) or "browsing" (which isn't even accurate due to javascript), then this pointless discussion is over.

 

[RudeCurve]

Let the rest race to 8-core and other silly solutions.

+1...Smart design is better than Stupid brute force design.

 

[silent_guy]

Irrespective of whether you lean one way or the other, can you stop calling each-other trolls? I see efforts to come with factual reasons from both of you. And both of you seem to agree that Apple will go 4 core at one point or the other, so what does it matter?
It's not as if Apple is going to say "we're going 4 core because of marketing reasons only": they'll always give some technical reason also, true or not.

All I know is that iOS 7 will support true multitasking, which should make it at least a little bit easier to come up with use cases that use more CPUs...

 

[Exophase]

I'd rather believe apple would introduce some littleBIG scheme with one or perhaps two weaker cores alongside dual main cores instead into their SoC than expect a straight, symmetrical quadcore design.

I personally don't think that Apple (or Qualcomm, nVidia, or Intel) will do something like big.LITTLE. A15's design is rather above the ideal power vs efficiency point for phones/tablets and A7's is somewhat below. Put together the hope is they'll more or less average out. But the reason why this makes sense for ARM is they're simultaneously targeting totally different markets where A7 and A15 alone make sense, while Apple and the like are mainly targeting phones/tablets directly. And their designs are closer to the sweet spot.

It's not that the approach can't potentially provide more benefit than not doing it, it's just a big endeavor to develop two completely separate uarchs and deal with all the software issues too. If it does happen it may be more likely that it'll be their core + vanilla A7, although I doubt that too (if anyone would do it'd be Qualcomm)

They might do asynchronous voltage and clock domains, though (like Qualcomm).

 

[anexanhume]

They might do asynchronous voltage and clock domains, though (like Qualcomm).

Are you suggesting that they don't already?

 

[Exophase]

Are you suggesting that they don't already?

I don't actually know if they do or not. I kind of doubt it.

 

[anexanhume]

I don't actually know if they do or not. I kind of doubt it.

Why do you doubt it? But I'm sure you can understand my confusion, as to my knowledge there are no profilers out there that would give us that fine of detail on iOS.

 

[Exophase]

Why do you doubt it? But I'm sure you can understand my confusion, as to my knowledge there are no profilers out there that would give us that fine of detail on iOS.

I doubt it simply because Qualcomm is the only other one doing it, and Apple couldn't have been doing it with their Cortex-A9 chips. This change means a move to more complex PMICs, unless the regulators are also on the SoC.

 

[Lazy8s]

My guess for the performance of the next iPhone's graphics comes from the amount of increased density in shader compute I'm figuring between SGX54x cores and a four cluster Rogue core, layered with even more assumptions as to how Apple might configure their phone solution and how that might perform in practice. So, it's a shot in the dark essentially, and I could see an equally convincing scenario that gives it about a one-third performance advantage (around what french toast anticipated).

I can even see the logic behind keeping a larger difference between iPhone and iPad (considering their relative resolutions), and equipping the iPhone with a SoC that reflects the small internal volume that a very thin 4" device can afford. Under that logic, it's not unreasonable to see that the next iPhone's graphics might not even reach the level of performance of a "superphone" like the Galaxy S4 LTE Advanced.

Still, I'm satisfied with my original guess.

What I'm wondering is, with the race to the bottom in current tablet design, does Apple even need to refresh the full-size iPad yet considering no one's really even challenging them there? A new iPad mini, though, is definitely needed, but that doesn't necessarily need quite as hefty a SoC as the larger iPad.

 

[ltcommander.data]

What I'm wondering is, with the race to the bottom in current tablet design, does Apple even need to refresh the full-size iPad yet considering no one's really even challenging them there? A new iPad mini, though, is definitely needed, but that doesn't necessarily need quite as hefty a SoC as the larger iPad.

I'd hope Apple has some internal plan for where they want tablets to go independent of what everyone else is doing. With resolution increasing 4 times over the iPad 2 and GPU power increasing about 4 times in the iPad 4, things have been pretty stagnant on a GFLOPs/pixel basis for the past 2.5 years so some movement on the full size iPad is overdo. A retina iPad Mini would likely be using a 28 nm shrunken A6X.

One thing I have been wondering about is whether there is anything similar to Intel's InstantAccess or AMD's HUMA in Apple's current SoC or any other mobile SoC? If not, does it require major changes to the CPU architecture which Apple controls and/or the GPU architecture (presumably Rogue) which they probably have less say on? Spending transistors on extra logic and buses to tightly integrate the CPU and GPU would be more interesting than just adding 2 more CPUs.

And despite the news that Imagination and TSMC are working together on optimized layouts, I wonder if Apple will just go ahead and hand-layout the A7/A7X GPUs? Given the GPU takes up more die area on the SoC than the CPU and presumably has a larger peak power load as well, if Apple is going to continue to hand-optimize the CPU layout, doing the GPU as well should be as fruitful if not more so.

 

[anexanhume]

I doubt it simply because Qualcomm is the only other one doing it, and Apple couldn't have been doing it with their Cortex-A9 chips. This change means a move to more complex PMICs, unless the regulators are also on the SoC.

We know for sure the current dialog chip doesn't do it? Their "custom" chips to easily correlate with existing IP?

 

[Lazy8s]

When Imagination made a press release recently recognizing MediaTek for their accomplishment with heterogeneous multi-processing in an upcoming SoC, I think it was primarily highlighting the fact that it had full cluster and core migration flexibility within its big.LITTLE implementation, so I'm guessing HMP between GPU and CPU with shared memory addressing and more for mobile isn't quite on pace with desktop development yet. But, as you mentioned, Apple with its custom ARM implementation could be one of the first to do it in this space.

I do agree that Apple's introduction of new product updates for the iPad, iPod touch, etc. aren't in reaction to their competitors, and I've been looking forward to this new generation in particular of large-sized iPad for precisely the reason of a SoC that would finally advance the GFLOPS/pixel ratio beyond the iPad 2 which I'm still using. Still, I'm wondering if Apple is considering becoming less aggressive on the full-sized iPad roadmap like they seemingly did with the iPod touch a while back as they evaluate the market trends.

As for my earlier prediction about the A7's graphics performance, I understand Nebuchadnezzar's point now about Adreno 330 already pushing the confines of a smartphone form-factor SoC after struggling to find examples of the Galaxy S4 LTE-A not actually getting throttled when run through the standard set of benchmarks. I knew the Adreno architecture wasn't leaving itself a lot of headroom this generation after the first 320 devices were having some throttling issues (which were later smoothed over), but it's clear that 330 is sucking down a lot more wattage under load (4+?!) than what I've seen of any iPhone and even iPad GPU thus far.

So, now that I understand that Adreno 330's power profile far exceeds what Apple would feel is rational for a 4 inch phone, I'd be plenty satisfied to see A7's graphics performance just approach that of the Galaxy S4 LTE-A. I'm definitely not predicting any 50% benchmark beating performance anymore...

 

[Lazy8s]

I'm guessing Apple will unveil the typical 4" iPhones today and then update the line several months later (not waiting anywhere close to as long as the typical annual refresh) to introduce the larger sized variant(s).

 

[Exophase]

We know for sure the current dialog chip doesn't do it? Their "custom" chips to easily correlate with existing IP?

I'm pretty confident that whatever PMIC they used for their Cortex-A9 based SoCs didn't have more big current (probably switching regulated) voltage rails than it actually needed.

No I don't know for sure their current one doesn't do it, I'm just saying they would have switched to a more complex one than before if so.

 

[Pressure]

Boom, 64-bit!

2x general-purpose registers
2x floating-point registers
Over 1 billion transistors
102mm2 die size

OpenGL ES 3.0 support

 

[Laurent06]

That's great news

 

[Deleted member 13524]

Boom, 64-bit!

Unless it bundles 4GB of RAM, that sounds a bit useless.

107mm^2 at 28nm. That doesn't leave a lot of space for a monster GPU.
Smaller than A5X, right?

 

[Lazy8s]

Too much focus on the CPU means not enough die devoted to the Rogue GPU. Disappointing to me.

 

[Deleted member 13524]

Too much focus on the CPU means not enough die devoted to the Rogue GPU. Disappointing to me.

They're claiming 5x iphone 5's GPU performance, though.

EDIT: it was a typo on the Verge. 2X better GPU performance.