Next-Gen iPhone & iPhone Nano Speculation

I'm surprised we are seeing benchmarks for the A6X chipset considering it's not being released until Friday? Where are these coming from?

 

If anything that iPhone 5 score looks odd heh.

 

Why? It looks like many other iPhone 5 results, be they qualified as ARMv6 or ARMv7

 

Not at all, for starters it is only shown as 1.20 Ghz and then it gets nearly 100 points more than all the other iPhone 5 phones.

 

Geekbench is notoriously bad at reporting frequency. And a large number of recent iPhone 5 results is in the 1650-1710 range (I used "iphone5" to search the DB).

And indeed picking from an "iphone 5" result, I get this: http://browser.primatelabs.com/geekbench2/compare/1213259/1219370. And it definitely doesn't look odd... Typical Geekbench

 

I guess if the CPU clock is 1.4GHz and they kept a SGX543MP4 we expect the clock speed to be 467MHz using a divider of 3? That would put it just under 2x performance improvement.

The other big Apple news is that Forstall is gone. With Ive in charge of interface design the expectation is that skeumorphism is out. I wonder if we'll now see more aggressive changes to UI and OS functionality? In particular, how much of the opposition to expanded multitasking/multi-application usage in iOS is innate to Apple and shared by all employees and how much was Steve Jobs' point of view shared by Forstall? With increased competition from Android and Windows/Windows Phone in the tablet/smartphone space, Apple could certain incorporate a bit more "power-user" functionality to reduce defection, especially if Ive can figure out a way to present it in an intuitive way to the user without copying existing solutions.

And it looks like Mansfield is out of retirement and leading the wireless and semiconductor teams. I'm guessing by wireless they mean existing antennae design teams rather than a move toward wireless chipset design which seems unnecessary.

 

Apple continues to establish that marginally higher performance profile for its full-size tablet versus its phone with the slight CPU bump. I expect the GPU clock rate of its 543MP4 to approach 500 MHz (like 467 MHz as ltcommander.data speculates or higher), keeping it atop the tablet SoC space for performance.

 

What about the "rule of 1/4th" that has existed since the iPhone 3GS?

Is it out of bounds to expect the A6X with dual-core Swift clocked at 1.40 Ghz and the SGX543MP4 clocked at 350 Mhz?

That or Geekbench is really bad at detecting the clock speed, like the first Geekbench result from the iPhone 5 which showed it as a 1.00 Ghz dual-core.

 

I think for iPhone, they will want to pack the smallest battery possible, keep it thin as possible

That probably means they won't enable "true" multitasking. It's been 2 years or so since they came out with the scheme they have and push notifications to substitute for backgrounded processes.

Sales haven't exactly suffered in that time. I don't know how much battery life and slim form factor contributes to iPhone sales but these are signature differentiators vs. Android.

Now, if they went to bigger displays and enabled multi window UIs of some kind, maybe "real" multitasking becomes more likely. However, they're loathe to give up their apps. compatibility advantage by supporting something like 720p or 1080p.

 

Yeah. The 1/4 rule has been consistent all this time. And 1.4GHz does seem like a very modest clock speed bump for the CPU given up to now the iPad has always been clocked 25% higher than the corresponding iPhone version of the SoC. Such a small clock speed bump would presumably make the A6X comparatively power efficient, yet Apple never claim battery life improvements which is curious.

It's definitely more a question for tablets. Fast-app switching with enough RAM to keep several recent programs resident should be sufficient for a smartphone with a modest screen.

I wonder what Forstall will do once his "consultant" period is up. He can't be happy with Apple, but the dislike for Microsoft, Google, and Samsung is probably too deeply ingrained for him to go that route. If he stays in the mobile OS world maybe leading HP's 2014 smartphone comeback could be an option or maybe shaking things up at RIM?

 

Thing is, he probably has enough money so that won't be a motivation.

But if he wants to punish Apple, then he could presumably try what Rubenstein and others tried, to come up with an alternate platform.

He would have to go to a big player. Palm/WebOS showed that if you try to ramp up, it won't work.

 

http://www.engadget.com/2012/10/30/apple-ipad-review/

Well Engadget got 42 fps in Egypt HD presumably onscreen. The iPad 3 typically gets around 22 fps onscreen so the 2x GPU speed-up seems to check out. How it was accomplished is still open of course.

 

What i find fascinating is that the new Apple core is roughly on par with Intels Core2Duo in IPC.
If you look at this chart you see that the Core2Duos that have clocks between 1.4-1.8 GHz have scores from 2025 to 2061, on average.

Now, comparisons within an architecture is difficult enough, between architectures the grains of salt need to be boulder size, but nevertheless - that's pretty damn impressive, because the Core2Duo had very good IPC. The subsequent Nehalem and descendants primarily improved main memory latency and bandwidth both, but as long as you kept on-chip, the Core2 family of processors did really well even vs its successors.

Also, on Intels 45nm, the most frugal of those chips had a TDP of 10W for the CPU only. Once you add the northbridge chip and graphics to those 10W, the advantage of the ARM platform is simply staggering.

 

The one thing this doesn't account for is that those scores came from all sorts of different Geekbench versions. In most cases they're probably much older than the current Apple ones. That means one could be compiled with compilers that are as much as 8 years older.

Geekbench is really not a good benchmark, it's so poorly regulated and reported :/

 

http://www.chipworks.com/blog/recen...-the-apple-ipad-mini-samsung-32-nm-confirmed/

In case there was any doubt, the iPad Mini is using a 32nm Samsung manufactured A5. I guess it's too early to start 28nm Samsung trials or a switch to 28nm TSMC.

http://www.chipworks.com/blog/recentteardowns/2012/11/01/inside-the-apple-ipad-4-a6x-to-be-revealed/

A6X analysis is underway and should be much more interesting.

 

I guess until we see benchmark results no lucky guess will work for possible frequencies on the A5 in the mini.

 

Entropy,

Congratulations for being right on spot:

http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Apple+iPad+4&testgroup=gl

SGX554MP4@ probably =/>260MHz

http://www.glbenchmark.com/phonedetails.jsp?D=Apple+iPad+4&benchmark=glpro25

5492 frames in 2.5 offscreen; triangle setup must be stronger on 554 I assume, otherwise I'm having a very hard time decyphering the triangle throughput results compared to iPad3.

http://www.glbenchmark.com/compare....tified_only=1&D1=Apple iPad 4&D2=Apple iPad 3

 

It seems the one inside ipad4 is SGX544MP4, near 3 times faster

 

The first link above actually verifies it.

 

Hmmm. I initially suggested it as a possibility a few posts earlier in post 1567
http://beyond3d.com/showpost.php?p=1674884&postcount=1567

One assumes relative gpu die size has increased again, given 554 includes dx compliance IP.

Assuming geek bench is reporting the CPU speed correctly @1400mhz, then it's not easy to get a.260mhz clock from that. A divide by 5 would get you 280mhz, which might explain the exact 13% increase in fill rate (280mhz is 13% increase on 250mhz)