Next-Gen iPhone & iPhone Nano Speculation

[Pressure]

Samsung S III (Quad-core @ 1.4Ghz) compared to iPhone 5 (dual-core @ 1.02Ghz).

Great single core performance and look at that memory performance.

So the A6 is ~6,4% slower on the final score with a 400Mhz deficit. Not bad.

 

[ltcommander.data]

Samsung S III (Quad-core @ 1.4Ghz) compared to iPhone 5 (dual-core @ 1.02Ghz).

Great single core performance and look at that memory performance.

So the A6 is ~6,4% slower on the final score with a 400Mhz deficit. Not bad.

With 2 less cores too. Does anyone know if Geekbench scales well with core count which may put quad cores at a disadvantage?

 

[Exophase]

These calculations are all assuming that Geekbench's CPU speed is representative of what it's actually running at. Does anyone remember when the iPad 2 first came out and there were Geekbench scores showing the CPU at varying clock speeds like 900MHz instead of 1GHz?

Who knows how it's reading clock speed. I'm not aware of iOS providing a value on this. DVFS behind the scenes can skew any calculation, and even a timing loop at a stable clock can be wrong if the assumptions about the uarch are wrong (although a reasonable loop would be based on a bunch of dependent ALU instructions, and it'd be odd if they had anything but a single cycle latency)

 

[Nebuchadnezzar]

With 2 less cores too. Does anyone know if Geekbench scales well with core count which may put quad cores at a disadvantage?

I'm getting 1854, 1574, 1292 respectively for 4, 3 and 2 cores. So the final score doesn't seem to scale that well with cores, which is obvious consindering there's single-core performance tests.

Here's the Exynos 4412@ 1Ghz with only 2 cores for a clock-for-clock comparison;

http://browser.primatelabs.com/geekbench2/compare/1032330/1030202

And here's @ stock 1.4Ghz with 4 cores, frequency locked:

http://browser.primatelabs.com/geekbench2/compare/1032351/1030202

Geekbench seems to take the frequency from max possible possible CPUFreq entry, I'm on my own kernel which has 1.8GHz as max but the real frequency is how I stated.

 

[Kaotik]

With 2 less cores too. Does anyone know if Geekbench scales well with core count which may put quad cores at a disadvantage?

When the major wins for A6 come from memory performance, saying "with 2 less cores too" is just wrong, as count of cores has nothing to do with what brings A6's score so much up

 

[RudeCurve]

Wow nice work guys, the A6 looks to be very competitive with what's out there.

 

[french toast]

Nubuchanezza: ok fair enough you have provided superior evidence and I am now at your way of thinking, nice explaination by the way and you did it with out any arrogance or "i told you so" ...in that case I'm disappointed at Samsung for being so misleading.

Yes that A6 chip looks pretty special..can't wait to see the efficiency of the thing, does this mean this is the first time arm uarch has caught up with x86 memory wise?..or was x86 never that much in front anyhow?

 

[Rys]

If 4412 clocks and powers all cores in lock step then I don't think it has to be that way. The CPU layout in hardware is clearly setup to do something else.

 

[Mariner]

A software issue, perhaps? It's not unheard of for the power saving aspects of mobile CPUs to be pretty much ignored by the phone makers in early software releases to ensure the devices are released on schedule. I can remember various occasions where new software releases have improved battery life a fair amount.

That said, I can't really remember seeing reports of users having too many problems with the SGS III as regards battery life or is it just a case of unreasonable expectations with some people? With a screen that big, something has got to give and expecting battery life equivalent to some of the smaller devices is perhaps a bit too much to ask.

 

[french toast]

I'm pretty sure Samsung has claimed (white paper?) That exynos 4412 can independently run each core at its own frequency, but I do believe what nubuchenezza says that it isn't currently doing that right now.

Perhaps what you say is true and this is an android ics issue and will be fixed with jellybean + future firmware updates.

Certainly I don't have a afterlife issue with those phone...it's awesome considering the super nova like screen....any improvement to that would be awesome.

Back know topic, having looked at the geekbench breakdown of the a6 it's now clear that the apple SOC is pound for pound the best one out there, despite being clocked as low as 1ghz..it is only just behind krait cpu score despite krait being clocked 1.5ghz!..

Memory controller is out of this world for arm, gpu is obviously going to smash current smartphone king Mali 400 mp4..
I'm thought impressed with the new iPhone with the exception that it has not moved on with new design ideas...the software is painfully out of date visually and far too simplistic and boring compared to rival offerings...although I really like new siri and the apple maps look great.

 

[Laurent06]

As far as I know, cores in an A9 cluster all have the same clock. All you can do, is power off a core independently.

 

[OlegSH]

http://www.phoronix.com/scan.php?page=article&item=odroid_panda_linaro&num=2
That's how should perform STREAM on Exynos 4412, wonder why it's almost 2 times slower in geekbench 2

 

[Exophase]

Is Geekbench (or any other third party benchmark) on Android compiled with the most recent NDK? Until recently they were using a really old version of GCC (4.4) and that can have a pretty serious performance deficit given all of the changes in ARM performance since then. Even now they're only up to 4.6.

It'd be great if you could somehow run the same binary on both for this..

 

[wco81]

iOS is still reportedly the smoothest UX out there. People say ICS on Nexus 7 stutters and requires reboots to clear up.

Maybe iOS 6 will slow things down again. I'm wondering if I should update my iPad 3. The new Maps sound like a mixed bag: unlikely to have the rich data set of Google Maps but Flyover is cool Also would get Siri in the software update.

But may have to weigh that against possible slowdowns on iOS 6.

 

[Rys]

No responsiveness or slowdown issues in the main UI for me at all on iPad gen3 for iOS6 and it's as great as ever.

 

[wco81]

That's great to hear, thanks.

 

[Mariner]

iOS is still reportedly the smoothest UX out there. People say ICS on Nexus 7 stutters and requires reboots to clear up.

Jelly bean is the release which aims to improve the "smoothness" of Android.

I've not been paying attention, to tell the truth, but I assumed Nexus 7 already had JB available!?!

The fact that quite a few phones have some pretty good JB ROMs which have been hacked together by enthusiasts in their spare time (not to mention cheap Chinese tablets) makes me extremely surprised to hear that the Google tablet is still stuck with ICS!

 

[BRiT]

Yes, my Nexus 7 device came with JB installed.

 

[wco81]

Yeah that's what I meant JB which came preinstalled on Nexus 7. That's according to people who have both an N7 and iPad3 on a discussion thread about N7.

 

[Mariner]

Ah, right.

As I mentioned, I've not really been paying attention.

I'm still waiting for a JB ROM on my cheapo (yet pretty good) Android Tablet which I bought from China. No great hurry, however, as the ICS ROM doesn't have any issues and I rarely notice any lag.