News - iPhone 5's A6 processor detailed and benchmarked

[HEXUS]

Geekbench scores beat all mobile and tablet opposition.

Read more.

[Badbonji]

And then there was Jelly Bean: http://browser.primatelabs.com/geekbench2/1026099

[Biscuit]

Ah yes, good old android improving performance with new operating system revisions. Still impressive results from a dual core 1Ghz CPU though.

[edvinasm]

Ah yes, good old android improving performance with new operating system revisions. Still impressive results from a dual core 1Ghz CPU though.

Impressive because it is custom designed GPU built in chip. Only dual cores means nothing these days - it's all about certain instruction sets and acceleration & optimisation of these. That's what ARM is pushed to do since the power usage must remain low - so.. more MHz is not an option. More operations per cycle however is.

[edvinasm]

Oh and regardless what the Apple does I will always opt for Samsung or Nokia handset.

[keithwalton]

The test seem's to be quite heavily weighted on memory bandwidth with the raw cpu performance actually quite a bit lower.
Increasing memory bandwidth only really matters when it becomes a bottleneck to performance.

I'm actually really surprised it's an A9 and not an A15 step it up apple!

Also got to wonder why the One X is so slow compared to the nexus 7 / transformer pads, since they all use tegra 3's.

At least with android os improvements come from code improvements instead of clockspeed increases ... who would do a thing like that

[watercooled]

It's still an A9/ARMv7 CPU so shares the same instruction sets as other CPUs on the market, it's the architectural improvements at work here. Not much is currently known about it but it's probably similar to Qualcomm's Krait. Core count is also important, of course; scaling isn't perfectly linear but you theoretically get about double the performance for dual cores given the same cores, and potentially power saving advantages depending on how they're used. For example, two lower clocked cores may give better power efficiency than one much higher clocked core, as power draw/clock also isn't linear.

The benchmark results are to be taken with a large grain of salt ATM, and if you do a quick search of the Geekbench results, plenty of stock Android devices outperform what are likely cherry-picked/peak results, if they're reliable at all, anyway.

Edit: @keithwalton: The One X may have lower scores as it uses DDR2 memory vs DDR3 used by the other devices.

[jnutt]

And then there was Jelly Bean: http://browser.primatelabs.com/geekbench2/1026099

Just shows you, never believe and Apple Marketing hype

[crossy]

Also got to wonder why the One X is so slow compared to the nexus 7 / transformer pads, since they all use tegra 3's.

Never mind those - look at the figure for the HTC One S, a "lesser" (dual core) device that - according to these figures - is steets ahead. I was tempted to point the finger at the Sense UI for the reason that the HOX is so damned dismal, but the HOS' marks kind of put the lie to that.

Anyone know whether this Geekbench is any good - it's not a benchmark that I've come across before? Looking at the "pretty picture" seems to suggest that it's poorly suited to quad core devices.

[watercooled]

It's nothing special really, you can see what it tests here: http://www.primatelabs.com/geekbench...enchmarks.html

[keithwalton]

The one s uses a dual a15 core chip (8260a) which is known to perform much better than the a9. If you look at the bigger picture the 8260 (a9 design) running at the same speed as the one s scores around 900.

Interesting point about the one x having lower memory bandwidth, I wonder how much that actually impacts any real world performance. I highly doubt it takes a 30% hit! Would show up the accuracy of the benchmark ..

[watercooled]

The 8260A is a Krait core, which fits somewhere between A9 and A15, but like I said the A6 seems to be somewhat similar based on what we currently know.

AFAIK the Samsung Exynos 5 will probably be the first A15 SoC to appear in devices.

[drowlin]

Why is it generation 6?

[crossy]

Why is it generation 6?

A4 was the first "Apple" processor - gen seems to say that it got this name because it was used in the iPhone4 - so Stevo thought it'd be good to have a link between processor and chip.

The iPhone4S uses an improved A4, the A5. In which case it's pretty logical that the improved 4S (the 5) would use an "A6" processor (since it's an improved A5).

I'm sure watercooled etc have a more technically accurate (and comprehensive) answer.

[keithwalton]

The 8260A is a krait core which is derived from the A15, the 8260 is a scorpion core which is derived from the A9.

The more I think about this apple surely started off with an A15 design and not an A9 to make this chip as they see to be making around the same gains arm did with the tweaks they made ... otherwise it's one heck of a coincidence!

[watercooled]

AFAIK it's just a name, but to confuse matters they chose a very similar naming strategy to what ARM use for their cores, so people often confuse the two.

There's also the A5X which contains a relatively massive GPU, used in the iPad 3.

To sum it up,

The Apple A4 uses a Cortex A8 core
The Apple A5 uses Cortex A9 cores
The Apple A5X also uses A9 cores but with a more powerful GPU than the A5
The Apple A6 uses something between A9 and A15 cores, like QC Krait, and claims to have similar graphics performance as the A5X.

Edit: I need to refresh before posting.
Yeah I meant 8260A in my post above, just a typo. Krait and A6 are sort of mid-way between A9 and A15; they are similar to A15 but missing some features/different microarchitecture. Scorpion has similarities to A8 and A9.