Talk:Apple A6

Cores included

The information that the A6 is a dual core ARM Cortex-A15 CPU is complete speculation 90.244.169.149 (talk) 21:21, 12 September 2012 (UTC)

It might, but it's credibly sourced. AnandTech is a very respected source in this matter. -- Henriok (talk) 21:26, 12 September 2012 (UTC)

Very dubious claim, even for anandtech (which has been getting more dubious over time, unfortunately). Citing sources and providing evidence seems to be something the digital publishing space has been shying away from. What a shame. Given everything I've seen so far, the power draw of the cortex-A15 makes it rather prohibitive in mobile SoC design, even down to 28nm. I'd bet on quad-A9s and expect to see a correction to the anandtech article imminently. Tinkuman (talk) 13:16, 13 September 2012 (UTC)

From what I have seen, an A15 clocked at moderately higher than 1 GHz would probably satisfy Apple's performance claims, and the power draw is not an issue at such low frequencies. The other A15 designs seems to aim much higher, and those are probably for tablets, netbooks and such appliances. -- Henriok (talk) 13:58, 13 September 2012 (UTC)

You forget, power is both dynamic and static. Even at low frequencies, the die size increase of the A15 (reported to be more than double the A9) would have significant leakage. This doesn't jive with the standby time at all. On the dynamic side, yes, they could have controlled the frequency (or supported multiple scaled frequencies), but I don't believe the architecture of the A15 was designed for mobile applications (instead, low-power server market to take on Intel). It's performance relative to size vs the A9 is probably less. Tinkuman (talk) 16:04, 13 September 2012 (UTC)

How about a big.LITTLE? (A15 A7 or 2 x A15 2 x A7) gnirre (talk) 09:25, 16 September 2012 (UTC)

Would be the first Cortex-A15 to market, wouldn't it? Apple has used Samsung technology previously and they're just releasing quad-core Cortex-A9s. Is the LTE integrated? If so, Apple could have been forced to use Snapdragon technology. Not sure...just discussing. Knurdtech (talk) 22:25, 12 September 2012 (UTC)

It is first to market. I did a search of all Cortex-A15 based parts I could find and no other has found its way into actual products for sale yet. Several parts from TI and Samsung has been demonstrated live at expos though. This is original research and must be sourced if it's to be included in the article. The relationship between Apple and the department doing chip manufacturing at Samsung is still rosy as far as anyone knows, and their equivalent processor generation is the Exynos 5. The LTE chips is not integrated in the A6, it's a separate chip from Qualcomm. -- Henriok (talk) 22:39, 12 September 2012 (UTC)

This is not confirmed to be an A15. Therefore A15 is a rumour and should not be on here. It could be 1.6GHz A9s, you know... 86.146.71.222 (talk) 19:58, 14 September 2012 (UTC) Also on the article it says it may be a quadcore... Apple has confirmed that is not true. 86.146.71.222 (talk) 18:29, 15 September 2012 (UTC)

It is a custom design and uses a SGX543MP3. 3|9|3|0|K (talk) 12:20, 23 September 2012 (UTC)

Designed by Apple? Or Samsung?

The article states that Apple designed this chip, but in fact, Samsung are the ones making the SoC, just like LG makes the screen. Apple hardly makes anything that the iPhone is built out of. Jørgen88 (talk) 23:19, 12 September 2012 (UTC)

Samsung is manufacturing the chip, but it is designed by Apple. Tweisbach (talk) 01:34, 13 September 2012 (UTC)

Do you have a source for that? Most sources regarding this and the previous chips says their designed and manufactured by Samsung. Apple just buys them to put them in their phones. Jørgen88 (talk) 10:56, 13 September 2012 (UTC)

The only source you really need is that Apple says they are designing their processors. You can watch either movie on this page to get that from Phil Schiller and Bob Mansfield, both Senior Vice Presidents of Apple, people why actually can be prosecuted for lying about stuff like this. And they have ample of resources to design stuff like that, they even bought the company that tweaked Samsung's process, Intrinsity. Besides that, no Samsung source that I know of have ever stated otherwise. So, between the two parties, they are in agreement on how these processors are designed, as should we. -- Henriok (talk) 11:41, 13 September 2012 (UTC)

That is absurd and untrue. You need some serious third party sources. Mansfield and Schiller cannot be "prosecuted" for making their statements even if they are grossly untrue and 99% of the design was decided by Samsung.71.252.100.185 (talk) 16:49, 17 September 2012 (UTC)

Of course they can. Under the SOX act high seated officials can be prosecuted if they lie, i.e. commit fraud. Lying about key production details and such as this would certainly be fraud, and if the DOJ doesn't care, I think several stock owners will and have a great case in a civil court. –– Henriok (talk) 13:48, 24 September 2012 (UTC)

The manufacturing part is trickier. I haven't seen actual statements from either Apple nor Samsung in this regard, but it's "widely known" and not disputed by any first party (like either Apple or Samsung or the actual manufacturer). Optical analysis of the processor die done by experts found extreme similarities in manufacturing between the A4, A5, A5X and similar chips from Samsung[1][2][3] giving a very strong indication that they are the manufacturer. They also found that Apple's chips are not anything that Samsung is selling. The markings on the package is also consistent with how Samsung is marking their chips. BUT, if anyone can actually provide sources that are disputing that Apple designed these and Samsung manufactured them, that'd be great. So we have something real to discuss. There are many thinking, and stating something, but there are far fewer that actually have some actual base for what they think, so is suddenly becomes more of a fantasy than a fact. -- Henriok (talk) 11:41, 13 September 2012 (UTC)

ARM v7s

It appears that the instruction set the A6 supports is a variant called ARM v7s. http://www.cocoanetics.com/2012/09/iphone-5/ Tweisbach (talk) 01:53, 13 September 2012 (UTC)

armv7 ==>Cortex A8, Cortex A9

armv7f ==>Cortex A9MP

armv7s ==> Cortex A15.

So this may mean that the A6 really does use Cortex A15s. — Preceding unsigned comment added by 68.38.184.189 (talk) 02:23, 14 September 2012 (UTC)

Do you have a source for ”armv7s ==> Cortex A15” gnirre (talk) 08:11, 14 September 2012 (UTC)

ARM itself say "ARMv7-A Cortex == Cortex-A15 MPCore". --Cvf-ps (talk) 22:13, 15 September 2012 (UTC)

Cortex A5, -A7, -A8, -A9, -A15 are all "Armv7-A":s http://en.wikipedia.org/wiki/List_of_ARM_microprocessor_cores but it makes sense coders would need to be more specific. gnirre (talk) 09:21, 16 September 2012 (UTC)

Yes, of course; for andandtech A6 ≠ ARM Cortex A15 MPCore. --Cvf-ps (talk) 11:57, 16 September 2012 (UTC)

Geekbench score

Should we add the things revealed from the GeekBench score? 3930K (talk) 15:21, 17 September 2012 (UTC)

It's still not confirmed, see Anandtech comment about this. --Cvf-ps (talk) 10:39, 18 September 2012 (UTC)

Performance claims

This is original research (using math) and info that can be found on other Wikipedia pages. I'd like for it to be included in the article. -- Henriok (talk) 13:23, 24 September 2012 (UTC)

The general performance gains of the A6 compared to previous processors are a combination of redesigned cores, higher frequency, number of cores, higher memory bandwidth, reduced manufacturing process and refined logic layout. Graphics performance of the PowerVR SGX family scales nearly linearly with the number of cores and frequency so the tri-core A6 running at 266 GHz (3 × 266 MHz = 800 MHz) is equal to a quad-core A5X at 200 GHz (4 ×* 200 = 800 MHz), or twice that of the dual-core A5 at 200 MHz (2 ×* 200 = 400 MHz). The CPU performance is doubled due to a combination of higher frequency (800 MHz → 1200 MHz), increased memory bandwidth (400 MHz 32-bit DDR2 = 6.4 GB/s → 533 MHz 32-bit DDR2 = 8.5 GB/s) and a revised micro architecture for the cores. ARM have stated that you can expect a 40% increase going from Cortex-A9 micro architecture to Cortex-A15, and Apple's A6 core claims to be similar to the A15. Performance regarding power consumption is increased due to a more efficient manufacturing process (45 nm → 32 nm) and advanced layout optimization of the logic, something that the company Intrinsity excels at. Intrinsity was bought by Apple in 2010. All in all, theoretical performance gains can be estimated up to three times so a claimed doubling in actual performance is reasonable.

Clock rate/frequency

GeekBench is not an official source and isn't a very reliable source for CPU frequencies by itself. Previously it approximated clock speeds at 1.2GHz, current version approximates it at 1.3GHz, next version might have different approximation algorithms. Unless some of you actually have an official clock numbers for base clock speeds and maximum clock speeds for A6 OR you have another 3rd-party source for it not dependent on GeekBench you should at least use word "approximately" next to its stated frequency... Rndomuser (talk) 04:00, 30 September 2012 (UTC)

Successor?

Why is the A6X listed as the "successor" to this chip, when the A5 is the "predecessor"? There are really two ways to categorize this:

1) The A5 is the predecessor to the A6 because these are successive SoCs used in the iPhone. (In this case, the successor will be the rumored-but-not-yet-revealed A7, not the A6X, which is iPad-only.)

2) The A6X is the successor to the A6 because it is the newest SoC design announced by Apple after the A6. (By this logic, the predecessor was the A5X, not the A5.)

I'd rather have consensus before fixing this, but the successor should be unannounced, not the A6X. Shelbystripes (talk) 18:03, 14 April 2013 (UTC)

Actually,I added that. I'll remove it now, silly me. Numbermaniac - T- C 02:51, 15 April 2013 (UTC)

Thanks! That was easy. Shelbystripes (talk) —Preceding undated comment added 03:51, 15 April 2013 (UTC)

Cool, but did you sign that with 4 tildes? (~~~~?) if you did, there's a problem with SineBot. Numbermaniac - T- C 09:26, 15 April 2013 (UTC).

No, I probably signed it with 3. I've made that typo before, when you only put three tildes it puts your name but not the time/date. Sorry about that. Shelbystripes (talk) 18:35, 15 April 2013 (UTC)

Nothing to be sorry about. I didn't realise that there was no date. And I didn't realise that SineBot was only adding the date. It's fine. Numbermaniac - T- C 00:29, 16 April 2013 (UTC)

Instruction set

It should be mentioned in the article that the CPU is a 32-bit CPU, or that it has a 32-bit instruction set. The problem is that anyone new to Apple A6 would not know about it, and the CPU's successor Apple A7 is 64-bit. -Mardus (talk) 04:06, 28 November 2013 (UTC)