• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

iPhone 6S' performance is monstrous

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I think, given the terrible level of optimization on x86, it's best to just take Geekbench as completely separate benchmarks on Android, iOS and x86 whose score values have no meaning between platforms.

I feel like everything is hearsay. Do we have any good sources? It wouldn't be at all surprising if Apple were looking to take on Intel and make their own CPUs going forward. ARM has a business model Intel simply can't compete with, you get the compute performance for a fraction of the cost. Unless they keep control of their bleeding-edge foundries (which they will), Apple would have no reason not to switch.
 
I feel like everything is hearsay. Do we have any good sources? It wouldn't be at all surprising if Apple were looking to take on Intel and make their own CPUs going forward. ARM has a business model Intel simply can't compete with, you get the compute performance for a fraction of the cost. Unless they keep control of their bleeding-edge foundries (which they will), Apple would have no reason not to switch.

All signs are pointing Apple dumping Intel sooner rather than later:
-Apple SoCs continues to deliver massive IPC and GPU gains over the last.

-Existence of iPad Pro.

-iDevices alone already beginning to outsell consumer PCs in volume with SoCs built on processes more or less at Intel's level, and Intel is struggling to maintain their process advantage lead.

-Aggressive pushing of software optimizations from OS to APIs like Metal to mitigate the raw hardware performance loss by dumping Intel through superior software engineering. Apple can easily pull this off with their vertical integration advantage.
 
-Aggressive pushing of software optimizations from OS to APIs like Metal to mitigate the raw hardware performance loss by dumping Intel through superior software engineering. Apple can easily pull this off with their vertical integration advantage.

I also wonder if Jim Keller leaving AMD was to go work with apple (or samsung?) on new ARM designs.
 
http://www.phonearena.com/news/Whop...xy-S7s-chipset-compared-with-Apple-A9_id73980

Hmm feels like we're close to a point where increasing performance won't have much real-world difference in the future. Scores from all OEMs are getting pretty insane - though hopefully power consumption is still reasonable.

Exynos 8890 - scoring 2304 single core and 8038 multi-core.

Yeah, I really wonder for how many people those great new benchmark number are actually relevant.
My old Nexus 5 still feels plenty fast for practically anything I am doing with it, for my needs, "good enough" speed was reached two years ago.

Now, long battery life in a smallish (<=5") form factor, THAT would pique my interest...
 
Yeah, I really wonder for how many people those great new benchmark number are actually relevant.
My old Nexus 5 still feels plenty fast for practically anything I am doing with it, for my needs, "good enough" speed was reached two years ago.

Now, long battery life in a smallish (<=5") form factor, THAT would pique my interest...

Agreed on the Nexus 5 still feeling fast today.

Sony Xperia Z5 Compact has three day battery according to review by The Guardian; http://www.theguardian.com/technolo...ria-z5-compact-review-best-smaller-smartphone
 
I think, given the terrible level of optimization on x86, it's best to just take Geekbench as completely separate benchmarks on Android, iOS and x86 whose score values have no meaning between platforms.

What is this terrible level of optimization you refer to? It's all C code using the same compiler. Do you mean because they don't use ICC on x86?
 
Yeah, I really wonder for how many people those great new benchmark number are actually relevant.
My old Nexus 5 still feels plenty fast for practically anything I am doing with it, for my needs, "good enough" speed was reached two years ago.

Yah, we're getting desktop class number on mobile now. Won't be long before my current i7 2600k build is eclipsed at this rate.
 
Agree on the "good enough performance" since the A7 and the S800, but improvement of this magnitude will be noticeable nonetheless. It is almost impossible not to notice nearly 50% higher performance, even with naked eyes. That is what happened with S805 -> Exynos 7420, and is what is happening with A8 -> A9.

Exynos 8890 - scoring 2304 single core and 8038 multi-core.

I would take any of those numbers with a large grain of salt. It is way too early for a legitimate leak and if those numbers really came out of Samsung I would suspect a controlled leak in order to mislead. Samsung must have known A9's performance long ago and have planned things accordingly, and with their new manufacturing node coming up (14 LPP instead of 14 LPE) I expect great things next year.

Back to iPhone 6s, the architectural changes of A9 over A8 should be of a very high interest. On the outside, two obvious changes are;

1. A9's IPC increase over A8 is much greater than that of A8 over A7.
2. A9 clocks much higher than A7 or A8. That indicates architectural changes to enable higher frequencies, and if that is the case the next question is how high it can go.
 
Yah, we're getting desktop class number on mobile now. Won't be long before my current i7 2600k build is eclipsed at this rate.

Apple is most likely still stuck with 14nm without a node shrink for the first time on the A10 and a smaller % ST performance increase like with the A7 -> A8. However, the real trump card of Apple is their software crew that is at worst still world-class which puts them leagues beyond purely-hardware players like Intel and Samsung and mostly-software people like MS and Google in terms of utilizing raw specs to the fullest.

Having iPhones and their chips outselling PCs in an era of declining PC sales and brickwall in node shrinks is surely going to make execs over at Intel losing sleep at night like a living nightmare unfolding before their eyes unthinkable just 9 years ago when Apple made the switch to x86.
 
Last edited:
The funny thing is that with the A10 if Apple still really needs to drive performance, they can always give it a third core. I know the guys at AT did a test with Android phones using big.LITTLE SoCs to see if they could utilize all of those cores and found out that in some cases they could, and while I don't know how well that translates to iOS, but considering that the A9 is still a dual-core SoC, they have room to grow performance by a big chunk again next year.

We're almost reaching the point where I think that notebooks will just be an empty shell with a bigger screen, better keyboard, and a lot of battery where you can dock your phone. Strange to think that Microsoft is the best equip company to deal with a world like that
 
The funny thing is that with the A10 if Apple still really needs to drive performance, they can always give it a third core. I know the guys at AT did a test with Android phones using big.LITTLE SoCs to see if they could utilize all of those cores and found out that in some cases they could, and while I don't know how well that translates to iOS, but considering that the A9 is still a dual-core SoC, they have room to grow performance by a big chunk again next year.

We're almost reaching the point where I think that notebooks will just be an empty shell with a bigger screen, better keyboard, and a lot of battery where you can dock your phone. Strange to think that Microsoft is the best equip company to deal with a world like that

You make it sound simple but power would be a major issue in a phone. These are big SoCs. Triple-core is even bigger obviously.
 
Yes. A9 should be near its power/thermal limit inside the iPhone 6/plus. You cannot simply tack another core because you feel the need. Heh. Only if it were so easy!
 
Yeah, I really wonder for how many people those great new benchmark number are actually relevant.
My old Nexus 5 still feels plenty fast for practically anything I am doing with it, for my needs, "good enough" speed was reached two years ago.

Now, long battery life in a smallish (<=5") form factor, THAT would pique my interest...

I'm already there, heck I've been there. My iPhone 6 is plenty fast, as was my iPhone 5. In day to day usage I really can't tell much of a difference in the speed. Safari and Facebook can only go so fast. I don't see many people doing a lot on their phones that takes a lot of processing power outside of games. Rather than speed I want more efficiency for better battery life. (Or just quite making the thing thinner and use the space for battery capacity!)
 
Some reports seem to be saying the A9 is in the >150mm^2 range. No idea how that compares to other chips, but it sounds impressive.
 
^ That would put the A9 almost twice the size of the Exynos 7420 which is 78mm². It'd be even bigger than the S810 which is rumored to be around 130~140mm² although the S810 is manufactured on 20nm, not 14nm process.

The A8 was 89mm² on 20nm. Larger than 150mm² for the A9 is.. difficult to believe. (how big the A9X would that make?!) But I won't rule out anything with the A9 at this point. lol.
 
I'll try and find a link. I know I read it in one of the reviews that are currently up, though who knows how reputable the site was.
Let me have a go through reviews.
 
Yes. A9 should be near its power/thermal limit inside the iPhone 6/plus. You cannot simply tack another core because you feel the need.

You can add all kinds of additional silicon and still stay within the thermal budget as long as you don't try to run all of it constantly. If the third CPU core never turns on when the GPU is under heavy load, it's probably not an issue.

In some sense, a little redundancy helps with the thermals as you can have the chip alternate which core is being limited, which distributes the heat across more silicon.

I don't think they have a good reason to add a third core outside of "because we can" reasons or them starting to build beefier chips that will also be put in their notebooks or other computers (although really that's just the X variant of their SoC)

I'm rather impressed that they got as much IPC improvements with the new chip as they did, as the clock increase doesn't explain all of the performance gains. Maybe they can do that again and it's enough for the A10, or perhaps they can go in the other direction and keep the performance fixed while significantly reducing the power consumption. The only other improvement I could see is that fab improvements let them nudge the clocks slightly higher to eek out a little more performance, but I don't see them going beyond 2.0 GHz, and 1.9 is probably a lot more likely.
 
Back
Top