Apple A10 Fusion is ** Quad-core big.LITTLE **

[Eug]

So, a lot of the tech blogs continue to say the Apple A10 will be on 10 nm TSMC. This is wrong from what I gather, correct? My understanding it will remain on TSMC's 16 nm. (No more Samsung.)

Nonetheless I'm curious what type of speed boost there will be going from A9 to A10. It seems Apple has consistently increased performance significantly from generation to generation, but in many cases a lot of that performance increase could be attributed to the smaller process used, although the A4->A5's increase was due to going dual-core at the expense of power utilization.

Apple A4: 45 nm
Apple A5: 45 nm
Apple A5: 32 nm with better battery life than A5 45 nm
Apple A6: 32 nm
Apple A7: 28 nm
Apple A8: 20 nm
Apple A9: 16 nm TSMC or 14 nm Samsung
Apple A10: 16 nm TSMC?

For comparative benchmarketing scores, check out this stock blog (not mine):

http://www.fool.com/investing/gener...-the-performance-of-the-apple-inc-a10-ch.aspx

Are you guys expecting a big boost in speed from A10 over A9, or will the fact it's on the same process as A9 be a big handicap? In the early days, Apple's designs improved by leaps and bounds, but A9 is comparatively much more mature, so one wonders why whiz-bang design and performance improvements PA Semi/Apple can come up with this time without the help of a process shrink.

I will likely buy an iPhone 7 next month with A10. Ironically though, CPU performance isn't my main concern. I'm currently on an iPhone 5S which has too little memory, and my aging eyes are having trouble with the small fonts these days. I really want to wait for Apple 11 and the iPhone 2017 with a new form factor, but the iPhone 7 will have to do. Hopefully the iPhone 7 Plus/Pro will come with 3-4 GB RAM.

EDIT Sept. 7:

Apple A10 Fusion: big.LITTLE Quad-core

Here are AnandTech pix from the live blog of the launch:

 

[imported_bman]

Apple has pulled off some impressive performance increases through architecture changes before, but I wonder if their architecture is mature and future IPC increases will only be marginal. I doubt 10nm, every rumor indicate that Samsung will be first to 10nm, 16nm with fanout. Fanout is supposed to improve thermals (~10%), reduce Z-height (~20%), and can increase performance (~20%) so I would still expect the A10 to be a decent improvement. The GPU could get a sizable increase if they move to the PowerVR Series 8XT (assuming it exists). Apple will push their new 'features' and marginal upgrades: dual camera on the pro, True Tone on the phone, new head phone setup (reason for ''), marginally bigger battery, 4k 60fps recording. The RAM upgrade seems questionable, Apple has always been chincy on the RAM, two upgrades in row would be something.

Apple has a lot of products to be upgraded, so it should be a good show. I hope they reveal the Watch 2, expecting a move to 16nm, better connectivity, GPS, and longer battery life. I suspect we will also get an improved Ipad Pro with A10, True Tone, and 4k recording. A refresh of the Mac Pro, iMacs, and the Macbook Pro.

 

[avAT]

A10 is rumored to be 16nm, and A10X is rumored to be 10nm but not ship until Q1 2017.

Purported A10 Geekbench shows 3548 single-core http://www.macrumors.com/2016/08/09/iphone-7-plus-geekbench-results-reveal-a10-3gb-ram/

 

[Eug]

Purported A10 Geekbench shows 3548 single-core http://www.macrumors.com/2016/08/09/iphone-7-plus-geekbench-results-reveal-a10-3gb-ram/

Primate Labs says those results are fake. I guess we'll have to wait a a bit longer for legit results, but they should show up in a few weeks. What I'm really hoping for though is the 3 GB RAM.

 

[sm625]

I expect it to score 1200 on Kraken 1.1, about 22000 on Octane v2. Not as big a leap as the last gen, but still way above any android device.

 

[avAT]

Primate Labs says those results are fake. I guess we'll have to wait a a bit longer for legit results, but they should show up in a few weeks. What I'm really hoping for though is the 3 GB RAM.

Bah, fake!

But yeah, A10 is widely expected to still be 16nm. I expect a minimum of 25% faster, but not as much as last time. I think we'll see 3GB, but only in the Plus/Pro.

 

[teejee]

Bah, fake!

But yeah, A10 is widely expected to still be 16nm. I expect a minimum of 25% faster, but not as much as last time. I think we'll see 3GB, but only in the Plus/Pro.

My guess is less improvement. Maybe 15% in single thread Geekbench or so ( about 50/50 IPC and frequency).

Sent from my LG-D855 using Tapatalk

 

[Lepton87]

It would be awesome if they added some kind of Simultaneous multithreading capability. It should make it competitive with Intel's skylake both in ST and MT or even faster! Normalized to the same clockspeed and core count of course.

 

[Eug]

But yeah, A10 is widely expected to still be 16nm. I expect a minimum of 25% faster, but not as much as last time. I think we'll see 3GB, but only in the Plus/Pro.

It would be awesome if they added some kind of Simultaneous multithreading capability. It should make it competitive with Intel's skylake both in ST and MT or even faster! Normalized to the same clockspeed and core count of course.

My guess is less improvement. Maybe 15% in single thread Geekbench or so ( about 50/50 IPC and frequency).

A 25%+ improvement in peformance on the same 16 nm seems rather optimistic to me, too. I personally am a n00b at this stuff, but with the discussion of the A9, Ryan Smith and Joshua Ho felt that it would be unlikely the A10 would achieve a 25% boost over A9:

However with the iPhone 6s, all of the stars are coming into alignment for Apple. On the one hand as this is an iPhone S release, even more is expected of them on the architectural side of matters. On the other hand between the power benefits of the FinFET processes and Twister’s place in Apple’s seeming 2-year cycle, Apple will get to run up the score twice: once with clockspeed and once with a more substantial architecture improvement.

In fact on the clockspeed front this is the biggest jump in CPU frequencies since Swift in the A6, where Apple went from an 800MHz ARM Cortex-A9 to the aforementioned custom Swift design at 1.3GHz. As a result Apple immediately gets to capitalize on a 450MHz (32.1%) clockspeed bump for Twister in the A9 versus the Typhoon-powered A8. That large of a clockspeed bump alone would be enough to give Apple a sizable performance boost, especially as competing designs are already at 2GHz+ and are unlikely to shoot much higher due to power concerns.

Apple has always played it conservative with clockspeeds in their CPU designs – favoring wide CPUs that don’t need to (or don’t like to) clock higher – so an increase like this is a notable event given the power costs that traditionally come with higher clockspeeds. Based on the underlying manufacturing technology this looks like Apple is cashing in their FinFET dividend, taking advantage of the reduction in operating voltages in order to ratchet up the CPU frequency. This makes a great deal of sense for Apple (architectural improvements only get harder), but at the same time given that Apple is reaching the far edge of the performance curve I suspect this may be the last time we see a 25%+ clockspeed increase in a single generation with an Apple SoC.

And memory bandwidth is likely no longer a low hanging fruit, as they already massively improved that with A9 over A8.

In fact, the more I read about this stuff, the less optimistic I am about CPU gains, esp. given Apple's general reluctance to increase phone thickness and battery size. More RAM and better camera seem like the better targets.

 

[avAT]

A 25%+ improvement in peformance on the same 16 nm seems rather optimistic to me, too. I personally am a n00b at this stuff, but with the discussion of the A9, Ryan Smith and Joshua Ho felt that it would be unlikely the A10 would achieve a 25% boost over A9

I'm a total silicon n00b. I'm basing it on the fact that the A_X line has so far been >= 25% each year (Geekbench single core), but looking at the phone chips I see they have been less at times. A7 to A8 was 15% so I guess that's my minimum expectation.

I haven't read anything about it recently, but I remember A9 was expected to be 16nm FF and A10 16nm FF+. Assuming that ended up being the case, I have no idea if there is much/any difference.

 

[Arachnotronic]

I'm a total silicon n00b. I'm basing it on the fact that the A_X line has so far been >= 25% each year (Geekbench single core), but looking at the phone chips I see they have been less at times. A7 to A8 was 15% so I guess that's my minimum expectation.

I haven't read anything about it recently, but I remember A9 was expected to be 16nm FF and A10 16nm FF+. Assuming that ended up being the case, I have no idea if there is much/any difference.

A10 is a 16FF+ product, not 10nm.

 

[jpiniero]

I remember an earlier rumor putting the CPU performance of the 7 Plus to that slightly behind that of the iPad Pro. Which would be like +20%ish.

 

[pcslookout]

I can't wait until the A10 !

 

[Eug]

The part leaks continue to come out, this time with a pic of part of the actual A10 package.

The production date is 1628, which is about a month ago, and the leak is from GeekBar, which has leaked legit pre-release info several times before.

http://www.macrumors.com/2016/08/10/apple-a10-chip-first-photo-weibo/

What is pictured here is not the complete A10 chip and may be the RAM layer that is stacked on top of the A10 wafer as one system-on-a-chip, rather than the processor itself. Apple A-series chips are typically labeled on all four edges, suggesting this chipset is in a mid-production state.

TSMC is expected to be the sole supplier of Apple's faster A10 chip based on its 16 nm FinFET WLP process. The new processor should bring performance improvements to the iPhone 7 and iPhone 7 Plus. The smartphones are expected to be announced on September 7 and released on September 16.

Now we just need actual benches, and info on the amount of RAM.

 

[Eug]

I'm not sure how much insight into chip technology MacRumors has, and some of this is old news, but I'll post this anyway:

How TSMC Won Back Exclusivity With Apple for the A10 Chip in iPhone 7

TSMC confirmed in conference call comments that its chip packaging changes have led to improvements of 20 percent in both speed and packaging thickness and 10 percent in thermal performance. This has a number of implications for future device performance and future foundry partner selection for Apple.

The promised performance improvements are certainly significant. A 20 percent improvement in performance is roughly equivalent to the improvement expected between successive foundry nodes (e.g. the change from 14 nm to 10 nm). With both TSMC and Samsung only offering refined versions of their 16 nm and 14 nm FinFET processes, this means that total performance gains could be in line with the same improvement seen from the A8 to the A9 chip, but driven by packaging improvements rather than a new process.

 

[Eug]

A9 on left, A10 on right.

http://cdn.macrumors.com/article-new/2016/08/a9_a10_pinout.jpg

 

[Eug]

Still no performance bench leaks.

I. Am. Disappoint.

I remember an earlier rumor putting the CPU performance of the 7 Plus to that slightly behind that of the iPad Pro. Which would be like +20%ish.

Where is that from? Sounds more like speculation, but +20% would be reasonable, even if it's a way lower increase than previous iterations. I'm hoping for more though.

 

[jpiniero]

I dunno, but there is a new leak saying that the clock speed is going to be close to 2.4 Ghz. That would be 30% faster just on clock than A9 and faster than the iPad Pro.

 

[Eug]

I dunno, but there is a new leak saying that the clock speed is going to be close to 2.4 Ghz. That would be 30% faster just on clock than A9 and faster than the iPad Pro.

That'd be nice, but that's not a leak. That's a prediction, from a guy with a decent track record, but still just a prediction.

http://www.macrumors.com/2016/09/03/iphone-7-5-colors-ipx7-12mp-cameras/

The claim is that it could max out at about 2.4 to 2.45 GHz, but he also hedges and says it could be less for heat & power reasons. If that's the case, maybe it would be say 2.2 GHz in the iPhone, but 2.45 GHz in an iPad later on this year. If 2.1-2.2 GHz, then that would be a 14%-19% clock speed boost over the 1.85 GHz of the A9. (The A9X in the iPad Pro is 2.26 GHz in the 12.9" iPad Pro, and 2.16 GHz in the 9.7" iPad Pro.) Add in a bit of extra IPC and you get say a 20-30% performance boost over the A9.

With an iPad Pro of 2.45 GHz, that would be an 11%-17% boost in clock speed over the iPhone at 2.1-2.2 GHz.

But, I'm just pulling numbers out of thin air.

 

[Eug]

Another Apple A10 Geekbench score was posted, as mentioned in this thread. Single-core 3379 and multi-core 5495, Geekbench 4.0.0 for iOS AArch64. I believe that is likely fake too, but it got me to download Geekbench 4, since that's new. In fact, Geekbench 4.0.0 just came out last week, and it's currently free for iOS, until early next week.

http://geekbench.com/

For reference, here are my iPad Air 2 and my iPhone 5S. I didn't bother running it on our 6s as the scores are readily available, including in my screengrabs.

iPad Air 2 (A8X):

iPad 5S (A7):

 

[Eug]

I don't know what this means but Apple has filed for a trademark on "A10 Fusion".

 

[Eug]

big.LITTLE Quad-core

Now we know why they call it A10 Fusion.

 

[ChronoReverse]

Big Little. I guess even Apple had to use it.

 

[witeken]

LOL. Sign of power hungry cores (like A15), even with FinFET?

Nice job, Apple!

 

[Andrei.]

Laughing my ass off here. Big middle finger to the big.LITTLE naysayers over the last few years. I guess that good micro-architecture just can't compensate for the laws of physics.