Apple A10X SoC

Apple announced iPad Pro updates with the A10X SoC today.

screen_shot_2017-06-05_at_11.47.02_am.png

(Image from AnandTech.)

30% faster CPU performance and 40% faster GPU performance than the A9X.
 
Wonder if the A11 will have Apple GPU cores instead of PowerVR.

Also with all the iOS 11 iPad-specific features, maybe they embed more RAM?

Better screens on the new iPad Pros compared to the previous ones. 120 Hz refresh rate.
 
The iPad Pro screens really felt like the stars of the show. Appropriate for devices where the screen basically is the device.
The SoC is obviously a beast, but I'd assume from signals from TSMC that it is a 16nm design, so if they have achieved an overall 30% single thread performance increase (which the slide doesn't claim, but nor can 30% refer to multi threaded benchmark performance), it would have to involve blood magic.
I want in on that, so I hope we get some benchmarks out soon or the temptation to test it out for myself may overwhelm my better judgement. :)

Whether the A11 incorporates an in-house graphics solution remains to be seen, but odds are the A10x doesn't. I can't imagine they would make such a shift without any kind of hint or comment. They were quite clear about designing their own CPUs, athough people doubted it anyway.

Also, Metal 2 is of B3D interest, but maybe not for this thread?
 
The SoC is obviously a beast, but I'd assume from signals from TSMC that it is a 16nm design, so if they have achieved an overall 30% single thread performance increase (which the slide doesn't claim, but nor can 30% refer to multi threaded benchmark performance), it would have to involve blood magic.
What are these signals from TSMC?

Previous rumors claimed that the A10X will be on 10 nm.

Also, Metal 2 is of B3D interest, but maybe not for this thread?
There's a thread here about it.
 
What are these signals from TSMC?

Previous rumors claimed that the A10X will be on 10 nm.
I know, but TSMC has made statements regarding shipping product and projected revenues that could only barely (if at all) fit with iPad Pros being available in Apple volumes right now. If the A10x was fabbed on TSMC 10nm, that would be worthy of a headline or two, particularly in the circles that is more connected to supply chain as opposed to consumers.
Until we know better, I'll assume that the A10x uses the same lithographic process node as the A10.
 
I wonder if the A10X performance improvements over the A9X will be similar at each iPad Pro form factor. The current 12.9" model is still far and away the most powerful device in its class.
 
I wonder if the A10X performance improvements over the A9X will be similar at each iPad Pro form factor. The current 12.9" model is still far and away the most powerful device in its class.
It would seem reasonable to broadly assume so. The larger iPad still has room for a larger battery (and still need to push around more pixels), so as long as battery times are quoted as similar, I think it's reasonable to assume that the larger iPad takes advantage of it's larger battery and larger power dissipating enclosure to increase max clocks a bit over its smaller sibling. We'll know soon enough, you can order one right now and benchmark it for us! :D
 
I wonder if the A10X performance improvements over the A9X will be similar at each iPad Pro form factor. The current 12.9" model is still far and away the most powerful device in its class.
Before the iPad Pro update, I thought that was likely to happen, but it turns out that the two A10X's have the same performance according to Apple's numbers.

From the iPad comparison page:
Code:
12.9-inch iPad Pro      10.5-inch iPad Pro

Compared to A8:         Compared to A8:
CPU: 2.5x faster        CPU: 2.5x faster
Graphics: 4.3x faster   Graphics: 4.3x faster

This wasn't the case with the previous 12.9" and 9.7" iPad Pros.
Code:
12.9-inch iPad Pro      9.7-inch iPad Pro

Compared to A8:         Compared to A8:
CPU: 1.9x faster        CPU: 1.85x faster
Graphics: 3.1x faster   Graphics: 2.7x faster

That being said, I don't know how Apple's measuring CPU and GPU performance so these numbers may not tell the whole story.
 
Before the iPad Pro update, I thought that was likely to happen, but it turns out that the two A10X's have the same performance according to Apple's numbers.

From the iPad comparison page:
Code:
12.9-inch iPad Pro      10.5-inch iPad Pro

Compared to A8:         Compared to A8:
CPU: 2.5x faster        CPU: 2.5x faster
Graphics: 4.3x faster   Graphics: 4.3x faster

This wasn't the case with the previous 12.9" and 9.7" iPad Pros.
Code:
12.9-inch iPad Pro      9.7-inch iPad Pro

Compared to A8:         Compared to A8:
CPU: 1.9x faster        CPU: 1.85x faster
Graphics: 3.1x faster   Graphics: 2.7x faster

That being said, I don't know how Apple's measuring CPU and GPU performance so these numbers may not tell the whole story.

So the performance gains in both are based over the previous 12.9 inches model.... cool

They seem to use Geekbench for CPU and GFXbench for the GPU.
 
Single threaded score is some 13% over the A10 in iPhone 7 at the same clocks.
Subtest scores do not scale linearly (would indicate fake entry). On the other hand, although there are subtests that swing more than the others both up and down, there is no clear outlier that skews the overall mean excessively. L2 cache has increased from 3MB to 8MB, and of course main memory bandwidth is almost doubled, so real world performance might be a bigger improvement than these 13%. And that is still ignoring the benefits of a third core completely.
 
Of course they won't scale linearly if the memory and caches changed that much. Some tests are more memory heavy than others.
Yes. The more intriguing cases are where performance has regressed. I'd be wary of drawing too detailed conclusions from a single benchmark run under undisclosed circumstances though. We have no idea what that system may or may not have been doing in parallel during the run.
 
Windows on arm was here and failed, it was Windows RT. Yes, not quite exactly the same, but no one liked having similiar yet different applications to x86 Windows. If anyone is able to switch to ARM it must have either every application built on a framework VM (like .Net CLI) or be purely Interpreted executable layer or have capabilities of translating x86 to ARM on the fly.
 
Windows on arm was here and failed, it was Windows RT. Yes, not quite exactly the same, but no one liked having similiar yet different applications to x86 Windows. If anyone is able to switch to ARM it must have either every application built on a framework VM (like .Net CLI) or be purely Interpreted executable layer or have capabilities of translating x86 to ARM on the fly.

Android on x86 was here also and failed. Or maybe better put x86 failed, and not due to a castrated OS like RT.
The ability of Windows on Apple hardware via boot camp probably is a key selling factor, at least it has been for me.
If Apple would switch to ARM CPUs, it still would have this possibility via Windows on ARM.
 
That must be becoming very favorably compared to the Kaby lake mobile Intel CPUs.
It must be very tempting for Apple to switch to ARM on its laptops at some point.
yes Intel will be shitting themselves, 'literally'. As I have written here before, It is inevitable that Apple will be moving to ARM for their laptops, this fanless CPU is within spiting distance of the top of the line intel macbook pro 'fan cooled' cpu's.
 
Back
Top