The iPad Pro was supposed to ship with A8X. A9X was pushed out 6 mos early.

[Eug]

http://www.bloomberg.com/features/2016-johny-srouji-apple-chief-chipmaker/

A little over a year ago, Apple had a problem: The iPad Pro was behind schedule. Elements of the hardware, software, and accompanying stylus weren’t going to be ready for a release in the spring. Chief Executive Officer Tim Cook and his top lieutenants had to delay the unveiling until the fall. That gave most of Apple’s engineers more time. It gave a little-known executive named Johny Srouji much less.

Srouji is the senior vice president for hardware technologies at Apple. He runs the division that makes processor chips, the silicon brains inside the iPhone, iPad, Apple Watch, and Apple TV. The original plan was to introduce the iPad Pro with Apple’s tablet chip, the A8X, the same processor that powered the iPad Air 2, introduced in 2014. But delaying until fall meant that the Pro would make its debut alongside the iPhone 6s, which was going to use a newer, faster phone chip called the A9.

This is the stuff that keeps technology executives up at night. The iPad Pro was important: It was Apple’s attempt to sell tablets to business customers. And it would look feeble next to the iPhone 6s. So Srouji put his engineers on a crash program to move up the rollout of a new tablet processor, the A9X, by half a year.

---

What I think this means is that the iPad Air 3 will have A9X, albeit probably slower clocked.

 

[StinkyPinky]

Which means it will be a good time to get the Ipad Air 2 for cheap. I have the Air 2 and it's fantastic, plus the 2gb ram makes it a good investment. I wonder if the Air 3 will have 3gb ram.

 

[manly]

Which means it will be a good time to get the Ipad Air 2 for cheap. I have the Air 2 and it's fantastic, plus the 2gb ram makes it a good investment. I wonder if the Air 3 will have 3gb ram.

To generalize, buying an Apple product just before it's replaced is usually a bad idea.

The iPad Air 2 itself is a great device and nobody buying it now should be displeased. However, you can literally wait a month to see if the Air 3 is substantially better. It's always possible the Air 3 will not be a good upgrade; i.e. if it came with a 4k display but didn't have the GPU or RAM to feed the beast.

For business reasons, I actually believe Apple clocked the A9 too high. They could've kept it somewhere around 1.7 GHz, and saved some performance for the subsequent A10.

 

[Eug]

To generalize, buying an Apple product just before it's replaced is usually a bad idea.

The iPad Air 2 itself is a great device and nobody buying it now should be displeased. However, you can literally wait a month to see if the Air 3 is substantially better. It's always possible the Air 3 will not be a good upgrade; i.e. if it came with a 4k display but didn't have the GPU or RAM to feed the beast.

Agreed. iPad Air 2 is very good, but it pays to just wait a month. I learned the hard way. I wanted a 12" PowerBook. Apple released a new PowerBook 2 months before the usual release date... but it was a 15" and nothing else (like in previous years). So I was dejected, but bought it. Then 2 months later at the usual time, they released the 12" and the 17" PowerBook.

For business reasons, I actually believe Apple clocked the A9 too high. They could've kept it somewhere around 1.7 GHz, and saved some performance for the subsequent A10.

Why are you assuming the A10 isn't going to be noticeably better? It's likely going to be on a 10 nm process for example.

 

[Blackjack200]

This is the stuff that keeps technology executives up at night. The iPad Pro was important: It was Apple’s attempt to sell tablets to business customers. And it would look feeble next to the iPhone 6s. So Srouji put his engineers on a crash program to move up the rollout of a new tablet processor, the A9X, by half a year.

JFC. Feel for those guys.

This kind of surprises me, I thought the A9X was necessary to drive all those extra pixels, but I guess not?

 

[manly]

Agreed. iPad Air 2 is very good, but it pays to just wait a month. I learned the hard way. I wanted a 12" PowerBook. Apple released a new PowerBook 2 months before the usual release date... but it was a 15" and nothing else (like in previous years). So I was dejected, but bought it. Then 2 months later at the usual time, they released the 12" and the 17" PowerBook.

Well historically, it was a bit difficult to predict new Mac releases unless you could closely follow the reseller channel. Except for January/February, it could happen at any time. In the Tim Cook era, Apple likes to stack multiple hardware releases in calendar Q4 which IMO makes the rest of the year somewhat stale.

Their business model is way too iPhone-dependent now, so you've see silly things like the 2015 15" rMBP released just before a suitable Broadwell U CPU is out. Or once again, the Mac Pro's last update was over 2 years ago...

Why are you assuming the A10 isn't going to be noticeably better? It's likely going to be on a 10 nm process for example.

I forget whose review it was, probably AT's that pointed out the A9 was the biggest Apple SoC performance upgrade in a very long time. Apple didn't know during development that Qualcomm Snapdragon would have an off year, or else they could've sandbagged a bit of the performance jump.

Intel has already pushed its 10 nm node out another year. I highly doubt TSMC or any other foundry is going to beat Intel to 10 nm. Hope you got the memo, Moore's Law is dead.

Hope I didn't segue too much in your thread.

 

[Eug]

I thought the A9X was necessary to drive all those extra pixels, but I guess not?

This all makes sense now. When the iPad Air 2 came out I kinda thought its SoC was overkill, almost like a desperation move to maintain the performance advantage, given its large size and triple cores. But now we know it wasn't even designed for the iPad Air 2 in the first place.

I forget whose review it was, probably AT's that pointed out the A9 was the biggest Apple SoC performance upgrade in a very long time. Apple didn't know during development that Qualcomm Snapdragon would have an off year, or else they could've sandbagged a bit of the performance jump.

Intel has already pushed its 10 nm node out another year. I highly doubt TSMC or any other foundry is going to beat Intel to 10 nm. Hope you got the memo, Moore's Law is dead.

TSMC will begin 10nm production this year, claims 5nm by 2020

10 nm tape out early this year with 10 nm production starting this year and higher volume 10 nm in 2017. In other words, they are claiming they will beat Intel to 10 nm. This doesn't mean that the iPhone 7 SoC will be 10 nm but it is within the realm of possibility.

It should be noted that some are saying TSMC's initial 10 nm won't be as good as Intel's 10 nm, and more like a hybrid of 10 nm and 14 nm with no EUV (whatever that is) but that's irrelevant for our discussion, since we're comparing against the TSMC (and Samsung) produced Apple A9, as Intel doesn't make Apple SoCs.

 

[manly]

TSMC will begin 10nm production this year, claims 5nm by 2020

10 nm tape out early this year with 10 nm production starting this year and higher volume 10 nm in 2017. In other words, they are claiming they will beat Intel to 10 nm. This doesn't mean that the iPhone 7 SoC will be 10 nm but it is within the realm of possibility.

It should be noted that some are saying TSMC's initial 10 nm won't be as good as Intel's 10 nm, and more like a hybrid of 10 nm and 14 nm with no EUV (whatever that is) but that's irrelevant for our discussion, since we're comparing against the TSMC (and Samsung) produced Apple A9, as Intel doesn't make Apple SoCs.

I'd be surprised if TSMC beats Intel to commercial production at 10 nm, but it's certainly possible with Cannonlake being delayed by a year.

It's unclear what TSMC means by "beginning production" in 2016 but I can guarantee you there's no way they can deliver tens of millions of units on a new process node starting in September.

 

[Rakehellion]

What I think this means is that the iPad Air 3 will have A9X, albeit probably slower clocked.

Or it'll have an A8X.

 

[Eug]

I'd be surprised if TSMC beats Intel to commercial production at 10 nm, but it's certainly possible with Cannonlake being delayed by a year.

It's unclear what TSMC means by "beginning production" in 2016 but I can guarantee you there's no way they can deliver tens of millions of units on a new process node starting in September.

Some of the CPU nerds say TSMC's "10 nm" is probably closer to Intel's 14 nm than Intel's 10 nm, but as mentioned, in terms of the Apple Ax line of chips it's irrelevant, since it will still be a generational improvement over TSMC's previous processes.

Or it'll have an A8X.

Not a chance. That's what Apple iPad Air 2 has.

 

[Arachnotronic]

This all makes sense now. When the iPad Air 2 came out I kinda thought its SoC was overkill, almost like a desperation move to maintain the performance advantage, given its large size and triple cores. But now we know it wasn't even designed for the iPad Air 2 in the first place.


TSMC will begin 10nm production this year, claims 5nm by 2020

10 nm tape out early this year with 10 nm production starting this year and higher volume 10 nm in 2017. In other words, they are claiming they will beat Intel to 10 nm. This doesn't mean that the iPhone 7 SoC will be 10 nm but it is within the realm of possibility.

It should be noted that some are saying TSMC's initial 10 nm won't be as good as Intel's 10 nm, and more like a hybrid of 10 nm and 14 nm with no EUV (whatever that is) but that's irrelevant for our discussion, since we're comparing against the TSMC (and Samsung) produced Apple A9, as Intel doesn't make Apple SoCs.

A10 is on 16FF+, not TSMC 10nm. Will use InFO packaging.

 

[Arachnotronic]

Or it'll have an A8X.

iPad Air has an A9X, iPhone 5se is A9. A8/A8X are old news, not to be used in future products.

 

[Arachnotronic]

Which means it will be a good time to get the Ipad Air 2 for cheap. I have the Air 2 and it's fantastic, plus the 2gb ram makes it a good investment. I wonder if the Air 3 will have 3gb ram.

Air 3 will likely have 4GB of RAM.

 

[stingerman]

Their business model is way too iPhone-dependent now, so you've see silly things like the 2015 15" rMBP released just before a suitable Broadwell U CPU is out. Or once again, the Mac Pro's last update was over 2 years ago...

The real problem is that the performance improvement of the Intel processors from year to year are so little that an upgrade is hardly justified.

 

[stingerman]

A10 is on 16FF+, not TSMC 10nm. Will use InFO packaging.

With InFO packaging Apple can mix nodes on the same Silicon Interposer. But, I do believe like you it will be 16FF+, though with some more improvements. What's exciting is the potential that they'll add the latest HBM v2 and other major component die to the same package.