Apple A8x

[Ajay]

Interesting.

http://www.notebookcheck.net/Apple-iPad-Air-2-A1567-128-GB-LTE-Tablet-Review.129396.0.html


Throttling under load. Worse screen. No improvements in battery life.

Apple needs to get off the 'thinner is better' marketing line and provide more battery life instead, IMHO. Then again, they are making billions in profits, so...

 

[Enigmoid]

Of course it throttles under load...it's a passively cooled mobile chip.

They released slides specifically touting the fact that there are no throttling, at least for the iphone.

 

[III-V]

Apple needs to get off the 'thinner is better' marketing line and provide more battery life instead, IMHO. Then again, they are making billions in profits, so...

I believe AnandTech has pointed out that Apple basically does not have anything else to gain from making devices thinner at this point. They need to find another tree to pick fruit from.

There's a point where the form factor essentially cannot be improved upon, and I think Apple's gotten very close to that. Next generations will likely be simple refreshes, with updated hardware in the same or similar chassis. I do think there's a market for a bigger tablet (as in length and width, not in thickness), provided they can scale up the iPad to 12-13 inches, sacrifice virtually nothing in the process, and keep it affordable. That might be too difficult to pull off at this point, though, in terms of the affordability thing.

I also think we'll start seeing Apple be more aggressive with pricing. There's certainly room for these things to come down.

 

[itsmydamnation]

There's a point where the form factor essentially cannot be improved upon, and I think Apple's gotten very close to that. Next generations will likely be simple refreshes, with updated hardware in the same or similar chassis. I do think there's a market for a bigger tablet (as in length and width, not in thickness), provided they can scale up the iPad to 12-13 inches, sacrifice virtually nothing in the process, and keep it affordable. That might be too difficult to pull off at this point, though, in terms of the affordability thing.

Personally i think Something like sonys XTZ in terms of form factor is significantly better , Ipad air feels like a brick in comparison, haven't been able to try ipad air 2 yet. But the reality is their target market doesn't care about cpu performance.

 

[III-V]

Personally i think Something like sonys XTZ in terms of form factor is significantly better , Ipad air feels like a brick in comparison, haven't been able to try ipad air 2 yet. But the reality is their target market doesn't care about cpu performance.

If their customers didn't care about CPU performance, they wouldn't be advertising it. Even if most consumers don't, developers do. Even then, consumers do care; they just don't know it.

 

[2is]

Personally i think Something like sonys XTZ in terms of form factor is significantly better , Ipad air feels like a brick in comparison, haven't been able to try ipad air 2 yet. But the reality is their target market doesn't care about cpu performance.

They absolutely care about performance. It's specs they don't concern themselves too much with. That's why Apple touts the performance and rarely if ever mentions specs.

 

[Arachnotronic]

They absolutely care about performance. It's specs they don't concern themselves too much with. That's why Apple touts the performance and rarely if ever mentions specs.

They also care about performance outside of pure benchmarketing. For example, they optimize the heck out of the operating system, the browser, APIs, and much more.

Apple is unique in that it doesn't just build operating systems or just processors -- it builds computers and actually cares about every detail, from the processor to the touchpad.

 

[Eug]

AnandTech's review:

http://www.anandtech.com/show/8666/the-apple-ipad-air-2-review

CPU Performance:

http://www.anandtech.com/show/8666/the-apple-ipad-air-2-review/3

GPU and NAND Performance:

http://www.anandtech.com/show/8666/the-apple-ipad-air-2-review/4

 

[Nothingness]

The review is missing some CPU throttling test. The T-Rex on screen GPU test doesn't show throttling. Battery life is similar to iPad Air (AT claims this translates into 16% better efficiency).

Definitely a nice device, and likely better than the (at the moment) disappointing Nexus 9.

 

[Eug]

AnandTech: The A8X GPU is NOT the 6650. It is a semi-custom 8-cluster, 2 x 6450 = 6850.

http://www.anandtech.com/show/8716/apple-a8xs-gpu-gxa6850-even-better-than-i-thought

 

[Arachnotronic]

AnandTech: The A8X GPU is NOT the 6650. It is a semi-custom 8-cluster, 2 x 6450 = 6850.

http://www.anandtech.com/show/8716/apple-a8xs-gpu-gxa6850-even-better-than-i-thought

Impressive. Another stepping stone to the eventual custom Apple GPU.

 

[kimmel]

The review is missing some CPU throttling test. The T-Rex on screen GPU test doesn't show throttling. Battery life is similar to iPad Air (AT claims this translates into 16% better efficiency).

Definitely a nice device, and likely better than the (at the moment) disappointing Nexus 9.

Hate to sound like a conspiracy theorist but almost every Arm review is missing throttling tests and yet every Intel review almost explicitly focuses on attempts to make the chip throttle. :hmm:

 

[Nothingness]

Hate to sound like a conspiracy theorist but almost every Arm review is missing throttling tests and yet every Intel review almost explicitly focuses on attempts to make the chip throttle. :hmm:

The current problem is that the devices using Core M seem to throttle a lot, so it looks like reviews are focusing on it. They are just discovering that the benchmarks they are used to run, do throttle on Core M. So they are making a lot of noise about it

You can see that differently: most smartphone benchmarks are so light that the devices don't throttle...

 

[Eug]

MacRumors has published Chipworks' die shot:

 

[kimmel]

The current problem is that the devices using Core M seem to throttle a lot, so it looks like reviews are focusing on it. They are just discovering that the benchmarks they are used to run, do throttle on Core M. So they are making a lot of noise about it

You can see that differently: most smartphone benchmarks are so light that the devices don't throttle...

Furmark+Prime tends to make mobile chips "throttle a lot" where as Geekbench doesn't. I think the simplest explanation is that the types of people who review phones and tablets, think watching videos and browsing the web as usage models to test where the types of people who review PC's expect testing models to revolve around the ability to render the next Pixar movie on your tablet.

The main reason I enjoyed Anand's reviews is that they attempted to apply a bit more critical thought to mobile platform testing instead of the fanboy based review sites. I hope that continues under the new management though I am doubtful. Interesting work is often much harder than just kicking off a few apps and regurgitating their output.

Back to A8X. An incredible feat of engineering, but one we still don't really know much about. Maybe David Kanter will get some more technical information out at some point.

 

[Arachnotronic]

Back to A8X. An incredible feat of engineering, but one we still don't really know much about. Maybe David Kanter will get some more technical information out at some point.

Kanter works for The Microprocessor Report...anything juicy he comes across will probably be exclusive to subscribers. Which is exactly why they'd hire someone with his talents

 

[Enigmoid]

Furmark+Prime tends to make mobile chips "throttle a lot" where as Geekbench doesn't. I think the simplest explanation is that the types of people who review phones and tablets, think watching videos and browsing the web as usage models to test where the types of people who review PC's expect testing models to revolve around the ability to render the next Pixar movie on your tablet.

The main reason I enjoyed Anand's reviews is that they attempted to apply a bit more critical thought to mobile platform testing instead of the fanboy based review sites. I hope that continues under the new management though I am doubtful. Interesting work is often much harder than just kicking off a few apps and regurgitating their output.

Back to A8X. An incredible feat of engineering, but one we still don't really know much about. Maybe David Kanter will get some more technical information out at some point.

Prime is heavier than anything you will encounter on a ARM SOC. Furmark is also significantly heavier than GFX bench which most seem to be using.

 

[Nothingness]

Kanter works for The Microprocessor Report...anything juicy he comes across will probably be exclusive to subscribers. Which is exactly why they'd hire someone with his talents

Since he's been recruited back on March, I have seen nothing significant coming from him... In fact he writes about things/products I'm not interested in such as LTE protocol or path tracing in latest NVIDIA GPU. I certainly miss his articles at RWT

 

[Eug]

This makes me think even more that there is a 12" MaxiPad hybrid waiting in the wings for release in early 2015.

 

[witeken]

This makes me think even more that there is a 12" MaxiPad hybrid waiting in the wings for release in early 2015.

Will it be powered by an ARM chip? If so, what would make it stand out compared to the regular Air?

 

[Eug]

Will it be powered by an ARM chip? If so, what would make it stand out compared to the regular Air?

If you're asking about the MacBook Air (not iPad Air) I'm guessing it will be an iPad, with optional keyboard, but one that is Apple sanctioned.

Truthfully, I'm more convinced about the iPad part, and less convinced about the keyboard part.

 

[Zink]

Will it be powered by an ARM chip? If so, what would make it stand out compared to the regular Air?

I think the key differentiating feature to have at launch for a ~12.5" ipad would be side-by-side multitasking. Having two iPad Mini sized apps open each taking exactly half the screen would be a very Apple like UI choice. They would hopefully increase the resolution to iPad Mini like DPI (3k resolution?) to make it super nice for productivity use.

With the extra core and bandwidth A8X has the performance to run any kind of heavy editing or content creation apps side by side with the same performance as a single app on previous iPads. If they wanted, Apple would have the opportunity to introduce mouse support for specific productivity applications if they really want to try to push ARM based content creation. Using a keyboard but no mouse is a bottleneck when editing text at a desk and probably for some audio/video creation apps.

 

[III-V]

MacRumors has published Chipworks' die shot:

Wow, look how synthesized that CPU is -- hadn't realized that until you posted this image. Swift and Cyclone were very obviously laid out by hand. Very interesting that performance still managed to move as far as it did.

 

[Idontcare]

Wow, look how synthesized that CPU is -- hadn't realized that until you posted this image. Swift and Cyclone were very obviously laid out by hand. Very interesting that performance still managed to move as far as it did.

A number of of Bobcat design engineers got snapped up by Apple.

 

[Nothingness]

Wow, look how synthesized that CPU is -- hadn't realized that until you posted this image. Swift and Cyclone were very obviously laid out by hand.

It certainly makes sense to massively uses synthesis when foundries move so fast at the moment and when you want to have multiple sources. Add to that it gives the ability to quickly make changes, and that you get the best of both worlds by just using custom cells where they make the most benefit, and you get a wise decision.

Very interesting that performance still managed to move as far as it did.

It's similar to what you see in software development: do the development at a high level (synthesis), then optimize what matters (custom cells), that will give you most of the achievable performance. At least that's how I see it, and how I have seen it done