What if the New 12" MBA Used an A8 Chip?

[iRun26.2]

well idk what you mean by make up for? A "retina" and a IGZO screen are pretty both just run of the mill high resolution screens?

An IGZO screen uses less power than a normal screen at the same size and resolution. Controlling more pixels require more power but because IGZO is more efficient you can have more pixels yet use a similar power.

Then there is the 30% power savings from Broadwell...

There are plenty of ways Apple can keep the same battery life in going to a retina display (without having to downgrade to an A8 processor).

With an A8 processor, you could greatly reduce the weight and increase the battery life even more (if you could live without x86 software)

 

[iStiggy]

The new Quad Core Atoms are much faster than the old Atoms. You can't compare the two (A7 vs Atom) anyways due to architectural differences.

The Atom Z3770 is a quad core with a base clock .16ghz higher than the A7's highest clock. Also Anandtech compared the 2 CPU's against each other.

Z3770: http://ark.intel.com/products/76760/Intel-Atom-Processor-Z3770-2M-Cache-up-to-2_39-GHz
Anandtech:http://www.anandtech.com/show/7335/the-iphone-5s-review/6

 

[ha1o2surfer]

The Atom Z3770 is a quad core with a base clock .16ghz higher than the A7's highest clock. Also Anandtech compared the 2 CPU's against each other.

Z3770: http://ark.intel.com/products/76760/Intel-Atom-Processor-Z3770-2M-Cache-up-to-2_39-GHz
Anandtech:http://www.anandtech.com/show/7335/the-iphone-5s-review/6

This is just a web browser test. I'd love to see Adobe After effects run on an A7 and actually play and edit AVCHD.. 2 different architectures can't be compared in anyway.

In LinPack I get 23gflops on my Dell Venue Pro 8. My iPhone 5s get 1gflops at most.

 

[cambookpro]

I quickly plotted some Geekbench (not the be-all and end-all, but a good comparative benchmark to use) scores of iOS vs Mac devices. I took the most powerful iOS device from that year, and the low-end Mac from that year.

OLD DATA - DISCARD Interestingly, the iPad Air benchmarks faster than the MacBook Air Apple was selling only a few years ago in late 2010. Possibly with more room inside a laptop, a A8 or A9 chip could be used in the future for a lower-end notebook.

Interestingly, the iPad Air benchmarks about 2/3 of the way between a 2010 and 2011 MBA. I've added a trend line, and if the rates of improvement continue at this pace (a big if, mind you) we may see Apple ARM chips be at a par with Intel by around 2015. Of course, ARM still doesn't present any real advantages, but if by 2017ish the current trend has continued, they may be as fast and much more power efficient. I think we may well see a fanless 12" MBA this year, though still expect it to use possibly an underclocked i3 or i5.

This doesn't mean that I think Apple will go down the ARM route - breaking compatibility with all x86 apps isn't going to be a popular choice. I just thought it'd be interesting to see how near Apple's ARM chips are to possibly ever powering something more than an iOS device.

 

[Nee412]

The A7 and the future A8 are still less powerful than the current crop of Intel chips. Maybe in a few years when we reach the A10, or whatever they have at that point, things will be different.

For now though the Intel chips win on performance and compatibility. Therefore Apple will continue to use Intel chips for the time being. It's common sense.

 

[iRun26.2]

I quickly plotted some Geekbench (not the be-all and end-all, but a good comparative benchmark to use) scores of iOS vs Mac devices. I took the most powerful iOS device from that year, and the low-end Mac from that year.

Image

Interestingly, the iPad Air benchmarks faster than the MacBook Air Apple was selling only a few years ago in late 2010. Possibly with more room inside a laptop, a A8 or A9 chip could be used in the future for a lower-end notebook.

This doesn't mean that I think Apple will go down that route - breaking compatibility with all x86 apps isn't going to be a popular choice. I just thought it'd be interesting to see how near Apple's ARM chips are to possibly ever powering something more than an iOS device. There's still rather a large gap between their chip and the chips used in their low end machines.

Thank you for that data.

To be fair (for the ARM processors), it would be nice to have an accompanying plot that compared the power consumed by each device in achieving their respective performance scores.

(although isolating the power measurement to the processor alone does not seem like something that would be easily accomplished)

 

[0007776]

I don't think they will, but I'm sure if they did 90% of consumers wouldn't know the difference or care. Even I wouldn't care, as long as it could match the performance of the i5 and run common programs.

They would know the difference once they bought it and all their old programs won't run unless the developer updates them to run on ARM which if it is only one computer they won't have a huge incentive to do.

 

[filmbuff]

They would know the difference once they bought it and all their old programs won't run unless the developer updates them to run on ARM which if it is only one computer they won't have a huge incentive to do.

It could run "optimized" versions of ipad apps though, which already includes a lot of desktop software.

 

[0007776]

It could run "optimized" versions of ipad apps though, which already includes a lot of desktop software.

Then what would be the advantage over an iPad with a bluetooth keyboard? All I would see is a loss of a touch screen, and a built in keyboard neither of which are great things if you are working with iPad apps.

 

[mattferg]

I don't think they will, but I'm sure if they did 90% of consumers wouldn't know the difference or care. Even I wouldn't care, as long as it could match the performance of the i5 and run common programs.

I think we're at least 3 generations away from an Arm powered laptop from Apple though.

Which is exactly what the A8 will never be able to do.

----------

Interestingly, the iPad Air benchmarks faster than the MacBook Air Apple was selling only a few years ago in late 2010

I think what's more interesting is that the iPad Air barely beats the original 2008 MacBook Air, and is beaten by the 2009. 6 and 5 year old models respectively.

 

[scaredpoet]

Thank you for that data.

To be fair (for the ARM processors), it would be nice to have an accompanying plot that compared the power consumed by each device in achieving their respective performance scores.

Even if the performance per watt matched the x86_64 chips, there's still an enormous expense required in recompiling the OS for ARM, and waiting for vendors to optimize their code. It took years for even Microsoft and Adobe to end their dependence on rosetta, and it's only within the last three years or so they've truly done it. Now this would be forcing a move again. And for what? The gains would have to be enormous to justify it.

(And for the record, the gains for moving out of PowerPC to x86 were enormous, among them being able to dual boot windows, and the fact that Apple could finally make laptops with CPUs more powerful than the G4 that wouldn't burn your house down with the heat generated as a result. PowerPC's performance curve hit a wall. Intel would have to do the same, pretty hard, to force a switch to anything else.)

 

[mattferg]

Even if the performance per watt matched the x86_64 chips, there's still an enormous expense required in recompiling the OS for ARM, and waiting for vendors to optimize their code. It took years for even Microsoft and Adobe to end their dependence on rosetta, and it's only within the last three years or so they've truly done it. Now this would be forcing a move again. And for what? The gains would have to be enormous to justify it.

(And for the record, the gains for moving out of PowerPC to x86 were enormous, among them being able to dual boot windows, and the fact that Apple could finally make laptops with CPUs more powerful than the G4 that wouldn't burn your house down with the heat generated as a result. PowerPC's performance curve hit a wall. Intel would have to do the same, pretty hard, to force a switch to anything else.)

I completely agree - at the moment I see it as this:

Benefits of switching to ARM:

1. Power consumption (and this is going down day by day more and more as Intel makes gains)
2. Not paying Intel a fee

Drawbacks:

1. Loss of ALL app compatibility
2. Massive loss of power
3. No bootcamp

As far as I see it, the only benefit once Broadwell is here is not having to pay Intel, and I doubt strongly that's a big enough reason for Apple to switch, considering all 3 of the drawbacks will ALWAYS be there.

 

[kwijbo]

I quickly plotted some Geekbench (not the be-all and end-all, but a good comparative benchmark to use) scores of iOS vs Mac devices. I took the most powerful iOS device from that year, and the low-end Mac from that year.

Image

Interestingly, the iPad Air benchmarks faster than the MacBook Air Apple was selling only a few years ago in late 2010. Possibly with more room inside a laptop, a A8 or A9 chip could be used in the future for a lower-end notebook.

This doesn't mean that I think Apple will go down that route - breaking compatibility with all x86 apps isn't going to be a popular choice. I just thought it'd be interesting to see how near Apple's ARM chips are to possibly ever powering something more than an iOS device. There's still rather a large gap between their chip and the chips used in their low end machines.

You use the multi core score for the MBA and single core score for the iPad. Update the data and that chart is alot more interesting

 

[cambookpro]

You use the multi core score for the MBA and single core score for the iPad. Update the data and that chart is alot more interesting

Whoops, so I did... hang on, I'll update it later on.

I thought the iPad Air seemed faster than the chart showed

 

[kwijbo]

Whoops, so I did... hang on, I'll update it later on.

I thought the iPad Air seemed faster than the chart showed

No worries, simple mistake. I only noticed because I plotted out the exact same data the other day.

Very interesting to see that 1) only the highest spec 2010 MBA is equal to the A7 and 2) the slope of improvement for the ARMs is much higher than the Intels. If the A8 is quadcore its going to be very close to (if not higher than) the 2012-2013 MBA scores.

Considering the narrowing performance gap and the massive cost difference between the two, an ARM based notebook is coming soon IMO.

 

[cambookpro]

Which is exactly what the A8 will never be able to do.

----------



I think what's more interesting is that the iPad Air barely beats the original 2008 MacBook Air, and is beaten by the 2009. 6 and 5 year old models respectively.

You use the multi core score for the MBA and single core score for the iPad. Update the data and that chart is alot more interesting

Thank you for that data.

To be fair (for the ARM processors), it would be nice to have an accompanying plot that compared the power consumed by each device in achieving their respective performance scores.

(although isolating the power measurement to the processor alone does not seem like something that would be easily accomplished)

I've updated the graph with the proper data (thanks, kwijbo ), and it shows that the power of the ARM chips, although still have a long way to go, are increasing at quite a pace.

 

[MartinAppleGuy]

First off, let's say the A8 processor is double the CPU performance. That means it should get a multi-core performance score of around 5000 on Geekbench. That is extremely close to what the Macbook Air is right now, but would get much better battery life. Apple could also increase the clockspeed of the A8 for the Macbook Air seeing as it would have a larger heat spread (and then an even higher clock speed for the 13" as it has an even larger battery than the 11", as well as a larger surface area). I don't see performance being a problem, I see compatibility being the real issue. Would software have to be rewritten to run on ARM based SoC's?

 

[\-V-/]

If Apple starts using ARM chips in their computers, I will probably stop using Apple products.

 

[MartinAppleGuy]

Which is exactly what the A8 will never be able to do.

----------



I think what's more interesting is that the iPad Air barely beats the original 2008 MacBook Air, and is beaten by the 2009. 6 and 5 year old models respectively.

First off, let's say the A8 processor is double the CPU performance. That means it should get a multi-core performance score of around 5000 on Geekbench. That is extremely close to what the Macbook Air is right now, but would get much better battery life. Apple could also increase the clockspeed of the A8 for the Macbook Air seeing as it would have a larger heat spread (and then an even higher clock speed for the 13" as it has an even larger battery than the 11", as well as a larger surface area).

 

[jdechko]

I completely agree - at the moment I see it as this:

Benefits of switching to ARM:

1. Power consumption (and this is going down day by day more and more as Intel makes gains)
2. Not paying Intel a fee

Drawbacks:

1. Loss of ALL app compatibility
2. Massive loss of power
3. No bootcamp

As far as I see it, the only benefit once Broadwell is here is not having to pay Intel, and I doubt strongly that's a big enough reason for Apple to switch, considering all 3 of the drawbacks will ALWAYS be there.

Agreed. The big question is "Is Intel getting better at making more efficient chips faster than ARM is getting better at making more powerful chips?" Right now, I think it's a pretty close race.

However, Intel is definitely better at shrinking the entire process. As far as I know, they're a whole generation ahead of the rest of the industry. Broadwell is going to be a 14nm process, while everyone else is just ramping up their 22nm processes.

Of course, if Apple went to ARM, they would control their entire technology stack, so there's that. But I think the drawbacks are too great for them to make this jump.

The other question is the industry as a whole. It's apparent, whether we want to admit it or not, that we are going mobile. Is this move practical for Apple, or are we eventually moving to ARM anyway as people replace their old computers with tablets?

 

[scaredpoet]

As far as I see it, the only benefit once Broadwell is here is not having to pay Intel, and I doubt strongly that's a big enough reason for Apple to switch, considering all 3 of the drawbacks will ALWAYS be there.

Just curious: Is there a particular reason why Apple shouldn't want to pay Intel?

Bear in mind that while Apple is designing their own ARM chips now, they still pay a license fee to ARM Holdings, which holds the rights to the ARM architecture and related intellectual property. They must then pay an external chip foundry (which is currently Samsung but may be both Samsung and TSMC later) to fabricate the chips.

If anyone has philosophical objections to paying Intel for silicon, I would imagine there's probably a bunch of other people who might not be too happy that Samsung's getting paid, either. And both are probably way outnumbered by people who don't care who gets paid, so long as their gadgets work as expected and the pricing isn't any higher.

So, it boils down to which is less costly: designing the chips in-house but cutting checks to at least two entities for IP and fabrication, or cutting a check to a single entity to do all the above: worry about the IP, design the chips and fab the silicon for you. It might even be true that the overall costs all come out about the same in the end. Though there is one thing we do know: an "Architectural" ARM license, which is the type Apple must have in order to design their own ARM chips, costs more than a "core" license, which is basically just using pre-existing ARM core designs.

One other thing about the Intel transition back in 2006: Rosetta offered a way for existing PowerPC apps to work under the Intel platform. Apparently, there was enough muscle on the Intel chips at the time to make the emulation/translation of PowerPC code on Intel bearable enough, that people didn't complain much.

I guess a similar emulation layer might be possible in an Intel to ARM switch, but I really doubt that current ARM chips have anything close to the performance required to make that same emulation tolerable without apps running VERY slowly.

 

[robvas]

I don't see performance being a problem, I see compatibility being the real issue. Would software have to be rewritten to run on ARM based SoC's?

Yes and no. You could flip a switch in Xcode and target ARM or even make a fat binary that runs on either ARM or Intel.

But if you don't have the code for something, let's say a 3rd party library or something you don't have the source code for anymore, you'll have to re-write it.

 

[kwijbo]

Image

Nice work on the updated chart

Just curious: Is there a particular reason why Apple shouldn't want to pay Intel?

Massive savings per device would be a motivator. An iPad Air BOM puts the price of the A7 at $18 compared to MSRPs of $342 and $454 for the i5 and i7 respectively. Apple probably pays less than MSRP for bulk purchasing but its nowhere near $18. Dropping the BOM cost of an Air $200+ would be huge. Or, if this were a new separate model, they could lower cost of entry to $699/$799 and create a new computing category for the company.

Other motivators could be vertical integration, tighter control over the HW/SW integration and the ability to leverage advantages of their own R&D that no one else would have access to.

--------------------------------

And finally, Anandtech has posted a detailed writeup on the A7 and its capabilities:

As I mentioned before, it's bigger than anything else that goes in a phone. Apple didn't build a Krait/Silvermont competitor, it built something much closer to Intel's big cores. At the launch of the iPhone 5s, Apple referred to the A7 as being "desktop class" - it turns out that wasn't an exaggeration.

http://anandtech.com/show/7910/apples-cyclone-microarchitecture-detailed

Its coming soon

 

[jamesjingyi]

I would like to have a different category of chip, i.e G8/P8 which is a chip designed by Apple, but is designed for computers and isn't tiny like the A series... Of course, everything would have to work on it otherwise I don't want it

 

[Cheffy Dave]

if its desktop class, why not?