Apple A9X the new mobile SoC king

[raghu78]

So basically A9 is a Core M inside of a phone, per Geekbench 3.

Sweet.

A9 at 1.8 Ghz has a single thread score of 2293. Multi thread score of 4293. Scaling of 87% with 2nd core.

http://wccftech.com/iphone-6s-galaxy-note-5-benchmark-scores-comparison-handset-takes-crown/

Skylake Core M 6Y75 with a 2.4 Ghz boost has a single thread score of 2711. Multi thread score of 5706 with 4 threads. Scaling of 110% due to SMT.

http://browser.primatelabs.com/geekbench3/2637682

A9X with a 2.0 Ghz clock single thread score should be around 2500 - 2600. This points to Apple's A9X IPC actually being even better than Skylake. The difference is A9X is more efficient as it has 3 true cores in the same TDP as Core M which has 2 cores with SMT. A9X multithread score should be 7000+ while 6Y75 is at 5706.

 

[ShintaiDK]

I think we already covered multiple times why Geekbench is a terrible benchmark that cripples x86.

 

[Arachnotronic]

A9 at 1.8 Ghz has a single thread score of 2293. Multi thread score of 4293. Scaling of 87% with 2nd core.

http://wccftech.com/iphone-6s-galaxy-note-5-benchmark-scores-comparison-handset-takes-crown/

Skylake Core M 6Y75 with a 2.4 Ghz boost has a single thread score of 2711. Multi thread score of 5706 with 4 threads. Scaling of 110% due to SMT.

http://browser.primatelabs.com/geekbench3/2637682

A9X with a 2.0 Ghz clock single thread score should be around 2500 - 2600. This points to Apple's A9X IPC actually being even better than Skylake. The difference is A9X is more efficient as it has 3 true cores in the same TDP as Core M which has 2 cores with SMT. A9X multithread score should be 7000+ while 6Y75 is at 5706.

The A9 is an impressive chip. Apple's chip team is strong and, IMHO, only getting stronger as the company is able to attract the industry's top talent.

Re: IPC. Per GB3 it looks like A9 CPU has higher perf/mhz than Skylake, but remember that Skylake scales to much higher frequencies. All I care about is total delivered performance, and in this respect, the A9 is a very impressive design. This kind of performance in your pocket is just straight up insane.

And, FYI, 6Y75 turbo boosts to 3.1GHz. http://ark.intel.com/products/88199/Intel-Core-m7-6Y75-Processor-4M-Cache-up-to-3_10-GHz

 

[asendra]

It still is a good way to measure advancements in a given platform. And however you look at it, It's a pretty impressive jump in performance, without even needing to add more cores.

A lot of people were saying last year when the iPhone only got a modest bump that we would not be seeing more big jumps in performance. Sure, it's also a node change, but IPC improvements are quite good.

If last year A8x was already impressive and started real discussions about Apple replacing Intel in some of their laptops because of how close performance was getting, this year A9x will bring them much more close to that possibility (If they ever decide to make the change).

 

[tempestglen]

anandtech will provide us the SPEC2000 score of A9/A9X. Be patient.

 

[raghu78]

I think we already covered multiple times why Geekbench is a terrible benchmark that cripples x86.

yeah anything which does not show your lordship Intel winning must be a terrible benchmark. :biggrin:

 

[raghu78]

The A9 is an impressive chip. Apple's chip team is strong and, IMHO, only getting stronger as the company is able to attract the industry's top talent.

Re: IPC. Per GB3 it looks like A9 CPU has higher perf/mhz than Skylake, but remember that Skylake scales to much higher frequencies. All I care about is total delivered performance, and in this respect, the A9 is a very impressive design. This kind of performance in your pocket is just straight up insane.

And, FYI, 6Y75 turbo boosts to 3.1GHz. http://ark.intel.com/products/88199/Intel-Core-m7-6Y75-Processor-4M-Cache-up-to-3_10-GHz

I am very impressed with what Apple has achieved with their custom ARMv8 core. Apple is actually going to be pushing Intel the most to improve the IPC of their cores. Apple is yet to implement SMT in their custom ARMv8 core. That would give another 15-20% improvement in multithread perf. Given the fact that their custom ARMv8 core is massive and very wide its a ripe candidate for SMT.

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

Right now Apple's custom ARMv8 core is without a doubt the best mobile CPU core irrespective of ISA. Given the fact that A9X at 2.0 Ghz will score around 2500-2600 and ARM A57 at 2.1 Ghz scores 1500 its a monster in single thread perf. In fact Apple can surely design server CPUs using their custom ARMv8 core. They have the chip design expertise and can design high performance coherent fabric to connect tens of such powerful cores along with cache and system I/O.

Apple is going to hurt Intel the most going forward. The way these guys have been improving performance on the Ax SoCs is frightening. Intel's tick tock model is history and Apple's tock tock model is the future. Apple has been adding 10% IPC every year to the Cyclone core from A7->A8->A9. The move to FINFET in A9 gave a huge clock speed jump too. Apple iPad Pro with A9X is going to take sales away from Wintel 2-in-1s. Going forward I expect Apple to move all their products to ARMv8. I think Apple might design 2 versions of their custom ARMv8 core - a high performance version for workstations, desktops and high end notebooks which can hit 4+ Ghz frequencies and a low power version which is used in tablets, phone and ultrathin notebooks. The low power version will use high density libraries and the high performance version will use high performance libraries and be targetted at maximum performance. Its not a question of will they but a matter of when ? I think sometime in 2016 or 2017 we could see this happen.

 

[ShintaiDK]

I see you are already in full "Fiji" mode again raghu78. But why learn from past mistakes

 

[Arachnotronic]

Right now Apple's custom ARMv8 core is without a doubt the best mobile CPU core irrespective of ISA.

:thumbsup:

 

[raghu78]

I see you are already in full "Fiji" mode again raghu78. But why learn from past mistakes

Apple is not AMD. I think your pride is taking a battering here with Apple's A9X kicking Skylake Core M. Don't worry we will see anandtech confirm the same with a range of benchmarks. I think we will see some industry standard benchmarks like SPEC CPU 2006 too from anandtech :biggrin:

Both Apple A9X and Skylake Core M chips are going after the same market - tablets and convertibles. Both of them are targetted at productivity. We will see who wins this war within a year.

 

[Arachnotronic]

Apple is going to hurt Intel the most going forward. The way these guys have been improving performance on the Ax SoCs is frightening. Intel's tick tock model is history and Apple's tock tock model is the future. Apple has been adding 10% IPC every year to the Cyclone core from A7->A8->A9. The move to FINFET in A9 gave a huge clock speed jump too. Apple iPad Pro with A9X is going to take sales away from Wintel 2-in-1s. Going forward I expect Apple to move all their products to ARMv8. I think Apple might design 2 versions of their custom ARMv8 core - a high performance version for workstations, desktops and high end notebooks which can hit 4+ Ghz frequencies and a low power version which is used in tablets, phone and ultrathin notebooks. The low power version will use high density libraries and the high performance version will use high performance libraries and be targetted at maximum performance. Its not a question of will they but a matter of when ? I think sometime in 2016 or 2017 we could see this happen.

Maybe, we'll see. I don't think you should underestimate Intel, though. They might not be so great at fielding competitive mobile cores/SoCs (frankly the execution here has been dismal), but they still have excellent CPU architects and the world's best manufacturing technology working on their "big" cores.

I also think the ROI is fairly limited for Apple to develop high performance CPU cores for the Mac line of products. These just aren't all that high volume markets and, frankly, there are a finite number of world-class CPU/SoC architects in the world. If I were running things at Apple, I would want these folks working to make the iPhone chips the best that they can be and not risk wasting their talent on reinventing the PC processor wheel.

We will see, though. If Apple can develop better PC processors than what Intel can, then as a consumer I would like to get the best hardware for my money and would welcome such a development.

 

[Leo9]

I see you are already in full "Fiji" mode again raghu78. But why learn from past mistakes

A8X scores 2400ms (less than Core i3Y) with iOS 9.1 beta. Useless benchmark.

 

[ShintaiDK]

Apple is not AMD. I think your pride is taking a battering here with Apple's A9X kicking Skylake Core M. Don't worry we will see anandtech confirm the same with a range of benchmarks. I think we will see some industry standard benchmarks like SPEC CPU 2006 too from anandtech :biggrin:

Both Apple A9X and Skylake Core M chips are going after the same market - tablets and convertibles. We will see who wins this war within a year.

What about all the benchmark that isnt Geekbench? 3Dmark physics, kraken, sunspider, webxprt. Hell, why not AnTuTu now we are at it since it doesnt seem to bother you?

 

[Idontcare]

It still is a good way to measure advancements in a given platform. And however you look at it, It's a pretty impressive jump in performance, without even needing to add more cores.

A lot of people were saying last year when the iPhone only got a modest bump that we would not be seeing more big jumps in performance. Sure, it's also a node change, but IPC improvements are quite good.

If last year A8x was already impressive and started real discussions about Apple replacing Intel in some of their laptops because of how close performance was getting, this year A9x will bring them much more close to that possibility (If they ever decide to make the change).

+1

And while I am no fan of any given benchmark, unless it is a close proxy for a specific app of interest to me, the bottom line is that if with the same benchmark Apple has increased benchmark performance this much while maintaining battery life and so forth, then isn't that rather remarkable and exciting in its own right?

I am both an Apple/Arm and an Intel/Windows user, and personally I don't care if the A9 is comparable to a Core M, it just isn't something that matters to me specifically. But it does matter to me how well the Core M performs relative to (in the same benchmark and battery life) its predecessors, just as much as it matters to me how well the A9 performs relative to the A8 in the same benchmarks (to the A8).

Should Apple actually come to be on par with Intel it would only impact me if the form factors actually overlapped. Can I get an ultrabook with an A9X? Or a 6S form-factor smartphone with a Core M?

Only at that crossover point do I truly care how one performs relative to the other in the Apple/Arm vs. Intel/x86 consumer space.*


*This is just me posting from a consumer/end-user perspective and not from a technologist/science perspective.

 

[raghu78]

Maybe, we'll see. I don't think you should underestimate Intel, though. They might not be so great at fielding competitive mobile cores/SoCs (frankly the execution here has been dismal), but they still have excellent CPU architects and the world's best manufacturing technology working on their "big" cores.

I also think the ROI is fairly limited for Apple to develop high performance CPU cores for the Mac line of products. These just aren't all that high volume markets and, frankly, there are a finite number of world-class CPU/SoC architects in the world. If I were Apple, I would want these guys working to make the iPhone chips the best that they can be and not risk wasting their talent on reinventing the PC processor wheel.

We will see, though. If Apple can develop better PC processors than what Intel can, then as a consumer I would like to get the best hardware for my money and would welcome such a development.

The Mac is growing 15% y-o-y while the PC market is shrinking and down 2 to 3%. Actually the more Mac gains against Windows the better for Apple. Apple can leverage their advantage of being able to provide the hardware and software together. Today Apple sells Intel CPUs with 4 cores and 8 threads in their Mac book Pro and iMac. Imagine if Apple could implement SMT in their custom ARMv8 core and provide 6 cores/12 threads in high end notebooks and iMac. For workstations they are stuck with Intel Xeons with 8 cores / 16 threads. Imagine if Apple could deliver a 16 core / 32 thread workstation CPU. The more important advantage is Apple's ability to tune the OS and their first party apps to make the best use of their hardware. In the past Apple did not design their own CPUs and thus they could not leverage their advantage. Now they do and it makes perfect sense to leverage it and beat Wintel across the computing spectrum.

 

[NTMBK]

+1

And while I am no fan of any given benchmark, unless it is a close proxy for a specific app of interest to me, the bottom line is that if with the same benchmark Apple has increased benchmark performance this much while maintaining battery life and so forth, then isn't that rather remarkable and exciting in its own right?

I am both an Apple/Arm and an Intel/Windows user, and personally I don't care if the A9 is comparable to a Core M, it just isn't something that matters to me specifically. But it does matter to me how well the Core M performs relative to (in the same benchmark and battery life) its predecessors, just as much as it matters to me how well the A9 performs relative to the A8 in the same benchmarks (to the A8).

Should Apple actually come to be on par with Intel it would only impact me if the form factors actually overlapped. Can I get an ultrabook with an A9X? Or a 6S form-factor smartphone with a Core M?

Only at that crossover point do I truly care how one performs relative to the other in the Apple/Arm vs. Intel/x86 consumer space.*


*This is just me posting from a consumer/end-user perspective and not from a technologist/science perspective.

Apple's iPad Pro is directly targeting the same market as the Surface Pro, and all other Windows 10 convertibles. It is a large tablet with a keyboard and stylus, meant to act as a productivity device which can replace your laptop. Plenty of people will be cross-shopping between the iPad Pro and the Surface Pro, and as such comparing Haswell-Y and the A9X- and presumably Skylake-Y and the A9X, when Microsoft updates their tablet.

 

[scannall]

Apple is going to hurt Intel the most going forward. The way these guys have been improving performance on the Ax SoCs is frightening. Intel's tick tock model is history and Apple's tock tock model is the future. Apple has been adding 10% IPC every year to the Cyclone core from A7->A8->A9. The move to FINFET in A9 gave a huge clock speed jump too. Apple iPad Pro with A9X is going to take sales away from Wintel 2-in-1s. Going forward I expect Apple to move all their products to ARMv8. I think Apple might design 2 versions of their custom ARMv8 core - a high performance version for workstations, desktops and high end notebooks which can hit 4+ Ghz frequencies and a low power version which is used in tablets, phone and ultrathin notebooks. The low power version will use high density libraries and the high performance version will use high performance libraries and be targetted at maximum performance. Its not a question of will they but a matter of when ? I think sometime in 2016 or 2017 we could see this happen.

I agree with much of what you had to say, but this paragraph kinda fell off the wagon. The Mac volume isn't enough to justify the expense of another chip design team. The reason they can throw so much money at the mobile SOC's is the volume they sell.

If they were to go ARM in their computers, then they'd lose (Wild guesstimate here) 30 to 40% of their Mac sales, by losing Windows compatibility.

To keep that loss from happening, it'd have to be able to emulate x86 very well. Apple is probably better than that than most, as they have gone through several ISA changes since 1984. But, you have to have a huge jump in CPU power to do so. If they, for the sake of conversation managed to get twice the raw CPU power that Intel gets, then yeah. You could at that point make the jump.

As good as they are with their ARM cores, they aren't ever going to get twice the performance Intel gets. If nothing else though, they seem to be the only one who has the ability to push Intel, now that AMD is nothing more than a blip.

P.S. I do hope the AMD Zen does well. It's not healthy for the industry to have a sole source, proprietary product for most of the worlds PC's, servers etc.

This isn't a bash on Intel's ability, they have some great engineers. And if they were to enter the mobile market via ARM I'd cheer them on, as they could likely do very well.

 

[raghu78]

What about all the benchmark that isnt Geekbench? 3Dmark physics, kraken, sunspider, webxprt. Hell, why not AnTuTu now we are at it since it doesnt seem to bother you?

We will see A9X and Skylake Core M put through all the tests and much more. Eventually this is a Wintel vs Apple/ARM contest. I am in the side of ARM as I believe ARM will eventually win over x86. It will take time. The mobile market has been conquered by ARM. Desktops , notebooks and servers will happen over a longer period of time. Maybe a decade. But it will happen.

 

[Arachnotronic]

The Mac is growing 15% y-o-y while the PC market is shrinking and down 2 to 3%. Actually the more Mac gains against Windows the better for Apple. Apple can leverage their advantage of being able to provide the hardware and software together. Today Apple sells Intel CPUs with 4 cores and 8 threads in their Mac book Pro and iMac. Imagine if Apple could implement SMT in their custom ARMv8 core and provide 6 cores/12 threads in high end notebooks and iMac. For workstations they are stuck with Intel Xeons with 8 cores / 16 threads. Imagine if Apple could deliver a 16 core / 32 thread workstation CPU. The more important advantage is Apple's ability to tune the OS and their first party apps to make the best use of their hardware. In the past Apple did not design their own CPUs and thus they could not leverage their advantage. Now they do and it makes perfect sense to leverage it and beat Wintel across the computing spectrum.

I can totally understand the enthusiasm for this, but at the same time, it would seem to me that this would distract Apple from crafting the best iPhone/iPad chips.

But, who knows what the future holds. Five years ago, if you told me that Apple would be putting something with the characteristics of the iPhone 6s in my hand, I probably would have been skeptical.

So, you may ultimately be right. As a consumer, I just want the best performance and the most features at the lowest cost. If Apple can deliver me better value with custom A-series chips inside of my next MacBook than with Intel silicon, then bring it on.

 

[raghu78]

I agree with much of what you had to say, but this paragraph kinda fell off the wagon. The Mac volume isn't enough to justify the expense of another chip design team. The reason they can throw so much money at the mobile SOC's is the volume they sell.

I refer to my post.

http://forums.anandtech.com/showpost.php?p=37716761&postcount=215

the x86 Wintel PC market is getting cannibalized by the ARM ecosystem. With the A9X, Snapdragon 820 and other powerful SoCs we will see that trend accelerate. ARM will gain at the expense of x86.

If they were to go ARM in their computers, then they'd lose (Wild guesstimate here) 30 to 40% of their Mac sales, by losing Windows compatibility.

Firstly you don't have proof that Apple will lose sales. Secondly Microsoft is rumoured to be developing a native Windows for ARMv8. Microsoft has shown they are no more wedded to Intel. They see themselves as a software company and any opportunity to maximize their software sales will not be spared. Windows for ARMv8 will be the eventual confirmation that ARMv8 is the future.

P.S. I do hope the AMD Zen does well. It's not healthy for the industry to have a sole source, proprietary product for most of the worlds PC's, servers etc.

This isn't a bash on Intel's ability, they have some great engineers. And if they were to enter the mobile market via ARM I'd cheer them on, as they could likely do very well.

I was a AMD supporter. But the way things are going I am skeptical that they will even survive to launch Zen.

 

[Nothingness]

What about all the benchmark that isnt Geekbench? 3Dmark physics, kraken, sunspider, webxprt. Hell, why not AnTuTu now we are at it since it doesnt seem to bother you?

JS benchmarks are useless to compare CPUs. 3dmark physics is the only outlier. Any other benchmark?

 

[Arachnotronic]

JS benchmarks are useless to compare CPUs. 3dmark physics is the only outlier. Any other benchmark?

Yeah, the state of mobile benchmarking is pretty dire.

At least in the PC world, we can just measure the performance of real-world applications. In mobile, this is very difficult to do.

 

[Nothingness]

Yeah, the state of mobile benchmarking is pretty dire.

At least in the PC world, we can just measure the performance of real-world applications. In mobile, this is very difficult to do.

Agreed. OTOH what is funny is that mere mortals don't have SPEC 2k results for recent Intel chips, but have them for A8 chips :biggrin:

 

[Leo9]

New bench results (even better!)

 

[Mondozei]

Apple's iPad Pro is directly targeting the same market as the Surface Pro, and all other Windows 10 convertibles. It is a large tablet with a keyboard and stylus, meant to act as a productivity device which can replace your laptop. Plenty of people will be cross-shopping between the iPad Pro and the Surface Pro, and as such comparing Haswell-Y and the A9X- and presumably Skylake-Y and the A9X, when Microsoft updates their tablet.

Exactly. So this means that Core M Skylake and A9X will be compared directly because now we're seeing them in devices that compete head on. I can't wait to see the iPad Pro vs Surface Pro 4.