Apple A11 is 6 core (2+4)

[Eug]

New W2 chip for Apple Watch Series 3 (which includes cellular).

New iPhone SoC is A11 Bionic 6-core with first Apple-designed 3-core GPU. So not 2018 after all. Also Apple designed image signal processor.

Two high performance cores, but only 25% faster than A10 high performance cores.
However, there are four high efficiency cores which are 70% faster than A10 high efficiency cores.

 

[deathBOB]

25% better performance on 2 high power cores.
70% better performance on 4 low power cores.

30% faster than A10.
4.3b transistors, A10 was 3.3 per wikipedia.

 

[raghu78]

25% better performance on 2 high power cores.
70% better performance on 4 low power cores.

30% faster than A10.
4.3b transistors, A10 was 3.3 per wikipedia.

Very very impressive. High power core performance improvement is really good. Multi thread perf is mind blowing. A11 is 80-100% faster in single thread and 45-55% faster in multi thread performance than Exynos 8895 and SD 835 respectively. A11 just crushes the competition.

 

[pjmssn]

Can the geekbench scores of desktop CPUs and phone CPUs be compared? I am really curious about whether the A11 is comparable to a modern x86 CPU in terms of raw power.
My Xeon E5-1650 V2 gets ~3400 ST and 16,500 MT with Geekbench 4.1.1.

 

[Arachnotronic]

A11 Bionic is ridiculous. Love it!

 

[raghu78]

Can the geekbench scores of desktop CPUs and phone CPUs be compared? I am really curious about whether the A11 is comparable to a modern x86 CPU in terms of raw power. My Xeon E5-1650 V2 gets ~3400 ST and 16,500 MT with Geekbench 4.1.1.

Definitely comparable to a Kabylake 15w CPU in terms of integer performance . Floating point performance would be better on Intel CPUs due to 256 bit AVX units.

https://browser.geekbench.com/v4/cpu/3971310

 

[Ajay]

A11 Bionic is ridiculous. Love it!

Now we know were some of Apple's cash goes. The bought PA Semi and dialed it up to 11. And they keep dialing it up year after year!

This sort of gains posted by @deathBOB really makes me question the long-term survivability of x86 for Mac Laptops. Total platform control seems to be so valuable that it must be on the minds of Apple Execs.

 

[raghu78]

Now we know were some of Apple's cash goes. The bought PA Semi and dialed it up to 11. And they keep dialing it up year after year!

This sort of gains posted by @deathBOB really makes me question the long-term survivability of x86 for Mac Laptops. Total platform control seems to be so valuable that it must be on the minds of Apple Execs.

imo Apple moving to a single ISA (ARM) is not a matter of if it will happen but a matter of when it will happen.

 

[deathBOB]

This sort of gains posted by @deathBOB really makes me question the long-term survivability of x86 for Mac Laptops. Total platform control seems to be so valuable that it must be on the minds of Apple Execs.

I bet the real question is not whether the chip itself is good enough (its probably been for a while if you gave it a higher performance envelope), but if its worth the transition and the cost of making another chip for the Mac when the Mac business is so small compared to iOS sales. They sold less than 20 million macs last year vs. 200 million iphones

Plus they would need to support a range of systems from ultrabooks to professional systems or continue to use Intel parts for the high end and support multiple sets of hardware.

 

[Eug]

The beauty of x86 in 2017 is that there are still two functional competing manufacturers to keep each other on their toes. This was not true for PowerPC at the high end. It was IBM and nobody else. Freescale didn't count. Sure AMD wasn't really a contender for a while, but at least at the higher performance end they have reasonable products now in 2017. Apple can push Intel from the bottom up, and AMD can push Intel from the top down, to keep Intel corporately humble.

In the meantime, Apple will simple expand the functionality of the iPad Pro lines and see where that goes.

You can be damn well sure Apple is compiling every single version of macOS on ARM though, and likely iOS on x86, to ensure the two OSes aren't straying too far from cross-platform compatibility. Who knows, I wonder if they still throw PowerPC into the mix.

 

[StinkyPinky]

Man, say what you like about Apple but their cpu's are beasts.

Kinda disappointed they didn't redesign the iphone 8 though. Bezels are bigger than texas.

 

[Eug]

I couldn't find any performance graphs this time around.

---

"Introducing A11 Bionic. With four efficiency cores that are up to 70 percent faster than A10 Fusion. And two performance cores that are up to 25 percent faster.

A second‑generation performance controller provides more power when you need it. While delivering the same great battery life.

The new Apple‑designed three‑core GPU is up to 30 percent faster than A10 Fusion."

---

"Introducing A11 Bionic. The most powerful and smartest chip ever in a smartphone, with a neural engine that’s capable of up to 600 billion operations per second."

---

Pricing on the iPhone X is ridiculous though. For 64 GB it's US$999 / CAD$1319 and for 256 GB it's US$1149 / CAD$1529. The Canadian pricing works out to US$1082 and $1254 respectively. They wouldn't even throw us a bone and price them at CAD$1299 and CAD$1499. All the more reason to skip this generation. See you next year!

 

[Ajay]

I bet the real question is not whether the chip itself is good enough (its probably been for a while if you gave it a higher performance envelope), but if its worth the transition and the cost of making another chip for the Mac when the Mac business is so small compared to iOS sales. They sold less than 20 million macs last year vs. 200 million iphones

Plus they would need to support a range of systems from ultrabooks to professional systems or continue to use Intel parts for the high end and support multiple sets of hardware.

If the ROI isn't there, then no, it's not going to happen. I think the only added complexity for the laptop line would be some equivalent to DMI on board to provide the full range of IO support. Otherwise, extending platform support to MacOS notebooks doesn't seem on the face of it to be that difficult*. The real question, I suppose, is would software vendors put up with another CPU change and the need to recompile all their code to create compatible binaries.

* There are a wide variety of unknowns as simple as the number of address bit and handling of virtual address spaces, the lack of AVX, etc. I don't know how closely Mac hardware is tied to the x86 platform.

 

[defferoo]

If the ROI isn't there, then no, it's not going to happen. I think the only added complexity for the laptop line would be some equivalent to DMI on board to provide the full range of IO support. Otherwise, extending platform support to MacOS notebooks doesn't seem on the face of it to be that difficult*. The real question, I suppose, is would software vendors put up with another CPU change and the need to recompile all their code to create compatible binaries.

* There are a wide variety of unknowns as simple as the number of address bit and handling of virtual address spaces, the lack of AVX, etc. I don't know how closely Mac hardware is tied to the x86 platform.

luckily Apple has been pushing compiling apps as universal binaries. They would just need to update Xcode to support a new target and have developers recompile their applications. high performance apps will need more tuning and customization, but a typical app might not require much work to get working on an ARM build of macOS.

If anything, we'll see this happen in the MacBook first, as the low-TDP CPUs are a closer match to the A-series in terms of capabilities (and A-series already has USB 3 support, and basically already has a SSD controller). Supporting other models would be a more involved endeavor.

 

[Ajay]

I didn't mean to take this thread way off topic - sorry.

 

[witeken]

Combination of new architecture (the totally reworked cache structure seems to indicate a big overhaul) and 10nm (even if it's not a huge performance jump) should allow Apple to do significantly better than the A10X in single-core perf.

This looks low -- only a 16% improvement in ST perf? But interesting that they changed up the cache structure; L1 cache sizes cut in half, but L2 much larger.

Nah. I would be shocked if they couldn't wring out more performance by increasing clocks + IPC. Let's wait for more reliable results.

Apple has hit the same wall as Intel . Welcome back in reality, where the physics is the same for every company.

 

[witeken]

This sort of gains posted by @deathBOB really makes me question the long-term survivability of x86 for Mac Laptops. Total platform control seems to be so valuable that it must be on the minds of Apple Execs.

Because Intel is not sitting still either.

 

[Ajay]

Apple has hit the same wall as Intel . Welcome back in reality, where the physics is the same for every company.

We'll see in reviews, but 16-25% boosts well ahead of Intel.

Because Intel is not sitting still either.

No, they are already at the flatter end of the curve (barring some breakthrough).

 

[scannall]

Because Intel is not sitting still either.

Core architecture is wrung out about as far as it will go. They've had nothing but little edges to pick at for a very long time now.

 

[LightningZ71]

Apple has been working both openly and behind the scenes to merge the iOS and MacOS platforms. It would not shock me in the least if they pulled a WindowsXP and merged their high end line with their mobile line to have one iOS with MacOS emulation to unite them all. And, with that move, they can merge the platforms into one CUP architecture based on their A series CPUs with maybe Nvidia dGPU units for the high end.

 

[Hans de Vries]

http://browser.geekbench.com/v4/cpu/3976700

ST: 4274
MT: 10248

 

[thunng8]

A new result just recently appeared on the geekbench database:

http://browser.geekbench.com/v4/cpu/3976700

Single: 4,274
Multi: 10,248

Impressive - this is faster than the recently released ipad pro with a10X processor

Approx same as the 25W Kaby Lake i5 in 2017 13" macbook pro in single thread and beats it in multithread

https://browser.geekbench.com/macs/414

Edit: Hans just beat me to it

 

[beginner99]

imo Apple moving to a single ISA (ARM) is not a matter of if it will happen but a matter of when it will happen.

And ti won't happen because it's not worth the hassle. Mac sales are like what? 1/10 of iPhone sales? It's a tiny market in unit and sales/income wise for Apple so ROI to change it all over is not really worth it.

Then the geekbench scores. Can we stop using some in real-life irrelevant artificial benches to compare CPUs on different platforms? It's close to meaningless. Are Apple SOCs great SOCs? Yes. Are they the best? Yes. But if you really think they can compete with a 25W kaby lake, don't be naive. Especially in sustained performance.

 

[darkswordsman17]

Apple has been working both openly and behind the scenes to merge the iOS and MacOS platforms. It would not shock me in the least if they pulled a WindowsXP and merged their high end line with their mobile line to have one iOS with MacOS emulation to unite them all. And, with that move, they can merge the platforms into one CUP architecture based on their A series CPUs with maybe Nvidia dGPU units for the high end.

Somewhat, although I'm not sure how deep that goes, and I think most of it is about unifying their software development so that they cut down on redundancy. I do think that unifying their platforms is going to happen, I just don't think they're in a rush to get there.

They're developing their own GPU architecture as well. I have a hunch that a licensing deal is the reason they've stuck with AMD despite Nvidia having much more compelling products. I also think that AMD and Apple's future align a fair amount (APUs/heterogeneous compute, and the multi-chip setup for high performance applications - I could see Apple moving the pro line to basically be add in cards that are just some CPUs with large GPUs and memory, and so if you want more computing power, you just add more cards).

And ti won't happen because it's not worth the hassle. Mac sales are like what? 1/10 of iPhone sales? It's a tiny market in unit and sales/income wise for Apple so ROI to change it all over is not really worth it.

Then the geekbench scores. Can we stop using some in real-life irrelevant artificial benches to compare CPUs on different platforms? It's close to meaningless. Are Apple SOCs great SOCs? Yes. Are they the best? Yes. But if you really think they can compete with a 25W kaby lake, don't be naive. Especially in sustained performance.

Actually I think your reasoning is exactly why it makes sense for them to overhaul (or deprecate) MacOS and effectively unify their platforms. They're wasting resources by developing things for two platforms. With iOS they've got a lean OS that has a lot of modern features integral to its operation, and they're getting hardware that is good enough that they are running out of reasons to keep things separate. And with how things are changing, having software be adaptable to input is going to be a requirement, so even the argument about mouse/keyboard versus touch, etc is dying. And that's going to keep changing as they add more wearable stuff and AR (Apple is developing an AR headset/glasses).

I just think Apple isn't in a rush. They see the reactions to things Microsoft has done, and I think they just are letting things happen gradually. This way they don't upset any of their current product lines. Apple no longer needs to set the world on fire, but they need steady revenue, which gradual development offers.

 

[SunnyNW]

One thing that caught my eye during the Apple event was the improvement mentioned regarding the efficiency cores. Going from A10 to A11 Apple mentioned a 70% improvement. When talking about the watch series 3 Apple also mentioned a 70% improvement to the core compared to the previous watch chipset. That got me wondering if they are the same core or at the very least have the same "core" foundation. I could be wrong but could it be that after Apple developed the first watch core that they then used it for the basis of the efficiency cores that are now used in the A-series?