Apple A10 Geekbench 4 Score

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

HiroThreading

Member
Apr 25, 2016
173
29
91
If Apple's processors are as good you say they are, they would be absolutely disruptive in the server space. Multiple generations ahead of anything Intel or IBM can come up with in terms of perf/watt (and even assuming Apple's clocks stay low, comparable overall performance considering how low clocked those massive core count Xeons are compared to their desktop brethren. In actuality, a 24-core up-clocked A10 would likely eat a 24-core Xeon for lunch).

No, not once you factor in things like quad channel memory controllers with ECC support, 40+ PCIE links, quad QPI links for MP scaling, hardware schedulers to manage 24+ cores and something like a bi-directional ring bus to connect the cores, cache, memory controllers and QPI links. Not to mention support that mimics x86 instructions such as AVX and AES.

It's part of the reason why ARM servers are struggling so much to get off the ground. Once you factor in a lot of the requirements of server CPUs, most--if not all--of the power efficiency of ARM designs disappear.

Yes, Apple is damn good at what it does. But there's a reason why the only players in the server CPU market are Intel and IBM (and Sun/Oracle if you want to count SPARC). It's pretty easy to underestimate the two blue giants.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
If anything, there's been more rumors that Apple is getting rid of the Mac Pro. I guess we'll find out soon enough if that's true but it would make going to ARM only macs that much easier.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
Yes, Apple is damn good at what it does. But there's a reason why the only players in the server CPU market are Intel and IBM (and Sun/Oracle if you want to count SPARC). It's pretty easy to underestimate the two blue giants.
Yeah, Intel covers an astonishing TDP range with Core, while Apple has to cover just 1 product.

5%20-%20Scale%20Mobile.jpg


During IDF, Intel stated that having an efficiency window from 4.5W to 91W is a significant challenge, to which we agree, as well as improving both performance and power consumption over Broadwell at each stage.

http://www.anandtech.com/show/9582/intel-skylake-mobile-desktop-launch-architecture-analysis/2
 

ksec

Senior member
Mar 5, 2010
420
117
116
Seeking Alpha makes a good point: QCOM has been the exclusive modem supplier for iPhone for a loooong time, which has now changed. I think the rumors from earlier, that Apple wants to integrate modem on die, have come a little closer to reality.

Edit: I wonder why INTC stock has reacted poor to the news. I hope we hear some things about it in the earnings call.

Edit2: Why else would Apple introduce versions of their phone that do not work with some providers?

You make it sound like Qualcomm modem on iPhone as default. But it was on small batch in iPhone4, and exclusive started in 4S only. So it accounts for half of the iPhone's life so far. ( Although the later half is where the volume are making ).

I believe there is a lot of saving switching to Intel. Mainly from patents cost. Although there is no way we could verfiy. As far as I am concern, Intel is betting on 5G, and they have no where to go ( Mobile aren't working for them, Cloud and DC wont grow forever and they have POWER9 and AMD coming which might lower their margin. ) Getting a modem inside iPhone is literally mission critical.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
I have to wonder if Low Power Mode would run exclusively on the Little cores. There's a bunch more geekbench results now, and perhaps 1.05 Ghz is the max speed of it?
 

majord

Senior member
Jul 26, 2015
433
523
136
Yeah, Intel covers an astonishing TDP range with Core, while Apple has to cover just 1 product.

5%20-%20Scale%20Mobile.jpg




http://www.anandtech.com/show/9582/intel-skylake-mobile-desktop-launch-architecture-analysis/2


Well, actually, no it's not that greater challenge, and as Intel correctly stated (Yet Anandtech somehow incorrectly stated), it's a 4.5-45w window, which is nothing special at all (despite their marketing spin) , and even then, it encompasses a large range of IGP spec's eating into that TDP range.

In fact, the most effective way to create a low power SKU is to design a high performance (i.e High IPC, and high frequency) core and down-clock it.. Core M style. The Problem with this is size - You can't have it both ways, and just look at the A9 and A10 as examples - single die SoC, at 105mm2 on a lower density process. On top of this, there's the reality of the situation - Core m is completely incapable of operating in the chassis of an Ipad pro, letalone an Iphone.. So whilst you may want to class both A10 and Core M as the same TDP class, the reality is, they're not.

Let's not beat around the bush here - if early performance indications are even 70% of reality, Apple are on a serious role here when it comes to low power SoC's. Keep intel, AMD, out of it. And this is coming from someone who is far from an Apple fan. But credit where credit is due, they are owning this space.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
I have to wonder if Low Power Mode would run exclusively on the Little cores. There's a bunch more geekbench results now, and perhaps 1.05 Ghz is the max speed of it?
Apple says the LITTLE cores run at about 20% of the performance of the big cores. Would you think that it takes 1.05 GHz of LITTLE to equal 20% of 2.33 GHz of big? It should also be noted that some of the benches have the clock speed listed at 396 MHz.

By the way, at least one of the benches has the single-core speed topping 3500 in Geekbench 4, while still maintaining > 1.6X performance for multi-core speed. Impressive.

https://browser.primatelabs.com/v4/cpu/240915

FWIW, the Samsung Exynos 8890 (Galaxy S7) at 1.6 GHz gets just over 1800 single-core. Since there are four of the big cores, it gets just over 5200 for multi-core (2.9X). How big is that chip?
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
423CBE13-2DFE-4E2A-8AE6-258124F80D9E_zpsyxkvb9iq.png


IIRC Apple said the little draw 1/5 the power, not have 1/5 the performance.
Ok I stand corrected then. That makes more sense actually.

Still, I wonder if those clock speeds are just errors in the software.
 

stingerman

Member
Feb 8, 2005
100
11
76
Apple says the LITTLE cores run at about 20% of the performance of the big cores.
Though your post was corrected, performance will not be an issue with the high efficiency cores. All four cores will run the same ISA but the hard scheduler will send threads that are memory bound to the efficiency cores whereas processor bound threads will go to the performance cores. Memory bound threads will run nominally faster on the performance core compared to the energy they would have used. The efficiency cores will be be paced by the memory latency, so threads that are memory bound will run near as fast as they would have on the performance cores.

Geekbench Is probably currently blind to what core is processing its benchmarks. It may require a couple of runs for the built in thread scheduler to best schedule the threads as it profiles them in its lookup tables. So I imagine we might see 20-25% variances in performance results.

I doubt the earlier reports that Apple is turning off cores, That would be highly inefficient and costly. More likely, the performance cores voltage scales up and down whereas the efficiency cores are always at a low fixed voltage. A hardware scheduler and coherent shared caches, would minimize the times the performance cores need to scale to their highest allowable voltage.

Just my educated opinion based on some research I've done on assymetrical multi processing (AMP).

On a mobile phone, I can see why Apple would limit to two simultaneous threads. Otherwise, imagine all 4 threads running simultaneously and all of a sudden power usage and heat would violate allowed envelope. Not to mention competition for memory resources between the threads.

There's no reason to believe that Apple couldnt keep all 4 cores busy on a different form factor. In fact, why not have two A10s with all 8 cores active? It would be a major leap forward with AMP having great advantages over SMP processors today. (Hoping for this in the Macs future.)
 

dark zero

Platinum Member
Jun 2, 2015
2,655
138
106
There has got to be a middle ground between "Geekbench 4 is amazing and the best benchmark ever" and "it sucks."

I think the benchmark is the best that we have and it should give a pretty reasonable idea of where two processors stand relative to each other.

I would say A10 is certainly Skylake Core m class. I don't understand why some people would want to downplay this achievement by Apple.
I don't think so. Is on the Pentium m Class (based on Core m). A10X will be the one which will be Skylake Core m/U Class.
 

ksec

Senior member
Mar 5, 2010
420
117
116
With iPhone release schedule like clock work every year, we already know there will be a A11 on TSMC 10nm next year, and A12 on 7nm in 2018.
TSMC 7nm should be on par with Intel's 10nm performance. Making it the first time in history someone catches up Intel in CPU fab tech. Since Cannonlake will just be a 10nm die shrink of Kaby Lake, and we already know what Kabylake's performance are like, I think it is fair to say Apple could have the best CPU by 2018.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
With iPhone release schedule like clock work every year, we already know there will be a A11 on TSMC 10nm next year, and A12 on 7nm in 2018.
TSMC 7nm should be on par with Intel's 10nm performance. Making it the first time in history someone catches up Intel in CPU fab tech. Since Cannonlake will just be a 10nm die shrink of Kaby Lake, and we already know what Kabylake's performance are like, I think it is fair to say Apple could have the best CPU by 2018.
You seem awfully optimistic for TSMC's 7 nm timeline. Remember that lots of people were saying its 10 nm would be in 2016 in volume. Now we all know how that went.

stingerman said:
There's no reason to believe that Apple couldnt keep all 4 cores busy on a different form factor. In fact, why not have two A10s with all 8 cores active? It would be a major leap forward with AMP having great advantages over SMP processors today. (Hoping for this in the Macs future.)
For a different form factor one would think they'd have true traditional quad-core or else octo-core big.LITTLE.
 
Last edited:

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
You seem awfully optimistic for TSMC's 7 nm timeline. Remember that lots of people were saying its 10 nm would be in 2016 in volume. Now we all know how that went.
Great point, I totally forgot about that :D. When Intel delayed 10nm it came as a shock because the others were gunning full on for 2016. Now were almost in Q4'16 and still dead silence about anyone's 10nm :D. (For Intel, we should here a few things in 70 days, and we know there will be 10nm in one year, we just don't know how much products.)

Of course, I expected 2017 for the foundries all along. Well, I expected 2018 tbh, so they will probably limit the damage I had forseen, so good for them.

You've always got to remember that with the foundries, you have to keep track of what they're saying and compare that to what actually happens. So you have to look closely at what they're not saying. I never saw an article saying TSMC 10nm delayed. So is it going to plan and was the marketing just inventing dates, or is it delayed?
 
Mar 10, 2006
11,715
2,012
126
Great point, I totally forgot about that :D. When Intel delayed 10nm it came as a shock because the others were gunning full on for 2016. Now were almost in Q4'16 and still dead silence about anyone's 10nm :D. (For Intel, we should here a few things in 70 days, and we know there will be 10nm in one year, we just don't know how much products.)

Of course, I expected 2017 for the foundries all along. Well, I expected 2018 tbh, so they will probably limit the damage I had forseen, so good for them.

You've always got to remember that with the foundries, you have to keep track of what they're saying and compare that to what actually happens. So you have to look closely at what they're not saying. I never saw an article saying TSMC 10nm delayed. So is it going to plan and was the marketing just inventing dates, or is it delayed?

From TSMC's July call:

Mark Liu

Good afternoon, everyone. Thank you for joining us. I'd like to deliver you the message firstly on leading-edge technology status that includes 10-nanometer, 7-nanometer and 5-nanometer.

First, on 10-nanometer, our 10-nanometer has been transferred from R&D to production. Our first 10-nanometer customer product has been produced with satisfactory functional yield. So far, three customer products have been taped out to us. More customer product tape-outs are expected later this year. Those product tape-outs will start a revenue stream starting first quarter 2017, which will ramp steeply throughout 2017.

On 7-nanometer, our 7-nanometer technology development is well on track. Its 256-megabit SRAM yield improvement is ahead of our schedule. In addition, we believe our 7-nanometer PPA that is power, performance and area density, with its schedule is ahead of our competitors. This technology has been aggressively adopted, not only by mobile customers, but also by high-performance computing customers. They all have aggressive product tape-out plan in first half 2017, with volume production planned in early 2018.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
From TSMC's July call:
Alright. You know what, I will leave my usual skepticism about TSMC/Samsung behind for a while.

So products around the same time as Intel 10nm, maybe earlier 'cause they're talking about tape-out in Q2. Although it might still well be only AAPL in '17.

Now, normally, I wouldn't believe this 7nm 1 year after 10nm and better yields than expected BS, but you know what, besides your subjective Apple comments, you do give some interesting information from time to time. So it might be that here we see the first prove of the materialisation of TSMC's parallel process node R&D. It very well could be, after their horrible 28nm node (which, I remind people, since people tend for forget quickly, it took them 4 years to get a successor for 28nm out of the fabs: 28nm AMD GPU early '12, 14nm GPU in '16).

So if TSMC pulls this off, then instead of them horribly falling behind Intel, which now seems unlikely, they will keep status quo with Intel, 2 years or so behind. I did not expect that at all (doing so well at these nodes) considering how much Intel is struggling with these nodes.

Of course we don't yet know precise timings and feature(s) (sizes), and with every passing node this gets worse since Intel and TSMC employ completely opposite node philosophies.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106

punkyDD

Junior Member
Sep 8, 2016
13
7
51
Though your post was corrected, performance will not be an issue with the high efficiency cores. All four cores will run the same ISA but the hard scheduler will send threads that are memory bound to the efficiency cores whereas processor bound threads will go to the performance cores. Memory bound threads will run nominally faster on the performance core compared to the energy they would have used. The efficiency cores will be be paced by the memory latency, so threads that are memory bound will run near as fast as they would have on the performance cores.

Geekbench Is probably currently blind to what core is processing its benchmarks. It may require a couple of runs for the built in thread scheduler to best schedule the threads as it profiles them in its lookup tables. So I imagine we might see 20-25% variances in performance results.

I doubt the earlier reports that Apple is turning off cores, That would be highly inefficient and costly. More likely, the performance cores voltage scales up and down whereas the efficiency cores are always at a low fixed voltage. A hardware scheduler and coherent shared caches, would minimize the times the performance cores need to scale to their highest allowable voltage.

Just my educated opinion based on some research I've done on assymetrical multi processing (AMP).

On a mobile phone, I can see why Apple would limit to two simultaneous threads. Otherwise, imagine all 4 threads running simultaneously and all of a sudden power usage and heat would violate allowed envelope. Not to mention competition for memory resources between the threads.

There's no reason to believe that Apple couldnt keep all 4 cores busy on a different form factor. In fact, why not have two A10s with all 8 cores active? It would be a major leap forward with AMP having great advantages over SMP processors today. (Hoping for this in the Macs future.)

I was lucky enough to talk to an Apple engineer responsible for getting their kernel running with the A10. He emphasized that from a software perspective the A10 is a dual core processor and explicitly said that the 2 cores that are not in use are power gated, and that it's not possible to offload work to the little cores while the big ones are in use. Geekbench and other applications see two cores available, and the A10 decides when to switch to the high power cores on its own.

He also spent some time talking about how big.LITTLE has been a disaster because scheduling is left up to software, and poor programming has limited how much power it is able to save. Internally Apple has been experimenting with big.LITTLE and the results have been poor. He said the A10 is nothing like big.LITTLE in his opinion because everything is done in hardware.
 

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
TSMC's track record is great.
Come on. I know you're also an NVIDIA bull, so you know that's not true. Maybe they're improving but they'll have to prove that the hard way. So far, they have only proved for 1 node. 20nm was horribly late, it was a huge mess, so it's nothing to sneeze at that it was in time for A8. From 20nm to 20nm+FF isn't much of a leap. So yeah, from 14nm to 10nm in 2 years will look great, but compared to 20nm, it's also 3 years, which is nothing to sneeze at. Let's see if they can turn things around with 7nm, if they can do that within 3 years which they have not done since 28nm.

Now, compare to Tick-Tock. That's what I call a real track record. With hardly any mishaps.

He also spent some time talking about how big.LITTLE has been a disaster because scheduling is left up to software, and poor programming has limited how much power it is able to save. Internally Apple has been experimenting with big.LITTLE and the results have been poor. He said the A10 is nothing like big.LITTLE in his opinion because everything is done in hardware.
And there you go.
 
  • Haha
  • Like
Reactions: NTMBK and pcp7

ksec

Senior member
Mar 5, 2010
420
117
116
20nm mess? 20nm was a huge success, took all the market share since they were the only one doing it, and move all capacity to 16nm. ( Samsung 20nm were not available to outsiders )

It isn't track record per se. It is just they are very Open about it. Read what they say in every conference call, no hiding in numbers or schedule, or graphs that are not to scale. And they are working with Apple hand in hand to get it all things smoothed out.

So yes, 10nm next iPhone, and 7nm 2018 iPhone. 2019 will likely be a 7nm+, before (possibly) moving to 5nm in 2020. Which again is what TSMC publicly stated.

And Copycat node? Even TSMC themselves admit their 10nm isn't on par with Intel 10nm. There is a difference between some random guy like Fudzilla reporting and actual conference call from CEO which himself is a brilliant engineer. 7nm will be the node that should be compared to Intel 10nm. And even then there will always be trade off, cost/density, power etc.
 
  • Like
Reactions: Arachnotronic
Mar 10, 2006
11,715
2,012
126
20nm mess? 20nm was a huge success, took all the market share since they were the only one doing it, and move all capacity to 16nm. ( Samsung 20nm were not available to outsiders )

It isn't track record per se. It is just they are very Open about it. Read what they say in every conference call, no hiding in numbers or schedule, or graphs that are not to scale. And they are working with Apple hand in hand to get it all things smoothed out.

So yes, 10nm next iPhone, and 7nm 2018 iPhone. 2019 will likely be a 7nm+, before (possibly) moving to 5nm in 2020. Which again is what TSMC publicly stated.

And Copycat node? Even TSMC themselves admit their 10nm isn't on par with Intel 10nm. There is a difference between some random guy like Fudzilla reporting and actual conference call from CEO which himself is a brilliant engineer. 7nm will be the node that should be compared to Intel 10nm. And even then there will always be trade off, cost/density, power etc.

What a great post.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
Call me a skeptic, but I'd be utterly shocked if the iPhone 2018's A12 SoC turns out to be on 7nm.
 
Mar 10, 2006
11,715
2,012
126
Call me a skeptic, but I'd be utterly shocked if the iPhone 2018's A12 SoC turns out to be on 7nm.

It's TSMC 7nm. Why do you think TSMC is in such a hurry to get 7nm into production by 1Q 2018? ~20% of the company's biz is dependent on Apple's A-series processors (and likely much more indirectly via foundry work for other Apple suppliers), so these guys are going to move mountains to make sure they deliver for Apple.
 
Last edited: