Apple A10 Geekbench 4 Score

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,685
136
Not talking about time travel at all.
No, you just picked the very best competition had to offer - Apple CPU, Intel process, Nvidia GPU - and pretended lack of competition would not only produce the same results, but manage to unify all in a glorious formula. If only all humans were as smart as the top 1%, we'd also have faster computers :)
 

SAAA

Senior member
May 14, 2014
541
126
116
I'm not sure we can really compare cross platform like this in any meaningful way. What I found earlier versions of Geekbench most useful for was comparing chips within the same platform. eg. iOS vs iOS.

Is Geekbench really better as a cross platform benchmark? And even if it is better, is "better" good enough as a true cross platform test? Call me a skeptic.

This point still stands even after geekbench 4 release, I won't confide too much on it for cross architecture and os comparisons.
I mean it might represent a fair comparison between the same class of processors over generations, but assuming it's close to SPEC as some are saying when it gives such different results is definitely a mistake:
http://www.anandtech.com/show/9766/the-apple-ipad-pro-review/4

Look at the spec_int results in that page and compare core-m, Broadwell 5-Y31 and Skylake 6-Y30 to A9X. To summarize, if you assume they are running at similar clockspeeds (~2GHz) during the tests half the benches were A9X leads it's 10-20% (only one is a 38% advantage and over Broadwell) but the other half it loses badly, scoring like half or even less.

I quickly put the data in Excel and got Skylake 13% faster than Broadwell, while both have an advantage of 71 and 44% over A9X respectively. The higher end core-M is twice as fast but it's also running at a much higher speed, anyway it should give you an idea of what SPEC says while geekbench puts A9 IPC at higher than Skylake...

Edit: as side note, if A10 does indeed score more than the best A9X, Apple is getting close to core performance in a phone but assuming they are using the sam 16nm process that must be at a power and/or die size cost. They already optimized the architecture a lot so I don't expect huge gains in IPC or performance/power until they move to 10nm next year.
 
Last edited:

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Is Geekbench really better as a cross platform benchmark? And even if it is better, is "better" good enough as a true cross platform test? Call me a skeptic.

It is a horrible cross platform bench. Extremely bad. Note I do not want to get in an argument with people on how bad it is, or people who say it is the greatest thing ever.

It is a wonderful tool though comparing iOS to iOS though or windows to windows. Aka same to same where the goal is to compare this years hardware vs 2 year old hardware.
 

Nothingness

Platinum Member
Jul 3, 2013
2,371
713
136
This point still stands even after geekbench 4 release, I won't confide too much on it for cross architecture and os comparisons.
I mean it might represent a fair comparison between the same class of processors over generations, but assuming it's close to SPEC as some are saying when it gives such different results is definitely a mistake:
http://www.anandtech.com/show/9766/the-apple-ipad-pro-review/4

Look at the spec_int results in that page and compare core-m, Broadwell 5-Y31 and Skylake 6-Y30 to A9X. To summarize, if you assume they are running at similar clockspeeds (~2GHz) during the tests half the benches were A9X leads it's 10-20% (only one is a 38% advantage and over Broadwell) but the other half it loses badly, scoring like half or even less.

I quickly put the data in Excel and got Skylake 13% faster than Broadwell, while both have an advantage of 71 and 44% over A9X respectively. The higher end core-M is twice as fast but it's also running at a much higher speed, anyway it should give you an idea of what SPEC says while geekbench puts A9 IPC at higher than Skylake...
The Intel part is using icc which is heavily tuned for SPEC, which makes any comparison useless. On top of that icc was generating 32-bit code which has a dramatic impact on some of the benches (such as mcf).
 
  • Like
Reactions: Arachnotronic

Nothingness

Platinum Member
Jul 3, 2013
2,371
713
136
It is a horrible cross platform bench. Extremely bad. Note I do not want to get in an argument with people on how bad it is, or people who say it is the greatest thing ever.
So you claim it's horrible but don't want to discuss your claim. Interesting :)
 

Andrei.

Senior member
Jan 26, 2015
316
386
136
It is a horrible cross platform bench. Extremely bad. Note I do not want to get in an argument with people on how bad it is, or people who say it is the greatest thing ever.
So you claim it's horrible but don't want to discuss your claim. Interesting :)
I think it's as close to a proper cross-platform bench as we can possibly get. It's really not the GB3 mess anymore.
 

FIVR

Diamond Member
Jun 1, 2016
3,753
911
106
Geekbench 4 is the most precise, accurate and significant cross-platform benchmark available.


I don't want to get into an argument with anyone about why it is the best, or to what degree. That's for another thread.

A10 is clearly competing with Skylake and atm appears to be winning by a quantifiable amount in terms of IPC, but it probably can't clock nearly as high.
 
  • Like
Reactions: Nothingness

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
So you claim it's horrible but don't want to discuss your claim. Interesting :)
Most people who think Geekbench 3 or 4 is a great cross platform bench were not convinced in a rational argumentative way of all its merits. They are unfamiliar with the merits or the flaws of Geekbench. They just take it as a matter of faith that is a great cross comparative tool for it claims to be a great tool, or they emotionally want it to be a great tool for there is nothing really good on the market for cross platform comparisons.

Aka it is an emotional argument. Me spending time trying to argue against an emotional argument is an unproductive use of time time that is draining for me and really poses little benefit. It is not worth my time.

And being the best we currently have does not make it a good crossplatform bench.
 

teejee

Senior member
Jul 4, 2013
361
199
116
Most people who think Geekbench 3 or 4 is a great cross platform bench were not convinced in a rational argumentative way of all its merits. They are unfamiliar with the merits or the flaws of Geekbench. They just take it as a matter of faith that is a great cross comparative tool for it claims to be a great tool, or they emotionally want it to be a great tool for there is nothing really good on the market for cross platform comparisons.

Aka it is an emotional argument. Me spending time trying to argue against an emotional argument is an unproductive use of time time that is draining for me and really poses little benefit. It is not worth my time.

And being the best we currently have does not make it a good crossplatform bench.
Ok, but guys like Andrei and Linus have said GB4 seems good. Linus experience with performance analysis of Linux and at Transmeta is hard to beat. And Andrei has Android kernel hacking as a hobby. Why should I listen to your opinion when you don't even show any arguments or facts?

Sent from my LG-D855 using Tapatalk
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Why should I listen to your opinion when you don't even show any arguments or facts?
I honestly do not care to change your mind. Will it improve my day if I convinced you? The answer is no. It is not worth the argument.

And I made that comment flat out at the start I am not interested in arguing this.

Me needing to spend my time to convince you, and you feel that I am obligated to do this, that I have some form of kantian duty to do so, is the definition of narcissism. Note by saying this I am not calling you a bad person or anything like that, I am telling you this is a game that is not fun, and I do not want to play. Expecting me to play to make you HAPPY is narcissistic.

One last thing bring up appeals to authority is once again keeping the argument not rational but instead emotional. Appeals to authority depend on trust another form of intuition. If you want to make a compelling argument you need to start bring up things like white papers which show the flaws of the benchmark. Oh Linus was not saying Geekbench 4 was a good cross platform benchmark in the quotes I have seen, he instead was saying Geekbench 4 was a good amount better than Geekbench 3 but he is still not satisfied and he sees serious flaws that need to be addressed for he can't vouch for the results.

I am now dropping this.
 

FIVR

Diamond Member
Jun 1, 2016
3,753
911
106
I honestly do not care to change your mind. Will it improve my day if I convinced you?
If you do not enjoy persuasive argument, why are you here? Is this just some place for you to post rants? Is it some sort of blog for you?
The answer is no. It is not worth the argument.

And I made that comment flat out at the start I am not interested in arguing this.

Me needing to spend my time to convince you, and you feel that I am obligated to do this, that I have some form of kantian duty to do so, is the definition of narcissism. Note by saying this I am not calling you a bad person or anything like that, I am telling you this is a game that is not fun, and I do not want to play. Expecting me to play to make you HAPPY is narcissistic.

One last thing bring up appeals to authority is once again keeping the argument not rational but instead emotional. Appeals to authority depend on trust another form of intuition. If you want to make a compelling argument you need to start bring up things like white papers which show the flaws of the benchmark. Oh Linus was not saying Geekbench 4 was a good cross platform benchmark in the quotes I have seen, he instead was saying Geekbench 4 was a good amount better than Geekbench 3 but he is still not satisfied and he sees serious flaws that need to be addressed for he can't vouch for the results.

I am now dropping this.
Oh, there's the rant. You answered my question.
 
Mar 10, 2006
11,715
2,012
126
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.
 

Nothingness

Platinum Member
Jul 3, 2013
2,371
713
136
If you do not enjoy persuasive argument, why are you here? Is this just some place for you to post rants? Is it some sort of blog for you?

Oh, there's the rant. You answered my question.
Indeed. Discussing with people who made their mind and don't even explain the reasons of their claims is a waste of time of everyone.

I have seen the source code of GB3 and GB4 and I know how John Poole worked hard on making GB better. GB4 is way better than any comparison done with SPEC when icc is involved for instance. As always one should not look at a single benchmark, but GB4 is really good.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
I have seen the source code of GB3 and GB4 and I know how John Poole worked hard on making GB better. GB4 is way better than any comparison done with SPEC when icc is involved for instance. As always one should not look at a single benchmark, but GB4 is really good.
I guess you're indirectly saying that SPEC performance on Intel is overly reliant on ICC if I'm reading you right. That makes sense. Anyhow, it's nice to hear that GB4 has some supporters, and that it's a large improvement for cross-platform comparisons compared to GB3, especially if you are a developer that sticks to the same compiler on multiple platforms. GB4 may not be perfect, but improvements are always welcome.

Of course, all this discussion pre-supposes that the Geekbench results posted are actually legit. Right now I'm not convinced of that. Nonetheless, this is a Mac OS X Core M 5Y71 1.3 GHz MacBook for comparison:

http://browser.primatelabs.com/v4/cpu/156809

Single core 2550
Multi core 4118

OTOH, a Windows 10 Core M 5Y71 1.4 GHz laptop scores much higher:

http://browser.primatelabs.com/v4/cpu/182257

Single core 3356
Multi core 5768

BTW, regarding the 4118 score for the Mac, my iPad Air 2 already scores over 4000 for multi core, but that's with three cores. However, I bought that 2 years ago. It's holding up well it seems.
 

Nothingness

Platinum Member
Jul 3, 2013
2,371
713
136
I guess you're indirectly saying that SPEC performance on Intel is overly reliant on ICC if I'm reading you right.
Not exactly. Intel performance on SPEC is excellent even with gcc; Intel microarchitectures are very good. But icc has broken some of the components of SPEC; as an example in libquantum it restructures data all over the program (AoS -> SoA) to enable better caching and vectorization. One could argue the benchmark itself is broken, but what icc does is extremely rarely automatically doable. Hopefully SPEC CPU2017 will be better (even though the speed variant still allows automatic parallelization of a single thread run :( ).

Back to the Apple score. I also wonder if it is legit. It looks like the scaling is similar for all components against A9 (except for AES and memory perhaps). That could hint at just a frequency bump. And even it that's "only" that, I'd still be impressed.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
FWIW:

My AMD Athlon II X3 435 2.9 GHz (triple core, 64-bit Win 10) gets 1675 / 3987.

My iPad Air 2 with A8X (triple core, 64-bit iOS 9) gets 1752 / 4045 (+5% / +15%).

The AMD is my main Windows desktop, and I'm typing on it right now. With SSD and 8 GB RAM, it runs just fine for what I use it for. Its main problem is it's a power hog, at 95 Watts TDP.
 

HurleyBird

Platinum Member
Apr 22, 2003
2,670
1,250
136
If apple had better ipc (or even within spitting distance) in addition to better perf/watt than Intel's best, then obviously they'd stop using Intel's chips in favor of their own, at least in anything short of Mac Pro. They haven't, so they haven't. Don't let wishful thinking based on a single benchmark prevail over common sense.
 
  • Like
Reactions: RussianSensation

Eug

Lifer
Mar 11, 2000
23,583
996
126
If apple had better ipc (or even within spitting distance) in addition to better perf/watt than Intel's best, then obviously they'd stop using Intel's chips in favor of their own, at least in anything short of Mac Pro. They haven't, so they haven't. Don't let wishful thinking based on a single benchmark prevail over common sense.
If they switched to ARM right now in their laptops, most of my software would not work in the new machines. Sure, iTunes would be fine, but Office, Photoshop, Lightroom, Toast, and what not would not. It'd be the PowerPC --> Intel transition all over again, but this time without Rosetta.

The last time around it made sense, since it meant going from one platform with one decent performance supplier to another platform with two suppliers with much higher performance, that made Rosetta viable, while at the same time saving money on the chips.

This time it would be going from a high performance platform to another mid to high performance platform at considerable cost. It's not about whether or not they can do it but the cost involved to compete at the desktop level with a new desktop-oriented chip line, only to get similar performance and a whole bunch of developer headache. I wonder how many billions of $ it would cost to design these chips to achieve the clock speeds to compete. And the other question is whether or not Intel would give them access to their foundries to make the chips on their most advanced processes at bargain basement prices, right after Apple dropped them as a supplier for their CPUs.
 
Last edited:

HurleyBird

Platinum Member
Apr 22, 2003
2,670
1,250
136
this time without Rosetta.

You sound pretty certain they can't accomplish something similar today.

The last time around it made sense, since it meant going from one platform with one decent performance supplier to another platform with two suppliers with much higher performance

For products like iMac and MacBook, perf/watt is far more important than raw performance, and if Apple has accomplished what you people think it has, then they are so far ahead of Intel in perf/watt that, architecturally speaking, they're way further ahead of Intel than Intel was of PowerPC back in the day. If Apple is actually matching Intel's best in phones, then they leave them in their dust in desktops, not to mention the super lucrative server market that is extremely perf/watt sensitive.

In fact, if apples SoCs perform anywhere close to as good as some people think, then servers are just low hanging fruit.

that made Rosetta viable, while at the same time saving money on the chips.

You can't get cheaper than making your own chips.

a whole bunch of developer headache.

Having your entire product stack on the same architecture makes things much easier in the long run.

I wonder how many billions of $ it would cost to design these chips to achieve the clock speeds to compete.

Besides the fact that Apple is absolutely loaded, getting a design to perform better at the expense of consuming more power isn't exactly the most challenging engineering problem.

And the other question is whether or not Intel would give them access to their foundries to make the chips on their most advanced processes at bargain basement prices, right after Apple dropped them as a supplier for their CPUs.

Why would Apple care in the slightest? They've gotten along just fine without Intel's fabs.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
My point was the last time they switched they saved money on chips AND got an enormous boost in performance, to the point where using Rosetta wasn't that much of a performance penalty vs native PowerPC, and they even saved on power utilization at the same time (vs the G5), while suddenly going from one niche supplier (IBM) to potentially two competing high volume suppliers (Intel and AMD). Motorola/Freescale didn't count because they had decided to focus on embedded.

This time around they'd at best equal the performance at huge cost and if they tried to implement Rosetta it'd be at a huge performance penalty, and with huge headache to third party developers. Yes I think they could do it but the financial benefit just isn't compelling at the moment, esp. in a market where desktop sales are waning.

Maybe this will happen in time, but now is not the time.
 

avAT

Junior Member
Feb 16, 2015
24
10
81
... Yes I think they could do it but the financial benefit just isn't compelling, esp. in a market where desktop sales are waning.

This is the key question to me. The rest of details don't matter much and have been argued to death over the last couple years.

Unless Apple thinks macOS only has another 10 years of relevance left, I have to believe that one way or another the bifurcation between iOS/tvOS/watchOS devices & macOS devices has to end some way eventually. It's just in their DNA to get everything on similar architectures & development frameworks. Maybe I just don't like the idea of the Mac dying a slow death of neglect.
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
You can't get cheaper than making your own chips.

That's just plain wrong. It only gets cheaper if you sell enough devices using them and the Mac market in amount of devices sold is tiny compared to mobile:

From Q3 2016:

40 Mio iPhones
10 Mio iPads
4.3 Mio Macs

Economics of scale and the scale is clearly missing for classical CPU market.
 

HurleyBird

Platinum Member
Apr 22, 2003
2,670
1,250
136
That's just plain wrong. It only gets cheaper if you sell enough devices using them and the Mac market in amount of devices sold is tiny compared to mobile:

From Q3 2016:

40 Mio iPhones
10 Mio iPads
4.3 Mio Macs

Economics of scale and the scale is clearly missing for classical CPU market.

You're talking about the same architecture on the same foundry, but you imply that there are zero economies of scale. There are, significant ones. After all, there are very few A9Xs relative to the rest of the stack, and they still get made. Actually, if Apple CPUs are as good as you people think they are, an up-clocked A10X could service a large amount of the Mac range as well, which would greatly improve economies of scale. And of course, no middle man taking a cut.

But whatever, it's far too easy for people to rationalise a plethora of weak reasons why Apple wouldn't switch to their supposedly magical, generations ahead of Intel, CPUs in their Mac range. Here's something a lot more difficult: You have a core with as good or better IPC than Intel, with a fraction of the power consumption, albeit at lower clocks. If this scenario is true, then you'd have to be a drooling idiot not to exploit that advantage in the ultra lucrative server market. Guess what?

Common sense says that Apple's cores are not as good relative to Intel's cores as this single benchmark says they are. If this were any other company, results of a single benchmark would be met with healthy scepticism. Because it's Apple, it's interpreted as holy writ.
 
Last edited: