Apple A13: Launch 2019-09-10 - Geekbench 4 & 5 scores available (GB4 multi-core 14070)

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

thunng8

Member
Jan 8, 2013
123
12
81
Apple shows graphs and brags about both CPU and GPU.
They claim to have unparalleled hardware x software integration and the iPhone ships with iMovie.
You are going way off topic.

CPU and GPU performance does not equate to video encoder performance as it has been stated many posts ago. It would have been great if Apple did increase the performance in this respect.

Maybe they concentrated on quality of the encode or supporting new encoding standards - just my speculation.
 

Kirito

Junior Member
Jun 20, 2017
12
7
51
I just asked:


Then people brought various things to the discussion.
Resolutions, codecs, VPU, CPU, GPU...

My question remains unanswered: "Why the iPhones are so slow here?"
Is just so Apple can sell more iPads?
Mate your question has already been answered by all the posts before this. Jeez. Just not the answer you want to hear but come on. So many different metrics being compared here so you can't just say the improvement encompasses everything the SoC ever does!

For example 5x faster random reads on my optane drive didn't get me 5x speedups in everything I do, does it? More like a noticeable difference yes, but not on every metric out there it doesn't!
 
Mar 11, 2004
20,050
2,267
126
I just asked:


Then people brought various things to the discussion.
Resolutions, codecs, VPU, CPU, GPU...

My question remains unanswered: "Why the iPhones are so slow here?"
Is just so Apple can sell more iPads?
I doubt anyone on here could answer that, and if I had to guess its because its something that is either still fast enough or is not used enough for end users to make much of a fuss about it. Apple likely has put most further development towards newer formats (didn't they switch over almost entirely to h265 for their own video like a gen or two back? I think they're in the process of doing the same thing to a new format, AV1 I think its called).

Its possible there's some other things happening that make the comparisons not fully valid. Which that's one of the things that's been bothering me about mobile, there's a lot of obfuscation that happens that makes it near impossible to properly show what is actually happening without fully knowing what the code is doing and then there's arguments over such. Heck, even on the PC side its difficult to actually compare two pieces of software both doing encoding even when they use the same encoder, because they'll have different default settings and tunes, where you'd have to delve into the code to find out which is which. And you'd need to understand the code as well as understand encoding and the encoder itself. Its at least possible though. I think it might be possible on mobile as well but I'm not sure and I'm not sure. But it seems like there is inherently extra layers making the situation more opaque.
 

Eug

Lifer
Mar 11, 2000
22,710
303
126
Apple likely has put most further development towards newer formats (didn't they switch over almost entirely to h265 for their own video like a gen or two back? I think they're in the process of doing the same thing to a new format, AV1 I think its called).
All new Apple iDevices support HEVC video recording. Furthermore Apple’s Live Photos now use HEVC, and such HEVC Live Photos as well as HEVC videos can be “texted” to other iDevices via iMessage. iTunes streaming is also HEVC if the client device supports it, and that includes any iPhone or iPad that has A9 or later.

IOW, h.264 is in transition to becoming a legacy format at Apple.

There is no support for AV1 yet.
 
Last edited:

Ajay

Diamond Member
Jan 8, 2001
6,982
2,309
136
Wonder when will they put one of their own SoC into some of their notebook.
Well, the golden handcuffs of x86 compatibility (large software base), probably means not for a while - if ever.
The Mac line is pretty small compared to the iPhone market, so it may never be worth the effort.
 

Eug

Lifer
Mar 11, 2000
22,710
303
126
Well, the golden handcuffs of x86 compatibility (large software base), probably means not for a while - if ever.
The Mac line is pretty small compared to the iPhone market, so it may never be worth the effort.
At this point it’s just rumour but Bloomberg has reported that the frameworks for the transition are in place and the work is already underway, with the target for the first ARM Mac being 2020 or 2021. Take that as you will.
 

Ajay

Diamond Member
Jan 8, 2001
6,982
2,309
136
At this point it’s just rumour but Bloomberg has reported that the frameworks for the transition are in place and the work is already underway, with the target for the first ARM Mac being 2020 or 2021. Take that as you will.
Apple ran MacOS X on x86 long before they moved over. At that point, the upward tragectory of Intel CPUs look very good (performance wise). I supposed that it's a mystery how well ARM will scale up for laptop applications and standards based components and peripherals (though a CPU headroom of 15-25 watts would be huge for Apple's ARM CPUs). Emulation, as a transitional surrogate, is very expensive. Apple are the only ones who really know.
 

scannall

Golden Member
Jan 1, 2012
1,659
1,016
136
At this point it’s just rumour but Bloomberg has reported that the frameworks for the transition are in place and the work is already underway, with the target for the first ARM Mac being 2020 or 2021. Take that as you will.
iOS is just OSX with the desktop components left out, and touch interface ones put in. To speculate they have ARM running on a laptop or a desktop in their labs is pretty easy.
 
  • Like
Reactions: moinmoin

soresu

Golden Member
Dec 19, 2014
1,226
464
136
Emulation, as a transitional surrogate, is very expensive. Apple are the only ones who really know
Not really, you can use Windows on ARM as a pretty good yardstick there, many big software packages on Mac exist on Windows too, so it's unlikely to get much better than MS's own efforts on that front.
 

Eug

Lifer
Mar 11, 2000
22,710
303
126
Apple ran MacOS X on x86 long before they moved over. At that point, the upward tragectory of Intel CPUs look very good (performance wise). I supposed that it's a mystery how well ARM will scale up for laptop applications and standards based components and peripherals (though a CPU headroom of 15-25 watts would be huge for Apple's ARM CPUs). Emulation, as a transitional surrogate, is very expensive. Apple are the only ones who really know.
I really like my 2017 MacBook Core m3, but it has literally just half the Geekbench 4 score of the consumer oriented A13.

And then Apple never updated that MacBook again, probably partially because Intel couldn't provide the chips. Not only did the CPU speed not advance significantly, they never put in the desired Thunderbolt controller. Then Apple released a MacBook Air in 2018 with a semi-custom Intel i5-8210Y CPU on Intel's 10 nm, and yet it gets a paltry 13% more performance than my 2017 MacBook Core m3-7Y32, which actually is the low end entry level model that 2017 MacBook year. The MacBook Air in 2018 actually does worse than the i7-7Y75 MacBook from 2017.

I think Apple is just sick of waiting on Intel and paying premium pricing too for those chips.
 

Ajay

Diamond Member
Jan 8, 2001
6,982
2,309
136
I think Apple is just sick of waiting on Intel and paying premium pricing too for those chips.
Good point. Jobs would probably be livid over Intel’s problems, though it seemed like he was mellowing some.
 

Eug

Lifer
Mar 11, 2000
22,710
303
126
Call me nuts but I went through all of the well over 100 Geekbench entries. o_O

All iPhone 11 are 3844 MB
All iPhone 11 Pro are 3759 MB
All iPhone 11 Pro Max are 3740 MB
There are no 6 GB models, contrary to popular rumour.

There are two clock speeds, 2.66 GHz and 1.86 GHz. I think they may actually be 2.6666666667 and 1.8666666667 GHz. Because with those numbers, the lower 1.86 GHz clock speeds is exactly 70% of the higher 2.66 GHz. Apple does have a low power battery mode that you can manually toggle on and off, so this may be reflective of that. However, on my 7 Plus with A10, while toggling on low power mode markedly reduces performance, it doesn't change the reported the clockspeed in Geekbench 5. Perhaps Geekbench reports clock speeds differently on A13, I don't know.

GB5 performance at 1.86 GHz is roughly only half (not 70%) of the higher 2.66 GHz. Multi-core vs. single-core multiplier is 2:1 at 1.86 GHz, but about 2.6:1 at 2.66 GHz.

Most of the speed variation concerns multi-core. Single-core speed is quite consistent, but multi-core has a wide range. Throttling? Background processes?
 
Last edited:

Nothingness

Platinum Member
Jul 3, 2013
2,153
397
126
GB5 performance at 1.86 GHz is roughly only half (not 70%) of the higher 2.66 GHz.
Isn't that Geekbench running on the small core? If so it's expected that it would get less than the 70% of clock difference.
 

Eug

Lifer
Mar 11, 2000
22,710
303
126
Isn't that Geekbench running on the small core? If so it's expected that it would get less than the 70% of clock difference.
Yes that could be it, but how would you run Geekbench on just the small cores? Is that something you can set in the developer tools?

And would that mean the small cores are running at a different clock speed? Or is that a low power mode clock speed?
 

Thala

Senior member
Nov 12, 2014
963
306
136
And would that mean the small cores are running at a different clock speed? Or is that a low power mode clock speed?
The small cores most likely running at lower voltage and frequency - very similar to most big.LITTLE implementations.
 
  • Like
Reactions: Eug

thunng8

Member
Jan 8, 2013
123
12
81
Thanks - I've never seen IBM power geekbench results before.

The Power results do seem a bit handicapped though. Crypto is slow - probably using software instead of hardware - and it doesn't seems to use Altivec either that's why SGEMM, SSFT is also slow.

Looking at the 3352 integer only score might be more representative of the single thread result.

In any case, these Power systems are optimised for server workloads and would not as fast for typical client workloads.

Eg. IBM Power posts some very impressive results - eg. SAP 2 Tier where an IBM Power9 server with 192 cores is approx 2X faster than the top end 224 core Xeon Platinum system.
 

DrMrLordX

Lifer
Apr 27, 2000
15,812
4,777
136
I wouldn't expect anything like GeekBench or (gag) Antutu to be suitable for POWER anyway. There were all kinds of odd performance inconsistencies in those POWER9 GB4 scores. Look at the memory, for example.
 
  • Like
Reactions: soresu

soresu

Golden Member
Dec 19, 2014
1,226
464
136
I wouldn't expect anything like GeekBench or (gag) Antutu to be suitable for POWER anyway. There were all kinds of odd performance inconsistencies in those POWER9 GB4 scores. Look at the memory, for example.
Agreed, it's built more for consumer workloads than something like a server oriented chip.
 

ASK THE COMMUNITY