Apple A9 SoC

Gubbi said:
So almost exactly 50% higher Geekbench scores than A8

If we assume DDR4-3200 is used, the 6s has twice the bandwidth of the A8 in iPhone 6. That alone would yield (2^2)^(1/10)=14.9% higher Geekbench scores (memory synthetics are weighted with 20% geometrically).

That means the remaining benchmarks improve by 1.5/1.149=30.5%. The would require a A8 clocked at 1.4GHz * (1.305^10)^(1/8) = 1.95GHz.

Since it is only clocked at 1.8GHz, we can infer an IPC improvement of 1.95GHz/1.8GHz = 8.5% (assuming limited sensitivity to raw bandwidth)

That's pretty close to what was predicted here.

Cheers
Click to expand...
NO.
Your distaste for GeekBench lead you astray, you are not looking at the actual data, as it turns out the memory bandwidth subtests yield lower improvements than the integer and floating point ones.

STREAM does not generally yield maximum theoretical bandwidth results. There is a reason why the test has several different components. Download the source, flip some compiler switches, watch the results fluctuate. As they should depending on architecture and compiler.
 
Entropy said:
Your distaste for GeekBench lead you astray, you are not looking at the actual data, as it turns out the memory bandwidth subtests yield lower improvements than the integer and floating point ones.
Click to expand...

Sort of guilty (the distaste part). Didn't look at individual sub scores.

Single thread memory score for A8/A9 is 1551 vs 2287, so not double, only 47.4% higher. so only 8% higher Geekbench score from memory, - and 17% performance improvement from IPC.

Cheers
 
ltcommander.data said:
http://mashable.com/2015/09/22/appl...w/?utm_cid=mash-com-Tw-main-link#Vmyb1kCkGkkG

Mashable has iPhone 6s Plus Geekbench results from it's 1.8GHz dual core A9:
Single-core: 2521
Multi-core: 4380

Slightly higher than those twitter iPhone 6s results, but since the CPU clock speed seems the same between the 6s and 6s Plus, I'm guessing that's just normal benchmark variance.
Click to expand...
Always amazed how outlets manage to mangle Android device scores in Apple reviews.

Edit: https://browser.primatelabs.com/geekbench3/3515465
 
Last edited:
Quite impressive scores indeed. FWIW (different platforms and whatnot...) those scores (both single and multi) are very similar to what higher-end (Broadwell) Core-M devices achieve (with the exception of those configured to run at higher TDP, such as the MacBook). Looks like the A9 still doesn't have to throttle (much) in pure cpu workloads at least with geekbench (in contrast to Core M, which doesn't show better multicore scaling despite benefitting quite a bit from multithreading).
 
Last edited:
They're not updating the regular iPad this year, so Air 3 should get the A10(X, possibly.)
 
Though there's speculation there might be an Air in the spring.

Yeah this Holiday season, they're going to see if they can skim extra profits from a higher-priced, higher-margin? product.

But honestly, instead of just a speed increase in the next iPad Air and mini, there might be more interest in support for the Pencil, if it proves popular.

Or maybe even the haptic layer to support "3D Touch."

Or both.
 
Nebuchadnezzar said:
Both Qualcomm and Samsung are using LPDDR4-3200 and have been for half a year now, it's safe to assume Apple with at least deliver the same. LPDDR4-4200 (or whatever the 2133MHz speed is called) is the next upgrade coming up. We're already at double the memory bandwidth off the bat.

In any case people need to remember that memory to a SoC doesn't mean the CPU cores themselves have access to it or can saturate the full bandwidth.
Click to expand...

In either case, it seems safe to assume LPDDR4 is in with the elevated numbers we're seeing. The power savings alone, regardless of operation speed, are worth it. They'd be pushing LPDDR3 higher than anyone else that I've seen, AFAIK.

Based on the L2 cache being 3MB and the SRAM being slightly larger perhaps it's 8MB after all.
 
Last edited:
tangey said:
First GPU score I've seen.
3dMark ice storm unlimited 28403, up from 16257 on iphone6, about 75% increase.AFAIK, in absolute terms, ios devices never score great on this BM.
http://mobilesyrup.com/2015/09/22/iphone-6s-and-iphone-6s-plus-review/
Click to expand...
http://www.futuremark.com/pressrele...results-from-the-apple-iphone-5s-and-ipad-air

Futuremark claimed Cylcone was held back when doing out of order memory accesses. Maybe Twister fixes this? Otherwise the GPU speedup would have to be massive to generate a high enough graphics score to push the overall score up 75% and overcome a CPU bottlenecked physics score.
 
This document has some information about workload size. Mobile version has smaller workloads than desktop version.

http://www.primatelabs.com/geekbench3/workloads.pdf

Some subtests have very small workloads that fit even 1MB cache, such as BZip2 and Lua. Some are larger than 3MB, such as JPEG (1.6Mpix).
Lua test is particularly interesting because it's considered a "hard" test for CPU core (the data set is so small that cache generally does not matter much). A9 performs very well on this subtest and that suggests some improvement on CPU core other than cache (probably branch predictor).
 
You must log in or register to reply here.

Similar threads

I
Apple A15 SoC
2
Replies
34
Views
20K
PSman1700
P
W
Snapdragon 845 announced
Replies
3
Views
3K
wco81
W
A
Holy Wall of Text: My impressions post Playstation 4 (PS4) unveil
2 3 4
Replies
70
Views
21K
(((interference)))
I
H
IBM backstabbing sony and neglected Apple ? (long read)
5 6 7
Replies
126
Views
48K
Fox5
F
E
New article on Xbox360 GPU
3 4 5
Replies
81
Views
14K
a688
A
Back
Top