Apple A9X Geekbench

Page 11 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Hi-Fi Man

Senior member
Oct 19, 2013
601
120
106
This PCSX2 forum post shows very limited speed up for AVX over SSE. It's between 0 and 20%.

I couldn't find any Dolphin number.

So yes definitely nice to have, but not something that in its own would make 256-bit width mandatory.

Due to the way emus work being shy of 100% speed can make a game unplayable or not very enjoyable which means that extra percentage is very important.
 

stingerman

Member
Feb 8, 2005
100
11
76
And, there are large parts of the Apple SOCs that are undocumented as to their functions. No doubt specialized circuitry to speed up certain types of functions that lend themselves to vector style processing.
 

khon

Golden Member
Jun 8, 2010
1,319
124
106
The benchmarks are nice, but what would you actually use that graphics power for ? The iPad Pro only supports phone/tablet apps, not the full versions. And it does not support a mouse or trackpad, so gaming will be extremely limited.
 

cytg111

Lifer
Mar 17, 2008
23,534
13,107
136
Due to the way emus work being shy of 100% speed can make a game unplayable or not very enjoyable which means that extra percentage is very important.

So what you are saying is that in performance constrained situations you are performance constrained. I would make that call one out of one times, all the time.
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Don't forget that the thing is having to drive a positively huge resolution screen and you can pan round graphics very, very fast on a tablet. Think that alone justifies it really. Might get used for some truly serious things down the line.

For me its nice as it can keep panning round 50k ordnance survey maps of the UK absolutely smooth. That drops to not quite so if I drop a bunch of preexisting tracks on there as well, so it does need much of the performance.

Its a whole lot smoother than the equivalent on my desktop but that's old, badly coded and running in wine to boot so not a surprise! There are though definitely some places where the software for iOS simply is ahead of desktop stuff.
 

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
So what you are saying is that in performance constrained situations you are performance constrained. I would make that call one out of one times, all the time.

Well there is a reasonable point that if you were to say, graph user enjoyment vs performance you'd see a bigger knee in the curve for a lot of emulators. For other apps 90% perf will make you a little more annoyed with things being slower, while in emulators it can totally break the experience (stuttering music, game speed that the developers never intended)

But Apple probably doesn't give a crap about emulator performance, certainly not to the point where it'd affect their hardware choices. And AVX+ isn't relevant for the vast majority of emulator setups anyway (people don't usually use the software renderer with PSCX2)
 

tipoo

Senior member
Oct 4, 2012
245
7
81

We already talked about the GFXbench results pages ago when Ars did the same test. On mobile it uses largely FP16, on laptops it uses FP32, so the scores aren't comparable.

Look at the 3DMark scores, where the Iris Pro goes from losing to the A9X to beating it by 3X. That's not the defining benchmark either, but it shows that beating laptop chips is an anomaly.

Remember when Nvidia called the X1 a 1Tflop part? that was FP16, and when using the same FP32 that every other card is ranked at, it performs under half of that.
 

Nothingness

Platinum Member
Jul 3, 2013
2,751
1,397
136
Look at the 3DMark scores, where the Iris Pro goes from losing to the A9X to beating it by 3X. That's not the defining benchmark either, but it shows that beating laptop chips is an anomaly.
Or 3DMark is an anomaly
 

Zodiark1593

Platinum Member
Oct 21, 2012
2,230
4
81
Or 3DMark is an anomaly

Or that these benchmarks royally suck for comparing cross-platform. No way in hell a Cape Verde GPU (radeon hd 7770) is losing to anything in a tablet today unless it was heavily penalized for something like latency (that 8 mb cache in the A9x probably helps a lot in tile rendering). Likewise, I also find it hard to buy an Iris Pro stomping the A9x gpu 3x over, maybe 2x would be more accurate.
 

Zink

Senior member
Sep 24, 2009
209
0
0
A naked A9X

It looks like there's no L3 cache. Surprising...

Wow that is weird. Anandtech left L3 of their initial iPad Pro table so it looks like it is true.
http://anandtech.com/show/9780/taking-notes-with-ipad-pro/2

I though L3 was mainly needed to improve latency but considering that A9 needed double the L3 of A8 the main reason for having a big L3 must be to help improve the throughput between L2 and main memory. A9X has enough memory bandwidth so it didn't need the L3, unless there is another technology in there to improve prefetch performance?

Atom also chooses not to use an L3 and it is aimed at similar workloads but it just seams strange to ship an 8 MB L3 in A9 and then drop it for A9X.
 
Last edited:

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
Wow that is weird. Anandtech left L3 of their initial iPad Pro table so it looks like it is true.
http://anandtech.com/show/9780/taking-notes-with-ipad-pro/2

I though L3 was mainly needed to improve latency but considering that A9 needed double the L3 of A8 the main reason for having a big L3 must be to help improve the throughput between L2 and main memory. A9X has enough memory bandwidth so it didn't need the L3, unless there is another technology in there to improve prefetch performance?

Atom also chooses not to use an L3 and it is aimed at similar workloads but it just seams strange to ship and 8 MB L3 in A9 and then drop it for A9X.

Lack of L3 seems strange at first, but then you consider that 3MB of L2 cache is actually a lot of LLC/core. It's comparable to Intel's Core i3 processors that are 2C4T with 3MB of L3 cache; they still do pretty well with it. They have 256KB L2/core in addition to that, but since the L3 is strictly inclusive that doesn't increase the overall capacity. And the latency benefit is somewhat offset by the L1 icache/dcache being twice the capacity in Apple's designs.

This leads me to believe that the large L3 in A9 is first and foremost there for bandwidth amplification and reduced power consumption for the GPU. However, when the L3 cache is missed it's actually worse for latency and power vs there being no L3 cache at all, so the hit rate has to be beyond a certain point for it to be a win. iPad Pro has a much higher resolution than iPhone 6s (5.6x) or 6s Plus (2.7x), to where 8MB of L3 comes nowhere close to even storing a framebuffer. Apple may have found that the GPU hit rates are just too low for the applications they're interested in.
 

Mondozei

Golden Member
Jul 7, 2013
1,043
41
86
The benchmarks are nice, but what would you actually use that graphics power for ?

Deep learning. Or more specifically, used in cars. Self-driving cars needs a ton of sensors and cameras, and the better the GPU is, the better the system.

Also, Apple is most likely thinking about this chip in more ways than just ultra mobile. We've been talking about Apple replacing Intel in their macbooks for some time. It used to be laughed at, then it became implausible, now it's openly speculated. Even if most agree that it won't happen anytime soon, the technical feasibility is there.

Given Apple's self-driving car project alone, massive GPU computational power in a very small power envelope is reason enough.
 

dark zero

Platinum Member
Jun 2, 2015
2,655
138
106
Maybe the A9X didn't had L3 since it would be so much overkill against Intel Core m and they would react improving dramatically their solutions.
 

StrangerGuy

Diamond Member
May 9, 2004
8,443
124
106
The benchmarks are nice, but what would you actually use that graphics power for ? The iPad Pro only supports phone/tablet apps, not the full versions. And it does not support a mouse or trackpad, so gaming will be extremely limited.

At the very least Apple can use it for a future killer app available for it's own OS. Can't say the same for Android.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |