Apple A9X Geekbench

[Arachnotronic]

So I think AnandTech will run SPEC2k6 on the iPad Pro and probably some Skylake systems, so we will get a "real" CPU comparison between A9X and Skylake soon.

I personally can't wait to see this comparison. Should be incredibly exciting to see

 

[Nothingness]

So I think AnandTech will run SPEC2k6 on the iPad Pro and probably some Skylake systems, so we will get a "real" CPU comparison between A9X and Skylake soon.

I personally can't wait to see this comparison. Should be incredibly exciting to see

Could be disappointing too. It's very easy to mess this up.

First the comparison will be against results compiled with icc SPEC compiler, unless AT also compiles using more fair compilers. Might be great if they did that on for instance a Surface Pro 4.

Second I'd be surprised if they are going to do what should be done regarding 32- vs 64-bit compilations; look at any SPEC 2006 score to see what I mean. But that'll be OK if they also run all tests in 64-bit on SP4.

Third will they play with the flags?

I sound pessimistic, but what I saw for SPEC 2k on Android by AT doesn't make me optimistic. But let's wait and see what we get

 

[Arachnotronic]

Could be disappointing too. It's very easy to mess this up.

First the comparison will be against results compiled with icc SPEC compiler, unless AT also compiles using more fair compilers. Might be great if they did that on for instance a Surface Pro 4.

Second I'd be surprised if they are going to do what should be done regarding 32- vs 64-bit compilations; look at any SPEC 2006 score to see what I mean. But that'll be OK if they also run all tests in 64-bit on SP4.

Third will they play with the flags?

I sound pessimistic, but what I saw for SPEC 2k on Android by AT doesn't make me optimistic. But let's wait and see what we get

Aw, you've taken a lot of the fun out of this for me

 

[Thanatosis]

Is anybody going to buy an iPad Pro to test this stuff? I have a friend who got one but I won't see him for a couple weeks. I do not feel I should spend $1k on an ipad when I already have an Air 2.



I am very curious what Adobe will be doing with their software for the ipad pro, if they give full functionality comparable to desktop it could be a very powerful and useful device. I don't see why they couldn't do the same for the air 2, though.

 

[Arachnotronic]

Is anybody going to buy an iPad Pro to test this stuff? I have a friend who got one but I won't see him for a couple weeks. I do not feel I should spend $1k on an ipad when I already have an Air 2.



I am very curious what Adobe will be doing with their software for the ipad pro, if they give full functionality comparable to desktop it could be a very powerful and useful device. I don't see why they couldn't do the same for the air 2, though.

I ordered one but cancelled it after using one in the store. I just can't justify it after using it.

 

[Thanatosis]

I ordered one but cancelled it after using one in the store. I just can't justify it after using it.

I had basically the same sentiment after using it. It is pretty though, but I already feel the Air 2 is a bit too heavy with the smart case for one-handed use comfortably and this will be worse.

 

[Arachnotronic]

I had basically the same sentiment after using it. It's pretty though.

Oh it's a gorgeous, well-engineered, finely crafted piece of Apple Tech and honestly I think Apple deserves somebody's money for how well made it is. Just...not my money.

I'd rather save that cash for the iPhone 7 Plus...or next year's iteration which will surely come in Rose Gold + include 3D Touch.

 

[jihe]

Who is going to drop $2600 on surface book when they can get an ipad pro for half the price that has better GPU and CPU performance, is thinner and lighter and looks better to boot?

People who actually, you know, do work on their tablet/laptops.

 

[stingerman]

People who actually, you know, do work on their tablet/laptops.

What kind of work are you referring to? All the key productivity Apps are on iOS and many many more

 

[jihe]

What kind of work are you referring to? All the key productivity Apps are on iOS and many many more

xcode? And I don't know how anyone doing anything serious can deal with the pain that is sharing files between apps. IOS is fundamentally not for production.

 

[Brunnis]

Out of curiosity, I put my 6700K at 2.2GHz without any Speed Step or turbo active. In other words, it's locked at 2.2GHz. Memory is 16GB DDR4 @ 2667MHz 15-15-15-36-1T. My Windows 10 installation is clean and I made sure there were no appliactions running in the background while Geekbench ran. Here's the comparison to the iPad:

http://browser.primatelabs.com/geekbench3/compare/4149721?baseline=4183457

If we assume that the iPad actually runs at 2.16GHz (and no higher), the performance/IPC advantage of Apples core is pretty big in some cases. Many of the test results seem to hover in the +30% area. That's if the results are actually meaningful at all. We can speculate, but I don't see any way of determining who's right with any certainty without having much more varied test results available.

For completeness, here's the bone stock Geekbench score of the system with otherwise exactly the same running conditions. Multicore enhancement disabled.

http://browser.primatelabs.com/geekbench3/4183670

 

[ShintaiDK]

We already established that geekbench is horrible biased towards mobile, specially IOS. The methods used for the windows version is an outright joke. Plus the dataset is also different. Not sure why people keep using this flawed benchmark.

 

[Arachnotronic]

Out of curiosity, I put my 6700K at 2.2GHz without any Speed Step or turbo active. In other words, it's locked at 2.2GHz. Memory is 16GB DDR4 @ 2667MHz 15-15-15-36-1T. My Windows 10 installation is clean and I made sure there were no appliactions running in the background while Geekbench ran. Here's the comparison to the iPad:

http://browser.primatelabs.com/geekbench3/compare/4149721?baseline=4183457

If we assume that the iPad actually runs at 2.16GHz (and no higher), the performance/IPC advantage of Apples core is pretty big in some cases. Many of the test results seem to hover in the +30% area. That's if the results are actually meaningful at all. We can speculate, but I don't see any way of determining who's right with any certainty without having much more varied test results available.

For completeness, here's the bone stock Geekbench score of the system with otherwise exactly the same running conditions. Multicore enhancement disabled.

http://browser.primatelabs.com/geekbench3/4183670

iPad runs at 2.26GHz but your point is still basically valid.

What's interesting is when I went down to a local Apple store and ran some JS benchmarks on Safari browser on A9X, Core M, and Broadwell i5 (quad core), I noticed that a 2.4GHz MacBook (1.1GHz base, 2.4GHz max turbo) is actually beaten by the iPad Pro's A9X @ 2.26GHz.

Nothing conclusive obviously, but I do think in terms of raw CPU st perf/clock there is a very real chance that Twister is actually ahead of Skylake. However, I do think more tests need to be done before we can say anything definitive. Also note that Skylake scales to much, much higher frequencies so it's not exactly apples to apples.

But yeah, Twister core from Apple? Good stuff.

 

[tempestglen]

Could be disappointing too. It's very easy to mess this up.

First the comparison will be against results compiled with icc SPEC compiler, unless AT also compiles using more fair compilers. Might be great if they did that on for instance a Surface Pro 4.

Second I'd be surprised if they are going to do what should be done regarding 32- vs 64-bit compilations; look at any SPEC 2006 score to see what I mean. But that'll be OK if they also run all tests in 64-bit on SP4.

Third will they play with the flags?

I sound pessimistic, but what I saw for SPEC 2k on Android by AT doesn't make me optimistic. But let's wait and see what we get

It is said that the difference between ICC and GCC(new version) is quite small, not so obvious as before.

And anandtech will disable PAR when testing SPEC06?

 

[Nothingness]

It is said that the difference between ICC and GCC(new version) is quite small, not so obvious as before.

Here is gcc on a 3.8GHz Haswell (most likely some 4670: https://vmakarov.fedorapeople.org/spec/spec2006.topka/gcc/home.html
Here is icc on a 4670: https://www.spec.org/cpu2006/results/res2014q3/cpu2006-20140715-30459.html

So that's base 35 / peak 37 vs 56/58. See what I mean?

Of course icc flags are playing with 32 vs 64-bit flags, while the gcc run is 64-bit only.

OTOH as was noted on RWT, gcc is competitive against icc for 403.gcc, because the latter is very difficult to break, so Intel found no trick to make its score shine. That's one of the reasons most people only care about 403.gcc.

And anandtech will disable PAR when testing SPEC06?

I think only icc (and other proprietary compilers) do autopar.

 

[itsmydamnation]

We already established that geekbench is horrible biased towards mobile, specially IOS. The methods used for the windows version is an outright joke. Plus the dataset is also different. Not sure why people keep using this flawed benchmark.

its pretty funny author of said benchmark says the difference between the two work sets is ~1% and in the next version all versions will be using the same work set. So what will be your excuse then?

 

[ShintaiDK]

its pretty funny author of said benchmark says the difference between the two work sets is ~1% and in the next version all versions will be using the same work set. So what will be your excuse then?

The dataset showed much more variation in different benches. But it still doesn't change the base setup.

You can call it what you want.

 

[Nothingness]

The dataset showed much more variation in different benches. But it still doesn't change the base setup.

The variations are not huge: http://browser.primatelabs.com/geekbench3/compare/3594692?baseline=3594646
But they exist, so results have to be taken with care.

 

[Arachnotronic]

its pretty funny author of said benchmark says the difference between the two work sets is ~1% and in the next version all versions will be using the same work set. So what will be your excuse then?

FPiednoel on Twitter showed that the difference can be much greater on some Intel mobile devices.

 

[podspi]

Anybody know what the TDP/power usage difference is between the 6200U and A9X?


A9X makes Core M look really bad...

 

[StrangerGuy]

The dataset showed much more variation in different benches. But it still doesn't change the base setup.

You can call it what you want.

Ohhhhh, so a few outliers means the benchmark is invalid. That never happens in pure PC master race land before, right?

Even you must know how lame your excuse is.

 

[jihe]

You know what's funny? The only reason that there is any argument over Apple's cpu performance is that you can't do jack shit with it.

Run a kernel compile, do some rendering, play a few common games, churn out a few F@H WU's, the real world performance of a cpu is really not that hard to gauge. But guess what? You can't do any of that on the ipad pro!

So the A9X might well have quantum computational powers but do you even care?

 

[dahorns]

Anybody know what the TDP/power usage difference is between the 6200U and A9X?


A9X makes Core M look really bad...

In geekbench and gfxbench (which isnt really comparable). Octane, kraken, sunspider, and webxprt are all roughly in line with top line broadwell core m. 3dmark favors broadwell core m.

 

[Hi-Fi Man]

In geekbench and gfxbench (which isnt really comparable). Octane, kraken, sunspider, and webxprt are all roughly in line with top line broadwell core m. 3dmark favors broadwell core m.

Web benchmarks are meaningless especially when you're using two different browsers.

It's silly that anyone could think A9 can beat skylake/haswell in IPC. Does A9 have 256 bit SIMD? No.

 

[Arachnotronic]

It's silly that anyone could think A9 can beat skylake/haswell in IPC. Does A9 have 256 bit SIMD? No.

Why is it silly? A9 is a wide design with a short pipeline. It could very well be a higher IPC design, but obviously Haswell/Skylake can go to much higher "C".