iPad Air 2 (Apple A8X) benchmarks reveal tri-core CPU surprise, 2GB of RAM confirmed

Apple unveiled the iPad Air 2, claiming the world’s thinnest tablet title, but the second generation of the airy tablet also packs some serious punch coming with the Apple A8X system chip…
This is a discussion for a news article. To read the whole news, click here

90 Comments

51. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

I'm not sure how to find the 32-bits versions of the iphone 5s. I thought you were referring to Geekbench2 and Geekbench3. The links I gave are correct (the phones are) iphone6,x is iphone 5S and iphone7,x is iphone 6. You can search for that yourself (search iphone6,1 in geekbench and it will give you the iphone 5S results. Important, no space between iphone and the 6). But same as vincelogman critizizes that about using Geekbench2 and 3 in my comparison can be said with yours. You use Geekbench 3.1.6 for iOS ARMv7s (32-bit) wich seems to be a defective version that gave low results, so not much can be inferred from that. I still would expect a mild upgrade in scores with the 64bit version of Geekbench for Android, but we'll see when it comes out.

41. spicy

Posts: 9; Member since: Oct 22, 2014

Check this (I cannot post links, so copy/paste is needed): browser.primatelabs.com/geekbench3/compare/1067175​?baseline=695393 Here you will see the difference in 32 vs 64 bit for the A7. I took some typical results (not the highest or lowest)

32. vincelongman

Posts: 5655; Member since: Feb 10, 2013

Yep that my point For example if its 40% like you say, then The K1's single would be ~2600 and multi would be ~4600 But are you sure its a 40% difference I think 40% sounds too high I though the difference would be tops 20% (that would mean ~2300 and ~4000)

37. spicy

Posts: 9; Member since: Oct 22, 2014

I have absolutely no idea if it will be 40%. The only example I have (32-bit vs 64 bit on the same type of processor: A7) is 40%. on Intel processors, this difference is much lower, but these also have heavy optimizations to run 32-bit code as fast as 64-bit. As the architecture is also different, the difference can be quite unpredictable (the tegra uses a code morphing system). Speed gains in ARMv8 will be linked to the optimization they have done for ARMv7.

56. janno

Posts: 144; Member since: Aug 19, 2014

Yeah the difference is anywhere from 10 to 30 percent for ARMv8 AArch64 mode.

55. janno

Posts: 144; Member since: Aug 19, 2014

Not to mention Denver is a 28nm chip (which affects both CPU and GPU). Apple's is 20nm, a whole process generation ahead. And Denver still beats it.

4. Tritinum

Posts: 471; Member since: May 06, 2014

damn..exceeds expectations, k1 still got the lead though

6. Zomer

Posts: 361; Member since: May 31, 2013

In benchmarks yes, in real life no, because it runs Android.

9. ArtSim98

Posts: 3535; Member since: Dec 21, 2012

You seem to still live in 2010.

18. kassi

Posts: 109; Member since: May 25, 2013

i hate to say its true

22. Tritinum

Posts: 471; Member since: May 06, 2014

No more, cause of Android 5.0

28. Zomer

Posts: 361; Member since: May 31, 2013

I live in 2014 and you have to be a fool to think that Android will 1 day be as smooth as iOS is. You can't compete with a company that develops the software and the hardware.

45. shuaibhere

Posts: 1986; Member since: Jul 07, 2012

It already as smooth on moto e for that matter!!

63. slh163

Posts: 140; Member since: May 19, 2014

actually it's even smoother fool. wait for an speed test between stock lollipop and ios 8 then maybe you gonna find another excuses for your beloved os. even kitkat in nexus 5 is on par or even faster than ios 8 on iphone 6. so what you're saying is just a load of BShttps://www.youtube.com/watch?v=hPhkPXVxISY&list=UUB4dWKbwgl3U3ohBmrT7k3w

74. Zomer

Posts: 361; Member since: May 31, 2013

Just use the browser and come again fool. It's funny how all those Android fanboys come up with shitty comparison video's, just use them both and you will know what i'm talking about.

68. VHMP01

Posts: 93; Member since: Aug 27, 2014

Then why did iOS 8.1 got so many bugs and was pull out from downloads? I mean like a month later iOS 8.2 has not made much noise! Yes, Android does not lag and iOS does in 2014!

5. Zomer

Posts: 361; Member since: May 31, 2013

Lol imagine when Apple puts a quad-core A9 in the iPhone 6S, it will not only bend, but it will also smoke of power.

7. ArtSim98

Posts: 3535; Member since: Dec 21, 2012

Tri core CPU's rock! (Wii U:)

14. vincelongman

Posts: 5655; Member since: Feb 10, 2013

Also Xbox 360 :)

8. gigaraga

Posts: 1454; Member since: Mar 29, 2013

Haha, thought the latest iPad needed 2GB RAM. It was in need of it (although not desperately)

10. TheCynicalTechie

Posts: 78; Member since: Nov 19, 2012

For the last time, cross-platform benchmarks are useless, because OS heavily impacts performance. The only way this would be slightly useful would be if it was using the SAME CPU to test the efficiency of the OS. Otherwise it's meaningless.

13. vincelongman

Posts: 5655; Member since: Feb 10, 2013

Also 64-bit Geekbench vs 32-bit, since Geekbench on Android has been update yet

30. spicy

Posts: 9; Member since: Oct 22, 2014

on the iphone 5s, it seems to be a difference of 40%: 1800 (32-bit) vs 2500 (64-bit). This is going to be interesting....

21. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

And you are quite wrong there. Some benchmark are useful to give you user experience, like the browser benchmarks. They are difficult to compare, but still useful as they give you the final experience of the device, including both HW and SW. But Geekbench is done in native code (what you call cross-platform), so very little left for the OS, they pretty much measure just the performance of the CPU/RAM. Example: you can see that the same chip (intel Atom Z3745) gets the SAME results in Android and in Windows.http://browser.primatelabs.com/geekbench3/search?dir=desc&page=2&q=Z3745&sort=score So yes> Geekbench can be used cross-platform. Same goes for some others, like Antutu, GFXBench (when comparing OpenGL ES in different OS), etc Even when comparing different Drivers/OS/Software they are far from MEANINGLESS, as they just compute the diffences on all those levels toguether. If you don't know how to interpret benchmarks is your problem, not the benchmarks one.

12. aesthetics

Posts: 128; Member since: Oct 02, 2014

Holy sh*****t ....!

15. kassi

Posts: 109; Member since: May 25, 2013

lol apple switched from singlecore king to multicore king, that was unexpected

16. aesthetics

Posts: 128; Member since: Oct 02, 2014

android fanboy excuses coming... :)))))))))) as usual :DDDDD

17. livyatan

Posts: 867; Member since: Jun 19, 2013

And there you have it. I predicted that the A8X will be faster than respective Intel chips (Haswell Core i3/i5Y). But it is even better than that, it will embarrass even the 14nm Core M Y variants (each of those will cost over 150$). I can barely imagine the ability of the 14nm quad core A9X - software compatibility aside, it should be good enough for the entire MacBook line, even the MacBook pro if clocked over 2.5Ghz. This is amazing, makes me want to have the iPad for the first time. It is almost as fast as the Core i5 Surface Pro 3, all while being incredibly light and thin. And it will also finally be able to multitask properly with that RAM and the upcoming split screen software update. A perfect tablet!

23. vincelongman

Posts: 5655; Member since: Feb 10, 2013

Nope, this doesn't match the Intel Core i5-4300U (2806 single and 5437 multi) The Core M 5Y70 Broadwell-Y is supposed to be faster than the Intel Core i5-4300U Early benchmarks show almost double CPU speed when fan cooled (see Sunspider which is more single core, 3DMark is more graphics and multi core), though will be a smaller gap when passive cooled The even the MacBook Air will still use Broadwell, not to mention the Pros Also because there's much more x86 Mac OS X apps than iOS ARM appshttp://browser.primatelabs.com/processor-benchmarkshttp://www.anandtech.com/show/8515/quick-look-at-core-m-5y70-and-llama-mountain

36. livyatan

Posts: 867; Member since: Jun 19, 2013

I didn't state that it matches the Core i5U, but that it almost matches it in processing speed. If you don't cherry pick the top geekbench results, you'll see that Core i5-4300U stands around 4500-5500 multi and 1800-2600 single. And that while being at at least double the TDP of this Apple chip! As for Core M-Y, note that it has a paltry 800 Mhz base clock in order to fit into something without a fan. Do not compare its unrestricted benchmarks because it is misleading. Besides, I didn't say A8X is ready to replace Intel chips in MacBook Air, it's still a bit too early for that. But A9X should be more than fine. X86 software could actually be recompiled to ARMv8 with Xcode really easily.
This copy is for your personal, non-commercial use only. You can order presentation-ready copies for distribution to your colleagues, clients or customers at https://www.parsintl.com/phonearena or use the Reprints & Permissions tool that appears at the bottom of each web page. Visit https://www.parsintl.com/ for samples and additional information.