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. So far, we only knew that Apple has ramped up the number of transistors in the A8X to a whopping 3 billion, up from 2 billion in the A8 and 1 billion in the A7. This remarkable achievement, however, did not reveal much about the inner workings of the Apple A8X: it’s most likely a 20nm chip, but what about its CPU and GPU?

The iPad Air 2 is yet to arrive in stores, but some folks have gotten an early look at it and ran it through benchmarks. The results are in and so are some very interesting revelations: Geekbench reports that the iPad Air 2 comes with a very unorthodox tri-core CPU and each core gets a bump in nominal clock speed from 1.4GHz in the A8 to 1.5GHz in the A8X.

This additional core explains a lot about the 1 billion increase in transistors, and it further cements the point about the iPad Air being in a whole different league in terms of productivity when compared to the iPhone. This border was blurred last year when both the iPhone and the iPad used largely the same Apple A7 chip, while this year Apple is eager to make a more clear distinction between the two.

The iPad Air 2 has also got 2GB of RAM, double the amount of memory on the iPhone 6 and 6 Plus, and that’s important for multitasking.

Geekbench results give us a well-rounded look at CPU performance:

*iPad Air 2, Nexus 9 results are preliminary

Higher is better
Samsung Galaxy NotePRO 12.2
936
Google Nexus 9
1903
Google Nexus 7 (2013)
637
Apple iPad mini 2
1355
Apple iPad Air 2
1811
Apple iPad Air
1466
Apple iPhone 6 Plus
1625
Apple iPhone 6
1630
Samsung Galaxy Note 4
1112.67

*iPad Air 2, Nexus 9 results are preliminary

Higher is better
Samsung Galaxy NotePRO 12.2
2736
Google Nexus 9
3166
Google Nexus 7 (2013)
1992
Apple iPad mini 2
2486
Apple iPad Air 2
4488
Apple iPad Air
2644
Apple iPhone 6 Plus
2918
Apple iPhone 6
2927
Samsung Galaxy Note 4
3259.67



Interestingly, we can clearly see that after years of enjoying being the undisputed king of single-core performance, Apple is finally seeing some solid competition from the Nexus 9 with its Nvidia Tegra K1 chip. Both the Apple A8X and the Tegra K1 are contemporary 64-bit platforms with large CPU cores, and it shows in the results. The K1 is actually a bit ahead of the Apple A8X in terms of single-core compute.

Cyclone vs Denver


Technically, the Denver CPU core is also a wider core than Apple’s Cyclone: it has a 7-issue superscalar in-order design with a large 13-cycle pipeline, and if you’re wondering why Nvidia has gone with in-order rather than out-of-order (OoO), the company claims this is because of a technique called “Dynamic Code Optimization”. Here is what Nvidia says about this:



Apple’s Cyclone core on its part is also a large, 6-issue wide OoO design with a mispredict penalty of 14 to 19 cycles, similar to that of Intel’s desktop architectures.

With all those technicalities in mind, it’s worth also pointing out that having three CPU cores results in a spectacular multi-core performance for the iPad Air 2. In fact, comparisons with the 2011 MacBook Air (with what is assumed to be a basic, Core i5 Sandy Bridge configuration, show that the iPad Air 2 actually beats the three year old notebook. Impressive!

Apple iPad Air 2 GPU benchmarked


Finally, we’re also seeing the first results for graphics performance on the iPad Air 2, courtesy of 3D Mark Ice Storm Unlimited where the iPad Air 2 scores 21660, less than the 25700-ish score of the Tegra K1. We still have no clear idea what is the exact GPU used in the new iPad Air 2, but we'd bet on the PowerVR GX6650 6-cluster graphics chip, an upgrade from the 4-cluster GX6450 on the iPhone 6 and 6 Plus. Let’s keep in mind, though, that all of these are preliminary benchmarks that we have not yet been able to run on retail devices. 

With this disclaimer, what do you think: does the Apple A8X and iPad Air 2 live up to your expectations?

Thanks for the tip, TylerGrunter!

Related phones

iPad Air 2
  • Display 9.7" 1536 x 2048 pixels
  • Camera 8 MP / 1.2 MP front
  • Processor Apple A8X, Triple-core, 1500 MHz
  • Storage 128 GB
  • Battery 7340 mAh

FEATURED VIDEO

90 Comments

1. iZman48

Posts: 8; Member since: Jul 17, 2014

Before a android vs Apple war starts have to say kudos to Apple for bringing in a great tablet with awesome benchmarks better than the nexus 9 in multi core performance ...and I thought that was a beast. These tablets are great.apple should have made their phones have a tri-core 2gb ram behomath as well

49. slh163

Posts: 140; Member since: May 19, 2014

so now the multi core score soddenly become more important than the single core after years and years where apple always been beated in multi-core score. but now when the results been inversed so the iStandards should inversed too. but still the K1 gpu performance(wich is the most important for us) still imbeatable :)

50. rantao333

Posts: 346; Member since: May 21, 2013

so suddenly k1 single core performance become so important now for fandroid?

58. slh163

Posts: 140; Member since: May 19, 2014

did i say this?

61. MySchizoBuddy

Posts: 159; Member since: Aug 23, 2011

look at the benchmark again. the old ipad air is beating nexus 7. Apple has always produced a product with less core and with similar or better performance than high core android phone. This isn't a new thing. you are just making it a new thing.

69. jk0907

Posts: 5; Member since: Sep 17, 2014

Good job beating a device that is less than half the price. costs. You must be really proud.

75. Chakra

Posts: 136; Member since: Dec 02, 2009

First and foremost, I'm neither an Apple or Android Loyalist. I'm a tech loyalist, and buy whatever I think fits my needs best. With that said, I completely agree with you. There's are countless threads in which apple fans harped on the "Single Thread is where is counts the most", And now it's not owning the Single thread benchmarks, they flip-flop and say it's the multi-thread that counts the most. My first thought when reading this article is, "I can see it now, the whole single vs. multi threads argument is going to flip-flop." and judging by the responses, that's exactly what happened. The main ingredient for fanboys is the lack of objective, and rational argument - y'all twist the context, and select only certain facts (and it's not even the same facts from generation to generation) that will help your argument. Go ahead and argue with me regarding hardware,you'll only prove my point as I'm not advocating on apple vs android. I'm just making an observation how you fanboys react to these types of "news".

52. LebronJamesFanboy

Posts: 671; Member since: Mar 23, 2013

Let me preface this by stating that I own a nexus 5.... Why are people trying to downplay the iPad 2 Air's performance? Android lovers, including myself, have always found multicore benchmarks to be more accurate. Let's be honest here. The iPad Air 2 wins this benchmark war. But oh, I forgot. It's made my Apple so you MUST hate it. Hate Apple all you want, but this tablet is looking like one of the best if not the best out there. I love how people are trying to bring up every excuse in the book to try to make the N9 look better.

57. Cicero

Posts: 1136; Member since: Jan 22, 2014

iPad Air2 have the best power procesor. That is a fact now. All my respect for this and now I'm waiting for Snapdragon/Exynos move. Until than I'm saying is pity to loose this power only to move static ios.

84. vincelongman

Posts: 5720; Member since: Feb 10, 2013

Can't say that right since the K1's result is for the 32-bit version of Geekbench From the Nexus 9 "Geekbench 3.2.2 for Android AArch32" From the iPad Air 2 "Geekbench 3.2.2 for iOS AArch64" And from the iPhone 5S before the Geekbench update "Geekbench 3.1.6 for iOS ARMv7s (32-bit)" Apparently the 32-bit A7 only got ~1100 single and ~2000 multi, which is far less than its 64-bit result which is ~1400 single and ~2500 multi (though this is only one result, may actually to a one of bad run) None the less, we have to wait and see the 64-bit result of the Nexus 9 before a proper comparison can be made Credit to spicy for having the 32-bit iPhone 5S resultshttp://browser.primatelabs.com/geekbench3/1067175http://browser.primatelabs.com/geekbench3/695393

60. slh163

Posts: 140; Member since: May 19, 2014

you know the processor contain a CPU and a GPU. those geekbench scores are just measuring the CPU. so yeah the A8X cpu is better BUT the K1 GPU have better benchmark score than A8X so K1 wins in graphics performance wich is the more important for gaming. so for me the K1 still the winner here :) peace out.

64. vinnyb4202

Posts: 23; Member since: Sep 18, 2014

This isn't the case though for multiple reasons. 1. Depends on the title and what its dependancies are. Some applications make better use of the CPU cores then it does the GPU, or vise versa. Any apps made for Tegra are going to target CUDA, so in that case that will be better. 2. Two different API's being used, metal vs OpenGL. 3. Tegra K1 actually has full CUDA cores you see in their computer GPU's, of course it's going to be faster in that aspect!

72. slh163

Posts: 140; Member since: May 19, 2014

no it is the case every game title will depend on your gpu not your cpu. but for other apps yes maybe..

89. SteveS

Posts: 51; Member since: Oct 15, 2013

Actually, no... the K1 won on this isolated benchmark. Go to gfxbench.com and pull up scores for the Air 2 and the Nexus 9. The Air 2 wins on a range of benchmarks that cover both on screen and off screen performance. What's more, if you're speaking of gaming performance, you should know that the major gaming engines have been updated on iOS for Apple's Metal API. For games, that will provide a dramatic advantage for the A8X.

53. janno

Posts: 144; Member since: Aug 19, 2014

It's funny that when every single Android chip out there beat Apple's chips in multi-core performance, nobody seemed to care and say that's irrelevant. What matters is single-threaded performance! Now that Apple can ONLY beat some Android chips in multi-core performance (probably won't beat Cortex A57-based ones, which come in quad-core), all of the sudden it doesn't matter as much that Apple gets beaten in single-core performance and GPU performance... This just goes to show that Apple fans will just look for whatever angle to keep telling themselves that their devices are the "best". " Android beats Apple in market share, UI, apps, revenue and whatever else - doesn't matter. Apple still beats them in PROFITS! MUAHAHA - We are the best!"

66. VHMP01

Posts: 93; Member since: Aug 27, 2014

The thing is that Apple beating in 'Profits' is the worst part for blinded iFans! Yet they rely on this silly excuse! You know, more 'Profits' means cheaper materials and hardware at a more expensive Price! Duh for iFans!

86. Neo_Huang

Posts: 1067; Member since: Dec 06, 2013

Its just too bad that iOS doesn't allow for multitasking.

87. JunitoNH

Posts: 1946; Member since: Feb 15, 2012

Is too bad Google makes more money from iOS than Android. So much so, Google spend more time polishing the apps for iOS e.g. Google drive, it even comes with a native pass code for privacy.

2. Sniggly

Posts: 7305; Member since: Dec 05, 2009

Well, holy shiznit, a mobile Apple product finally got 2 GB of RAM!

67. VHMP01

Posts: 93; Member since: Aug 27, 2014

Yes finally, welcome iFans to 2012! LG Optimus LTE2 breaks cover: the first phone with 2GB of RAM... http://www.phonearena.com/news/LG-Optimus-LTE2-breaks-cover-the-first-phone-with-2GB-of-RAM_id29787

73. chokkest

Posts: 1; Member since: Oct 22, 2014

Yeah yeah 2gb ram, welcome to 2012, heard all about it.. But still beats every device released in 2014 in all round performance

81. iushnt

Posts: 3122; Member since: Feb 06, 2013

And in functionality, any good 2012 android device will beat it

85. BattleBrat

Posts: 1476; Member since: Oct 26, 2011

IPhones are not smooth and fluidlike as applefans claim, get a Note 3 with nova on it and compare it to An iPhone 6+, the note is smoother.

3. RebelwithoutaClue unregistered

Too quote some Apple users.. it's all about single core performance, multi core speeds don't matter, more cores are not needed, I say fine piece of engineering once again :)

11. vincelongman

Posts: 5720; Member since: Feb 10, 2013

Also we are comparing Geekbench 64-bit on the iPad and Geekbench 32-bit on the Nexus 9, which gives the iPad a unfair advantage And the Nexus 9 still gets ~1900, the highest single core score Though credit to Apple, the A8X is still seriously impressive and now has the best multicore Except maybe when the Note 4 (Samsung's 8433) gets Android L, currently gets ~4200 while 32-bit

19. spicy

Posts: 9; Member since: Oct 22, 2014

If you actually look closer to the benchmark comparison between the two you will notce some interesting facts: - Integer performance is almost identical (tegra on 2 cores, vs A8X on 3 cores). - Within the integer performance, encryption algorithms are 2 to 3 times faster on the Tegra - On the other integer tests, the only places where the A8X wins are smaller or equal to the margin of the extra core. - The Tegra looses heavily in Floating point benchmarks. As the Geekbench on the Nexus is compiled to ARMv7, these results might change heavily with a 64-bit compiled one (e.g. the floating point handicap)

24. spicy

Posts: 9; Member since: Oct 22, 2014

EDIT: To make a comparison to 32 and 64-bit: Geekbench lists results from the iphone 5s with the 32-bit and 64-bit version. It seems the 64-bit one is 40% faster (1800 multicore on 32-bit vs 2500 muticore in 64-bit) So things can still improve for the tegra

31. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

Sorry, but your calculation is completelly off... The iphone 5s has a 2250 multicore score in 32 bits:http://browser.primatelabs.com/geekbench2/2498324 And a 2500 in 60 bits:http://browser.primatelabs.com/geekbench3/79332 Same goes for the iphone 6:http://browser.primatelabs.com/geekbench2/2498746http://browser.primatelabs.com/geekbench3/884499 So that's what I would expect as an improvement when the 64bits version is out: around 10%, not 40% That would mean Tegra K1 could get somewhere over 2000 in single core, but that's about it.

34. vincelongman

Posts: 5720; Member since: Feb 10, 2013

"Geekbench 2.4.3 for iOS" That's not Geekbench 3 Geekbench 2 is a their older benchmark, Which is why there's one score instead of single and multi Though I agree 40% sounds too high I think the difference would be 10-20%

35. spicy

Posts: 9; Member since: Oct 22, 2014

the first link you gave of 2250 is not from the 5s. In you last links, you also show two different models.

Latest Stories

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.