x PhoneArena is hiring! Reviewer in the USA
  • Hidden picShow menu
  • Home
  • News
  • Apple iPad Air 2 GPU benchmarks are out: graphics neck and neck with Nexus 9's Tegra K1

Apple iPad Air 2 GPU benchmarks are out: graphics neck and neck with Nexus 9's Tegra K1

Apple iPad Air 2 GPU benchmarks are out: graphics neck and neck with Nexus 9's Tegra K1

After seeing the first preliminary results from the iPad Air 2 testing revealing that the Apple A8X uses a peculiar, tri-core CPU, today the company’s new tablet appeared on GFX Bench, showing off its capabilities as a gaming device.

Silicon battles have definitely heated up this fall as Apple’s A8X arrives with a whopping 3 billion transistors, and chances are that it is also one of the first to come with Imagination Technologies’ new PowerVR GX6650 GPU with six clusters (up from the four-cluster GX6450 on the A7).

The biggest rival on the GPU and gaming front on the Android side of the fence is the Tegra K1 chip with Kepler graphics, a chip used in the upcoming Google Nexus 9. Nvidia uses a different approach to graphics with Kepler so there is no use comparing the small, 192 shader cores directly with the large, hexa-cluster setup on the Img Tec’s GPU, but benchmarks do testify that performance-wise, the two are neck to neck. Take a look below:

Manhattan on-screen

*iPad Air 2, Nexus 9 results are preliminary

If the T-Rex HD component of GFXBench is demanding, then the Manhattan test is downright gruelling. It's a GPU-centric test that simulates an extremely graphically intensive gaming environment that is meant to push the GPU to the max. that simulates a graphically-intensive gaming environment on the screen. The results achieved are measured in frames per second, with more frames being better.

score
Higher is better
Apple iPhone 6
25.8
Apple iPad Air 2
24.1
Google Nexus 9
22
Samsung Galaxy Note 4
11.2
Apple iPad Air
9.1
View all

Manhattan off-screen

*iPad Air 2, Nexus 9 results are preliminary

score
Higher is better
Apple iPad Air 2
32.4
Google Nexus 9
29.9
Samsung Galaxy Note 4
18.5
Apple iPhone 6
17.8
Apple iPad Air
13.1
View all

T-Rex on-screen

*iPad Air 2, Nexus 9 results are preliminary

The T-Rex HD component of GFXBench is a demanding, GPU-centric test that simulates a graphically-intensive gaming environment on the screen. The results achieved are measured in frames per second, with more frames being better.

score
Higher is better
Apple iPad Air 2
52.2
Apple iPhone 6
48.9
Google Nexus 9
41
Samsung Galaxy Note 4
25.9
Apple iPad Air
23.2
View all

T-Rex off-screen

*iPad Air 2, Nexus 9 results are preliminary

score
Higher is better
Apple iPad Air 2
70.4
Google Nexus 9
55
Apple iPhone 6
42.3
Apple iPad Air
28.7
View all


Let’s also not forget that real-life performance is also dependant on software and here come APIs like Apple’s new Metal for graphics that promises to be up to 10x faster, and Nvidia’s CUDA. For all it’s worth, these two new chips are certainly setting a new standard for mobile graphics.


source: GFXBench via GSM Arena

83 Comments
  • Options
    Close




posted on 23 Oct 2014, 08:30 14

1. darkvadervip (Posts: 365; Member since: 08 Dec 2010)


Oh the truth has came out! I give congrats to both but y'all going to hate on apple anyway

posted on 23 Oct 2014, 11:10 3

46. bendgate (unregistered)


I hate Apple but it did impressed me.

posted on 23 Oct 2014, 13:33 5

57. brrunopt (Posts: 742; Member since: 15 Aug 2013)


Has it ???

http://www.futuremark.com/hardware/mobileice_storm_unlimited/filter/iosandroid

nVidea Shield Tablet (32 bit version) = 30753 points
iPad Air 2 = 21509 points

posted on 23 Oct 2014, 14:52 1

62. Ninetysix (Posts: 2340; Member since: 08 Oct 2012)


Hi. Read and understand.

http://www.futuremark.com/pressreleases/understanding-3dmark-results-from-the-apple-iphone-5s-and-ipad-air

posted on 23 Oct 2014, 15:35 6

67. brrunopt (Posts: 742; Member since: 15 Aug 2013)


The A7's CPU can significantly outperform the A6 for other tasks with less complex data structures. However, we believe that benchmarking with complex tests that mirror real-world scenarios is more useful than measuring performance with simplistic low-level tests.
Using a modified version of the Bullet Physics Library that optimized the data structures resulted in a 17% improvement in performance on the iPhone 5s, no change on the iPhone 5, and a slight decrease for some Qualcomm powered Android devices.

-----------------------------

so basically you can get up to 17% more by optimizing it for the A7 ...
witch will be a very small part , i'm not seeing developers rewriting their games to get that increase on just 1 chip; in real world it wont really matter

But even graphics scores alone , still losses... :
- nVidea Shield Tablet (32 bit version)
Graphics test 1 - 209 FPS
Graphics test 2 - 126 FPS
- iPad Air 2 = 21509 points
Graphics test 1- 147 FPS
Graphics test 2- 127 FPS

posted on 23 Oct 2014, 15:53

68. Altsein (Posts: 12; Member since: 23 Oct 2014)


The issue with IOS 3Dmark benchmarks is the physics portion of the test. Good thing we can look at things separately and exclude the physics portion from the test.

K1 still manages to outperform the A8X in Graphics score.

posted on 23 Oct 2014, 17:18 1

74. rangoj (Posts: 46; Member since: 29 Mar 2014)


This test needs to be optimized for CUDA core for real results - iPad on his side have really good optimization! I hope they will optimize benchmark for CUDA cores - and them, make test again! :)

posted on 23 Oct 2014, 08:33 17

2. Pattyface (Posts: 1584; Member since: 20 Aug 2014)


Wow I'm pretty impressed actually! Can't wait to hear the people that said it would get crushed by k1

posted on 23 Oct 2014, 09:03 6

15. Finalflash (Posts: 2867; Member since: 23 Jul 2013)


Don't know who said that but even this is amazing for both. They should compare the image quality in both to see how equal they are as well. Plus, an Android chipset matching a PowerVR on iPads is quite amazing. It really hasn't happened this evenly before and I wouldn't be surprised if the Apple chipset turned out to be much larger than the K1 in this case too. Good times ahead if this is the new set of competitor chips to QC.

posted on 23 Oct 2014, 09:19 1

24. tury694 (Posts: 167; Member since: 25 Aug 2014)


That thing is a monstrosity. Too bad the software is garbage, but Apple is good at embarrassing people when it comes to mobile gpu benchmarks.

posted on 23 Oct 2014, 09:31 2

27. jcdz89 (Posts: 78; Member since: 01 Mar 2013)


Wow impressive. . But gsmarena article mentioned about nvdia sheild tab not N9. . ..

posted on 23 Oct 2014, 11:43 1

51. madmikepr (Posts: 136; Member since: 09 Aug 2011)


That's Right even GsmArena Says
"We're eagerly waiting to see how much better the 64-bit Tegra K1 will be than the 32-bit one in the Shield Tablet and Mi Pad 7.9. As it happens GFXBench lists a Nexus 9 unit (the HTC-made Google tablet with the 64-bit Tegra K1) with scores in both T-Rex and Manhattan but there's nothing in the Info screen to indicate this is legitimately the Nexus 9.

posted on 23 Oct 2014, 08:42 5

3. joseram695 (Posts: 79; Member since: 28 May 2012)


Impressive Apple. Coming from a android enthusiasts. I can't wait to try both the nexus 9 and air 2. I'll probably still stick with a nexus 9 tho, just because I like to be able to cutomize my tablet and not get stuck with the way it comes out the box

posted on 23 Oct 2014, 10:01 1

34. Zomer (Posts: 268; Member since: 31 May 2013)


I always hear Android peeps talking about how you can customize everything, but they never seem to explain what they want to change?

posted on 23 Oct 2014, 10:44 2

41. CreeDiddy (Posts: 667; Member since: 04 Nov 2011)


Exactly!!! Homescreen and widgets. Oh and maybe a different launcher. LOL!!! Wasted time.

posted on 23 Oct 2014, 11:09 5

45. alrightihatepickingusernames (Posts: 474; Member since: 29 Dec 2013)


Doesn't make a big difference to you but Android users (like myself) see grid after grid of squares as pretty ugly.

posted on 23 Oct 2014, 11:00 4

44. Jobayer (Posts: 167; Member since: 22 Feb 2013)


Stock android is boring for me (so is iOS), so I change to custom roms .

posted on 23 Oct 2014, 13:18

55. Joulukas (Posts: 66; Member since: 05 Jun 2013)


My phone is crucial for me fo every day tasks, when i open it up i need tons of information then and there, not open up apps to get after half an hour of browsing my phone. Sadly only phone to give me that option, is sadly android.

posted on 23 Oct 2014, 08:44 6

4. Jason2k13 (Posts: 1213; Member since: 28 Mar 2013)


technology is going so fast, i feel like benchmark results is just used for fanboys to hate on each other and other brands. I dont think anybody will see the difference with real life performance on any of those phones and tablets.

posted on 23 Oct 2014, 08:46 3

5. Felix_Gatto (Posts: 900; Member since: 03 Jul 2013)


It's not surprising, the A8X is more advanced in manufacturing process than the K1 (20nm vs 28nm). Nvidia have made a mistake again, they should design the Denver variant of K1 in 20nm.

posted on 23 Oct 2014, 09:04 3

17. Finalflash (Posts: 2867; Member since: 23 Jul 2013)


They will most likely, but the fact that they match it right now with a larger process means the future is looking pretty damn good for the Android crowd.

posted on 23 Oct 2014, 09:13 2

21. Felix_Gatto (Posts: 900; Member since: 03 Jul 2013)


It's not about performance alone but also power consumption, both of A8X CPU & GPU should have less power to draw than K1 while outperform it in perfomance. So, the conclusion is: A8X > K1. Nvidia will never made to defeat Apple as they are still trailing in 28nm.

posted on 23 Oct 2014, 09:56 5

32. TylerGrunter (Posts: 1515; Member since: 16 Feb 2012)


You can see it the other way around: they are neck to neck even when they are one node apart, so the future is very bright for Nvidia and Tegra.

posted on 23 Oct 2014, 10:56 4

42. synot (Posts: 277; Member since: 14 Sep 2012)


This article is posted with wrong information.The test was not done by a tech site but some individual sent in the information and the chip was not the K1 Denver used in the Nexus 9 but the one used in the Xiaomi Mi Pad which is the Nvidia Shield and the chip was also 32 bit and not the 64 bit in the Nexus 9.

posted on 23 Oct 2014, 13:02

53. renz4 (Posts: 317; Member since: 10 Aug 2013)


i don't think that's a mistake for nvidia part. initially denver wasn't supposed to come out until gen 6 tegra (2015 tegra). but somehow nvidia want denver out in developer hands early instead of waiting until 2015. if they wait for 20nm they might not able to win nexus 9 deal.

posted on 23 Oct 2014, 08:47 2

6. UglyFrank (Posts: 1622; Member since: 23 Jan 2014)


NVidia is still trailing as they are on a 28nm process

posted on 23 Oct 2014, 11:00 2

43. synot (Posts: 277; Member since: 14 Sep 2012)


This test was not done by Phone arena but unconfirmed information sent by unknown individual and the chip tested was not the one in the Nexus 9 but Nvidia Shield used in the Xiaomi Mi Pad.

posted on 23 Oct 2014, 08:47 3

7. hortizano (Posts: 291; Member since: 22 May 2013)


Well played Apple. You win this round. Anyhow I can't get an iPad because of my "condition"... My loss I guess..
Long Live Nexus!!

posted on 23 Oct 2014, 09:21

25. Felix_Gatto (Posts: 900; Member since: 03 Jul 2013)


Don't forget, Samsung is still have their trump card : The Exynos 5433. Apple is not yet win. Sadly, Snapdragon have become dominant in Samsung's flagship line since the S5. This is the first time Exynos has become a minority in Samsung's flagship line.

posted on 23 Oct 2014, 09:57 6

33. TylerGrunter (Posts: 1515; Member since: 16 Feb 2012)


Exynos 5433 is worse in GPU and CPU than Tegra K1, just saying. No much hope there.

posted on 24 Oct 2014, 07:11

79. alex3run (Posts: 711; Member since: 18 May 2014)


It is more or less comparable with Apple A8X in terms of CPU but yeh GPU in Exynos cannot compete with Apple.

posted on 23 Oct 2014, 10:03

35. Zomer (Posts: 268; Member since: 31 May 2013)


Apple will simply always win, because they develop the software and hardware.

posted on 23 Oct 2014, 11:12

47. bendgate (unregistered)


Really? They are still good in some things especially CPU but that isn't always.

posted on 23 Oct 2014, 08:49 1

8. byte11 (Posts: 16; Member since: 05 Nov 2013)


This is such a click-bait article. Does PhoneArena not read the articles attached to its sources? The GSMArena article states, "As it happens GFXBench lists a Nexus 9 unit (the HTC-made Google tablet with the 64-bit Tegra K1) with scores in both T-Rex and Manhattan but there's nothing in the Info screen to indicate this is legitimately the Nexus 9."

posted on 23 Oct 2014, 08:54 1

10. Victor.H (Posts: 653; Member since: 27 May 2011)


We do say it's a preliminary result, and we've written in earlier articles that the results are too early to draw final conclusions.

posted on 23 Oct 2014, 09:32 2

28. networkdood (Posts: 6330; Member since: 31 Mar 2010)


Then why do the article? Wait until you get all the info before creating a story...

posted on 23 Oct 2014, 22:20

78. Kaiser2007 (Posts: 173; Member since: 16 Nov 2013)


Victor. You twisted it. GSMArena article is about Xiaomi Mi Pad 7.9. It didn't say anything about Nexus 9.

posted on 23 Oct 2014, 08:58

12. Felix_Gatto (Posts: 900; Member since: 03 Jul 2013)


They already tested the 32-bit K1 and still behind A8X. l think the GPU performance of both 32-bit and 64-bit K1 should be similar because both of it are still manufactured in 28nm.

posted on 23 Oct 2014, 08:53 1

9. datsyuk (Posts: 179; Member since: 11 Jan 2012)


PowerVR GPUs are monsters

posted on 23 Oct 2014, 08:55 2

11. janno (Posts: 144; Member since: 19 Aug 2014)


Shame Nvidia always seems to lag in process technology. They promised the Maxwell-based Tegra will jump straight to 16nm FinFET next year, but Tegra K1 will probably remain at 28nm planar until then.

posted on 23 Oct 2014, 09:06 2

19. Finalflash (Posts: 2867; Member since: 23 Jul 2013)


That is because TSMC is feeding Apple most of their chips at 20nm which is also a half node jump. 16nm should provide better results and I think Samsung will be making those as well as Global foundries, so that should make next year very interesting.

posted on 23 Oct 2014, 13:10

54. renz4 (Posts: 317; Member since: 10 Aug 2013)


so erista still expected to use 16nm finfet? i only remember nvidia old roadmap mentioning parker using 16nm finfet. but the roadmap has change since. parker has been moved further down tegra roadmap and for 2015 nvidia will come out with erista instead of parker. i thought erista will be 20nm and parker will use 16nm finfet. does TSMC 16nm finfet will be ready if nvidia going launch erista early next year? (most likely at CES)

posted on 23 Oct 2014, 08:58 2

13. Fuego84 (Posts: 342; Member since: 13 May 2012)


Android only user here, congrats Apple! now if you'd only stop being a closed platform and become just as open and free as Google I could consider using/buying one of your many device but until then only android for me!

posted on 23 Oct 2014, 09:02 1

14. jove39 (Posts: 1838; Member since: 18 Oct 2011)


PowerVR knows how to push pixels...both gpu have same core count but gx6650 trumps k1 in Pixels/Clock and Texels/Clock.

http://www.anandtech.com/show/7796/imagination-technologies-announces-new-high-end-mobile-gpu-powervr-gx6650

This round goes to gx6650...let's hope nvidia's maxwell core performs better than gx6650...and competition continues......................

posted on 23 Oct 2014, 09:27 1

26. Felix_Gatto (Posts: 900; Member since: 03 Jul 2013)


At the time Nvidia's Maxwell went to mobile, they no longer have to compete with GX6650. Perhaps it would be G7XXX series.

posted on 23 Oct 2014, 10:03 2

36. TylerGrunter (Posts: 1515; Member since: 16 Feb 2012)


You forget that K1 has been out for 6 months already, so I would expect Maxwell to go mobile around March next year. That will give it 6 months competing with GX6650 (in iPad air) and with the even older and much weaker G6430 (in iPad mini) that even the current K1 crushes.
In other words: it's normal that the new chip outperforms the old one, as jove39 says.

posted on 23 Oct 2014, 09:03 1

16. Andrew5190 (Posts: 49; Member since: 23 Oct 2014)


Very impressive, but it takes three cores for them to match up to the Nexus 9 (if this is the Nexus 9 at all, as byte11 said). Benchmarks don't impact my decision on a phone or tablet at all.

And if you're going to include the iPhone, add the 6 Plus please. Those extra pixel make a big impact on the benchmark scores between 6 and 6 Plus.

posted on 23 Oct 2014, 09:04 1

18. technosiastic (Posts: 33; Member since: 18 Sep 2014)


A8x is a real beast.
No one can stand in front of ipad air 2.

posted on 23 Oct 2014, 09:11 2

20. technosiastic (Posts: 33; Member since: 18 Sep 2014)


look at note 4 scores ..
hahaha so called beast.

posted on 23 Oct 2014, 09:14

22. HomerS (Posts: 178; Member since: 19 Sep 2014)


Why are there only the Note 4 On Screen results and not off Screen results, as the Note 4 with 1440p has to push alot more pixels on screen than the 1080p devices?

Yes the others are still higher even on Off Screen but they are tablets with less power restrictions.

posted on 23 Oct 2014, 09:17

23. YamiYukiSenpai (Posts: 4; Member since: 24 Sep 2014)


Those are impressive results!

posted on 23 Oct 2014, 09:34 2

29. networkdood (Posts: 6330; Member since: 31 Mar 2010)


Nexus 9 is definitely the better value here...lower price, and more freedom always wins

posted on 23 Oct 2014, 13:45

58. CellPhonesExpert (Posts: 45; Member since: 10 Oct 2014)


LOL-----

posted on 23 Oct 2014, 14:31

60. kwmwtaj (Posts: 1; Member since: 23 Oct 2014)


You're laughing because your lost

posted on 23 Oct 2014, 09:46 3

30. Chuck007 (Posts: 1090; Member since: 02 Mar 2014)


Android fans go Nexus 9 while iOS lovers go iPad Air 2. No need to fight ok guys?

posted on 23 Oct 2014, 09:51

31. trqster (Posts: 97; Member since: 17 May 2013)


Apple as usual shows the way with far less impressive hw theoretical specs... Not to mention the more integrated hw/sw combo that unfortunately will never be matched by Android. BTW both iOS and Android top-tier devices here...

posted on 23 Oct 2014, 10:06

38. TylerGrunter (Posts: 1515; Member since: 16 Feb 2012)


Ignorance must be a bless for you. Where do you exactly see the "far less impresive HW theoretical specs"?
Because the specs of the A8X have been always higher than the ones of the K1, that you (and may Apple fans) know that it's just a question of lack of reading (or reading comprehension)

posted on 23 Oct 2014, 11:40

50. TylerGrunter (Posts: 1515; Member since: 16 Feb 2012)


Correction:
you (and many Apple fans) don´t Know that, it´s a question...

posted on 23 Oct 2014, 10:06

37. stein (Posts: 22; Member since: 17 Oct 2014)


I'm an ipad air owner.
Is good to see many power tablets on the market, i hope it will push developers to make greatest apps and games for both plataforms.

posted on 23 Oct 2014, 10:26 2

39. Peace (Posts: 23; Member since: 22 May 2014)


Some times I wonder what should happen for android lovers to admit something is better on the other side of the fence?
I am not idiot enough (or paid(I'm not mentioning samsung)) to blindly follow a brand and comment like a troll, just admitting that apple won this race at the moment...

Want to comment? Please login or register.

Latest stories