x PhoneArena is hiring! Reviewer in the USA
  • Hidden picShow menu
  • Home
  • News
  • Nexus 9 storms through Geekbench, Tegra K1 outperforms Apple iPhone 6's A8

Nexus 9 storms through Geekbench, Tegra K1 outperforms Apple iPhone 6's A8

Posted: , by Victor H.

Tags :

Nexus 9 storms through Geekbench, Tegra K1 outperforms Apple iPhone 6's A8

The Nexus 9, a device previously referred to as the HTC Volantis, is the first to come with Nvidia's 64-bit Tegra K1, a dual-core chip with two large Nvidia Denver CPU cores, and a Kepler GPU with 192 shader cores.

This sounds good on paper, but we now have the first benchmark scores from the Nexus 9 to see if the K1 lives up to expectations. The benchmark used is Geekbench, a CPU and memory-testing app (it does not test graphics performance).

The score is in, and it smokes the competition. The Nexus 9 with Tegra K1 and Denver CPU has nearly double the single-core performance of even top Android devices like the Galaxy Note 4, and it finally closes the gap with iOS devices, the recent single-core performance champions.

Nexus 9 with Tegra K1 finally matches and outdoes Apple's single-core performance

We cannot do a full apples to apples comparison until we see Apple's new iPads, but since we do expect them to feature a chip similar to the Apple A8 in the iPhone 6 and iPhone 6 Plus, it's safe to assume that the Nexus 9 will be in the same league when it comes to CPU compute.

Nvidia unveiled the Tegra K1 in two version early this year, and the Nexus 9 is the first to arrive with the 64-bit variation of the K1.
Nexus 9 storms through Geekbench, Tegra K1 outperforms Apple iPhone 6's A8
The curious narrative around the Denver CPU core used in it, is that it was originally planned as a desktop-grade solution. Hence, it has a large, 7-issue design with a wide, in-order pipeline. 

Interestingly enough, though, the benchmark lists the Tegra K1 as running at a clock speed of 2.5GHz, while we have seen the Nexus 9 officially unveiled with a 2.3GHz clock speed. This is yet another fact that should remind you to take these preliminary results with some reservations.

You can see the breakdown of the Nexus 9 performance in each of the tests in Geekbench in the slideshow right below, and do keep in mind that those are preliminary results that we'll confirm once we have the device in our hands. For the time being, though, these promising scores show a lot to be excited about.

Thanks for the tip, Tyler and Jawad!

Geekbench single-core

*Nexus 9 results are preliminary

score
Higher is better
Google Nexus 9
1903
Google Nexus 6
1062
LG G3
950
HTC One (M8)
888
Samsung Galaxy S5
944
Sony Xperia Z3
974
Samsung Galaxy Note 4
1112.67
Apple iPhone 6 Plus
1625
Apple iPhone 6
1630
View all

Geekbench multi-core

*Nexus 9 results are preliminary

score
Higher is better
Google Nexus 9
3166
Google Nexus 6
3295
LG G3
2545
HTC One (M8)
2613
Samsung Galaxy S5
2900
Sony Xperia Z3
2660
Samsung Galaxy Note 4
3259.67
Apple iPhone 6 Plus
2918
Apple iPhone 6
2927
View all

source: Primate Labs

86 Comments
  • Options
    Close






posted on 16 Oct 2014, 08:31 21

1. SuperAndroidEvo (Posts: 4857; Member since: 15 Apr 2011)


I knew the Nexus 9's Nvidia's 64-bit Tegra K1 would out perform Apple's A8. Google is killing it & the rest of the world is on notice!

posted on 16 Oct 2014, 08:35 6

3. hassoups (Posts: 473; Member since: 06 Jun 2013)


i'd wait for the A8x first. I doubt any of them will oust the other in day to day use.

posted on 16 Oct 2014, 08:43 6

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


because ? we are talking about CPU , not GPU .. (until now AXx versions only had better GPU's )

posted on 16 Oct 2014, 08:55 2

16. slh163 (Posts: 140; Member since: 19 May 2014)


whaat! say it again.. A8x have better GPU!!
the K1 is known for how great his GPU is with his 192 cores that can even play some consoles games like Serious sam 3 and Trine 2. do you think that A8 or A8X can run these games like K1 do??

posted on 16 Oct 2014, 09:10 2

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


:facepalm:

I'm talking comparatively to non-x version, not comparing to K1 GPU
- > A6x better GPU than A6 , A5X bettter GPU than A5 ...

posted on 16 Oct 2014, 09:12 2

24. TylerGrunter (Posts: 1533; Member since: 16 Feb 2012)


I think you just missundestood what brrunopt was saying. He never meant A8x will have better GPUs than Tegra K1.
He only meant that AXx GPUs only have GPU improvemnts over the AX versions, not CPU ones.
EG: A8X will have a better GPU than A8, but the same CPU.

posted on 16 Oct 2014, 08:48 4

12. SuperAndroidEvo (Posts: 4857; Member since: 15 Apr 2011)


Did you read the article? They are implying that the A8X based on these numbers might not be able to topple the Tegra K1. This is a BEAST processor. Finally taking the single-core performance away from Apple.

Imagine if this had 4+ GB of RAM instead of 2GB...

posted on 16 Oct 2014, 08:35 1

4. Benoit (Posts: 57; Member since: 16 Oct 2014)


A tablet that outperforms a smartphone, good job...
Wait for the comparison with the new iPad before claiming supremacy

posted on 16 Oct 2014, 10:10 4

40. UglyFrank (Posts: 1732; Member since: 23 Jan 2014)


That's a weak first post.

posted on 16 Oct 2014, 11:06 2

53. TechieXP1969 (Posts: 10700; Member since: 25 Sep 2013)


Its still the same hardware whether its in a tablet or phone. The Apple A8 is the phone CPU, the A8X is the tablet version. Same CPU but with a slightly high clock speed...the GPU will liley also be the same but maybe with a slightly higher clock speed. Taking the same chip and making run faster is no big deal.

After all the iPad is just an iPhone without cellular capabilities and a larger display with more pixels. That is why Apple makes the processor slightly better as far as speed. The iPad also has a bigger battery. Increasing the clock, surely increases the drain.

posted on 22 Oct 2014, 01:13 2

88. Rokkun (Posts: 1; Member since: 22 Oct 2014)


You must be feeling pretty silly right now given how powerful the A8x is.

And yes i made an account just to reply to you.

posted on 16 Oct 2014, 21:14 2

75. g2a5b0e (Posts: 3725; Member since: 08 Jun 2012)


Your noob is sticking out. You might want to tuck it back in.

posted on 17 Oct 2014, 07:55

77. SuperAndroidEvo (Posts: 4857; Member since: 15 Apr 2011)


lolololololololololololololololol

+1

posted on 16 Oct 2014, 08:41

8. Victor.H (Posts: 678; Member since: 27 May 2011)


I'm also happy to see some real competition in single-core performance. Let me just say this here as well (I mention it in the article): keep in mind that those are preliminary results. The final Nexus 9 should run at 2.3GHz, while the Geekbench score registers it at 2.5GHz, so there might be some slight variation.

posted on 16 Oct 2014, 08:51 3

13. SuperAndroidEvo (Posts: 4857; Member since: 15 Apr 2011)


Competition is always good. Apple is already modding the A8 & turning it into the A8X. Imagine what a modded Tegra K1 will do...

posted on 16 Oct 2014, 09:12

25. Iodine (Posts: 1351; Member since: 19 Jun 2014)


If there would be an A8X it will definitelly not be a derrivate of A8, but a SoC developed in paralel with it, there is no "modding" in there.

If we take the A6 and A6X for example, the only part that was the same were CPU cores.

posted on 16 Oct 2014, 09:42 1

35. SuperAndroidEvo (Posts: 4857; Member since: 15 Apr 2011)


Thus modding cause they share something. It's NOT a 100% completely different unit.

posted on 16 Oct 2014, 09:13 1

28. sorcio46 (Posts: 435; Member since: 27 Jul 2011)


Geekbench on Android shows the max frequency read in the kernel, that can be different from the max frequency of the OS.

Or that 2.5GHz is a turbo boost used only by the SoC in single threaded tasks, but in dual core mode it runs @ 2.3GHz

posted on 16 Oct 2014, 09:14

29. TylerGrunter (Posts: 1533; Member since: 16 Feb 2012)


We will see that when it comes, but the preliminary results are promising.
By the way, are you sure Denver is 9-issue or is it a typo?
I always read it is a 6-issue (6-fe wide) core.

Ah, and remember that is is geekbenk for 32bits, so some performance increase is expected when the 64bits version is out (even I doubt it will be more than a 5-8%)

posted on 16 Oct 2014, 09:35 1

34. RebelwithoutaClue (Posts: 3037; Member since: 05 Apr 2013)


Is the 9-issue related to the number of instructions per cycle it can do? Since the Denver K1 can do 7 instructions per cycle (the Apple A8 6 and Snapdragon 810 can only do 3)

posted on 16 Oct 2014, 10:01 1

37. TylerGrunter (Posts: 1533; Member since: 16 Feb 2012)


Is related to the number of paths an instruction can take, so how many instructions could be issued "in parallel" (6-issue wide)
So Denver should be 6 or 7 (I read both, 7 at first and later 6), instead of 9. That's why I'm asking if it's a typo or he has more information about the internals of Denver.
The intructions per cyle (IPC) relates to how good the OoO engine is dispatching instructios, not only about how wide the core is.
In other words, the width tells you some "theoretical" maximum IPC under perfect conditions (that are never to be met).
More or less, as I'm not 100% sure about what I'm writting ;D

posted on 16 Oct 2014, 10:15 1

42. RebelwithoutaClue (Posts: 3037; Member since: 05 Apr 2013)


Than it's 7, since it's a 7 way superscalar cpu :)
Dunno how he came up with 9 though. That would be stellar.

lol I know the feeling, have that all the time lol

posted on 16 Oct 2014, 12:26 1

64. TylerGrunter (Posts: 1533; Member since: 16 Feb 2012)


Fully on target, it´s seven. I must have mixed it with something else.
I would have been wonderfull to have a 9-wide one though, performance would have gone through the roof!
One word about it: they are not ARM cores, but transmeta, so in reality the performance is about half. Let´s say that is is the equivalent of a 3.5 wide ARM core with 64 bits.
Which taking into perspective that A57 is still 3-issue, 3-dispatch wide is great news.

posted on 16 Oct 2014, 10:16 1

43. Napalm_3nema (Posts: 2227; Member since: 14 Jun 2013)


6 per cycle for the A7/A8, I thought. I might be wrong.

posted on 16 Oct 2014, 10:24

48. RebelwithoutaClue (Posts: 3037; Member since: 05 Apr 2013)


Yup, the A7/8 has 6 per cycle, double the amount of the standard ArmV8 guidelines.

posted on 16 Oct 2014, 10:19

46. Victor.H (Posts: 678; Member since: 27 May 2011)


Typo! I must have been thinking about the Nexus 9... It's 7.

posted on 17 Oct 2014, 16:06

85. Neo_Huang (Posts: 1067; Member since: 06 Dec 2013)


The XiaoMi MiPad also has a Denver Tegra K1. You may want to test that.

posted on 16 Oct 2014, 08:44

11. masi0 (Posts: 4; Member since: 19 May 2012)


note apple is dual core 1.4ghz CPU - this Chihuahua is mad as pitbull :)

posted on 16 Oct 2014, 09:13 1

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


and that means exaclty what ?
until you measure the power consumption of booth you you cant make any assessment about witch one is more power efficient...

posted on 16 Oct 2014, 10:18

45. Napalm_3nema (Posts: 2227; Member since: 14 Jun 2013)


We already know the K1 is not cool running or power efficient. If it was, the Nexus 6 would be 64-bit K1 instead of 32-bit SD805.

Want to comment? Please login or register.

Latest stories