x PhoneArena is hiring! Reviewer in the USA
  • Hidden picShow menu
  • Home
  • News
  • Android storage speed comparison: which phone has the fastest IO performance?

Android storage speed comparison: which phone has the fastest IO performance?

Posted: , by Victor H.

Tags:

Android storage speed comparison: which phone has the fastest IO performance?

One of the most underreported smartphone specifications is the speed of the internal NAND storage. This, however, remains a key factor for a smooth, stutter-free performance on Android, and this became painfully obvious a couple of years ago when the slow internal storage of the original Nexus 7 tablet and the lack of TRIM support caused huge bottlenecks. The end result was many unhappy users that felt their device was too slow for comfortable use.

Following Android versions starting with Android 4.2 Jelly Bean added software fixes, but hardware is also an important part of the speed of a phone’s internal storage.

In order to have a perfectly smooth experience on Android, the IO performance is key, and today we take a look at the most popular Android flagships and how they compare in terms of NAND performance. We've used the awesome AndroBench tool, a free app on the Google Play store.

We test sequential and random speeds

Testing is divided in two categories: sequential speeds and random reads. Substantial speeds happen when big files are written or read to the storage, things like a movie, which is stored in ‘sequential’ blocks of the memory. The faster the result here, the faster you’d be able to move around larger files in the system.

The second and more important factor for the smooth Android experience are random speeds, which measure the speed of reads and writes in different blocks of the NAND storage. Random writes and reads is exactly what happens when apps update in the background (many small bits of data written and read), as well as when you are multitasking. Greater random speeds usually translate in fewer dropped frames.

The IO performance speeds below are all in MB/s, and the higher the values, the faster a phone’s internal storage is. Take a look at the fastest Android smartphones right below, and don’t forget that our benchmark section is where you can find these and many other measurements.

AndroBench random read speeds (4KB files, in MB/s)

Androbench - Random read
Higher is better
Samsung Galaxy Note 4 20.56
LG G3 15.71
Sony Xperia Z3 14.78
Motorola Moto X (2014) 14.32
Huawei Ascend Mate7 14.04
Samsung Galaxy S5 13.7
HTC One (M8) 13.13
Google Nexus 5 11.09
Google Nexus 6 7.93
Androbench - Random write
Higher is better
Samsung Galaxy Note 4 3.23
LG G3 1.46
Sony Xperia Z3 9.26
Motorola Moto X (2014) 2.51
Huawei Ascend Mate7 1.3
Samsung Galaxy S5 1.79
HTC One (M8) 1.29
Google Nexus 5 0.88
Google Nexus 6 1.54
View all


AndroBench sequantial read speeds (256KB files, in MB/s)

Androbench - Sequential read
Higher is better
Sony Xperia Z3 217.17
Samsung Galaxy S5 176.5
LG G3 164.08
Motorola Moto X (2014) 160.1
Samsung Galaxy Note 4 147.05
HTC One (M8) 90.65
Google Nexus 5 88.4
Huawei Ascend Mate7 64.01
Google Nexus 6 31.62
Androbench - Sequential write
Higher is better
Sony Xperia Z3 44.43
Samsung Galaxy S5 24.2
LG G3 22.29
Motorola Moto X (2014) 31.2
Samsung Galaxy Note 4 34.26
HTC One (M8) 24.4
Google Nexus 5 17.23
Huawei Ascend Mate7 14.65
Google Nexus 6 22.4
View all


60 Comments
  • Options
    Close




posted on 04 Feb 2015, 09:26 22

1. boosook (Posts: 1419; Member since: 19 Nov 2012)


Ok, let me understand... the Z3 is three times faster than the Note 4 in random write and the winner both in reading and writing in Androbench, and which parameter do you choose to order the phones? The only one where the winner is a Samsung...

posted on 04 Feb 2015, 09:51 9

13. Neros (Posts: 1016; Member since: 19 Dec 2014)


Z3 clearly cheated...

Go sammy :D! Note 4 forthe win!

posted on 04 Feb 2015, 13:56 12

41. bugsbunny00 (Posts: 1989; Member since: 07 Jun 2013)


sammy is the one caught cheated.

posted on 04 Feb 2015, 16:04 4

47. cheetah2k (Posts: 1507; Member since: 16 Jan 2011)


How did the Z3 cheat? looks to me like a clear winner. Note 4 sucks on random write which is where it matters most especially when taking those 20MP pics :)

posted on 04 Feb 2015, 18:06 1

49. gigaraga (Posts: 1454; Member since: 29 Mar 2013)


If the Note 4 'sucks' on random write test, then Huawei and Nexus 6 must be beyond trash at the test then.

posted on 04 Feb 2015, 19:32 1

50. joey_sfb (Posts: 5339; Member since: 29 Mar 2012)


How do you know Z3 cheated? Really... all sort of accusation get pick out of thin air.

Good to know the internal storage speed of phone as it will affect overall performance of the device.

Hope to see PA make these a regular update for new devices.

posted on 04 Feb 2015, 12:15 10

31. romeo1 (Posts: 705; Member since: 06 Jan 2012)


You see this kind of weird science more than often on pa and other sites.
Everybody who can read will see that z3 is the winner here

posted on 04 Feb 2015, 15:26 8

46. xperian (Posts: 292; Member since: 10 Apr 2014)


But they still put note first so fanboys won't get angry

posted on 05 Feb 2015, 16:38

60. romeo1 (Posts: 705; Member since: 06 Jan 2012)


Yes but now you could get mad sony fanboys but they could probably live better with that as there aren't as many or they just don't get paid or paid less by sony

posted on 04 Feb 2015, 12:42 2

35. TechieXP1969 (Posts: 8353; Member since: 25 Sep 2013)


Which of the things is your phone going to perform most of? The phone performs more random reads vs writes. Correct?

What would you phone randomly be writing all the time?

Read speeds are what need to be the fastest, not writes.

Everything you do on a phone requires reads. The only time the devices writes is when you download files and an application writes some data like saving app states for example.

But writing should be slower to allow for faster reads.

I mean isn't this have motorized drives work?

When Windows is done with an application, writes are fast because Windows simply drops the data to the next empty space that passes the head to help increase the speed. But over time this slows the reads,because data for an app is everywhere, which is why you must defrag.

With Flash, the data is stored in blocks where the data is pretty much kept in the same place. So since the data is basically going where it came from, writing isn't as important as when it is time to load the data which needs to be fast since Flash doesn't have moving parts.

SO the winner for random reads is correct because your device will perform more random reads than writes.

Hopefully I am understanding it correctly as well.

posted on 04 Feb 2015, 14:06

43. brunelian92 (Posts: 79; Member since: 04 Feb 2015)


It mentions in the article that they are ordered based on random read speed. Considering this is the most likely function you will use it seems perfectly appropriate to do so.

posted on 04 Feb 2015, 09:26 2

2. redmd (Posts: 1239; Member since: 26 Oct 2011)


Benchmarks and actual user experience is entirely different.

posted on 04 Feb 2015, 09:35 17

8. terabyteRouser (Posts: 457; Member since: 18 Oct 2011)


Benchmarks provide excellent indicators into what the limitations of that experience will be.

posted on 04 Feb 2015, 09:43 2

12. rd_nest (Posts: 1591; Member since: 06 Jun 2010)


In case of storage performance, benchmarks actually provide user experience.

But you need to look into relevant benchmarks to make some conclusion. Also after certain point, further increase in performance is not likely to give you significantly visible advantage

posted on 04 Feb 2015, 10:05

16. vuyonc (Posts: 987; Member since: 24 Feb 2014)


iPhones have high single core performance in benchmarks and it shows in UI fluidity. Tegra processors score high in graphic benchmarks and play Half Life 2 at 1080p between 40-60fps. Benchmarks aren't obviously perfect but some are good enough. Any complaints from Xperia Z3 users?

posted on 04 Feb 2015, 10:31

24. rd_nest (Posts: 1591; Member since: 06 Jun 2010)


Again, you don't grasp the difference. For storage, this is not same. I make a living out of consulting with storage industry. Storage benchmarks is not similar to Antutu where you attach a empirical value to particular test. In storage, your experience is directly related to numbers. It doesn't mean every benchmark will tell you something. Looking into relevant statistics is an art that not many people understand. A simple question like whether I should look into IOPS, Bandwidth or queue depth should be enough to grasp which number is more important than another. It's again waste if you run SQLIO on a mobile device, waste of time..

posted on 04 Feb 2015, 09:26 8

3. Anshulonweb (Posts: 430; Member since: 07 Feb 2014)


well nexus 6 results were a disappointment...... not sure but may be due to default encryption on lollipop ...

posted on 04 Feb 2015, 09:41 6

11. rd_nest (Posts: 1591; Member since: 06 Jun 2010)


Check Anandtech article on encryption and store performance on 5.0

posted on 04 Feb 2015, 14:05 1

42. vincelongman (Posts: 4104; Member since: 10 Feb 2013)


Yep its the encryption
Basically Moto some how forgot about encryption and didn't add anything to hardware accelerate the encryption

E.g. the Nexus 9 uses its ARMv8 based K1 for hardware acceleration
Hence why AnandTech said the IO speeds were the same for the Nexus 9 with and without encryption

posted on 04 Feb 2015, 10:02 1

14. boosook (Posts: 1419; Member since: 19 Nov 2012)


Yes, it's definitely due to encryption IMO.

posted on 04 Feb 2015, 09:28 4

4. TyrionLannister (unregistered)


Z3 and Note 4 are owning it.

posted on 04 Feb 2015, 09:29 12

5. RebelwithoutaClue (Posts: 2702; Member since: 05 Apr 2013)


Keep in mind the Nexus 6 has encryption default on, all files will be encrypted on the local storage. And it already has been proved this makes the Nexus 6 a lot slower than a device without encryption. So the internal storage might still be a lot faster if encryption is turned off.
If you want to make a real comparison, turn on encryption on all the other devices too.

posted on 04 Feb 2015, 09:35 7

7. Victor.H (Posts: 651; Member since: 27 May 2011)


Good point.

posted on 04 Feb 2015, 10:19 2

20. FlyingDutch (banned) (Posts: 97; Member since: 30 Jan 2015)


Benchmarks should be performed with default settings since defaults are defaults for good reasons.

posted on 04 Feb 2015, 13:29 1

40. RebelwithoutaClue (Posts: 2702; Member since: 05 Apr 2013)


They should be benchmark-ed with defaults,but an annotation would be in place that with the default setting, the Nexus 6 does offer encryption while the others don't.

posted on 04 Feb 2015, 10:04 2

15. boosook (Posts: 1419; Member since: 19 Nov 2012)


Well, I think it makes sense to use the default configuration, as it will reflect the user experience (you cannot turn encryption off on the N6).
However, it's fair to remark that on the N6 you have the advantage of a greater security.

posted on 04 Feb 2015, 10:24

23. Downtime24g (Posts: 24; Member since: 04 Aug 2011)


rooted my own nexus 6 and the r/w speed is noticably faster, worth performing the test again atleast

posted on 04 Feb 2015, 09:31 1

6. polybius (Posts: 32; Member since: 03 Feb 2015)


OEMs why you no put flash storage in flagship phones?

posted on 04 Feb 2015, 10:12

18. wargreymon (Posts: 671; Member since: 05 Nov 2013)


You might wanna read up on what eMMC is.

posted on 04 Feb 2015, 10:20

21. polybius (Posts: 32; Member since: 03 Feb 2015)


Multimedia cards are not bad, but they are still nowhere as fast as regular SSDs. If people are willing to spend money on a phone with the best possible specs, why not? I should have been more specific tho.

posted on 04 Feb 2015, 10:50

26. wargreymon (Posts: 671; Member since: 05 Nov 2013)


If I dont remember wrongly, SSDs take up way too much space and drains too much power to be usefull in a phone. Plus it costs more. eMMC are getting fast though, its not like the old days and Im pretty sure eMMC 5.0 kicks most of the first generations of SSDs.

posted on 04 Feb 2015, 11:23

28. polybius (Posts: 32; Member since: 03 Feb 2015)


Still, given the ever-increasing internal and external batteries, some people would be more than willing to sacrifice battery life in favour of more efficient internal storage.

posted on 04 Feb 2015, 12:16 1

32. strudelz100 (Posts: 644; Member since: 20 Aug 2014)


.1% doesn't qualify as a market.

Why do you want faster storage when Read speeds are the only thing that matters 99% of the time? Not worth the battery drain. Even if SSDs could even be made to perform on a smartphone with very limited power envelope compared to a Windows laptop or premium Windows tablet.

posted on 04 Feb 2015, 12:40

34. polybius (Posts: 32; Member since: 03 Feb 2015)


I guess the battery drain aspect is subjective. The need for a faster memory, mostly when playing graphical intensive games like NOVA3 and MC5 (just to name a few of many), is however present, and to people who enjoy gaming on the go, me included, would represent a great selling point.

posted on 04 Feb 2015, 12:48

37. TechieXP1969 (Posts: 8353; Member since: 25 Sep 2013)


I'm jus curious. How much power do you think a single SSD chip with 32GB of space is going to use? lol

posted on 05 Feb 2015, 02:14

55. wargreymon (Posts: 671; Member since: 05 Nov 2013)


You kinda hit the nail on the head with that one. The main advantage of SSDs speeds are because they use several chips. The chips in a SSD and eMMC are the same actually, but with a large SSD they can RAID more chips together.

What we need is stacked memory so you can stack several chips in a small space and RAID them, then we will get a much faster IO speeds.

posted on 04 Feb 2015, 12:46

36. TechieXP1969 (Posts: 8353; Member since: 25 Sep 2013)


I like this simply because I can see the funny big head character that goes with those!

posted on 04 Feb 2015, 13:12

39. polybius (Posts: 32; Member since: 03 Feb 2015)


Thanks, I see you know your memes.

posted on 04 Feb 2015, 09:36

9. antmiu2 (Posts: 302; Member since: 19 Jun 2011)


my oneplus one is pretty fast, can you please add it to the list?

posted on 04 Feb 2015, 09:38 8

10. Anterv (Posts: 172; Member since: 09 Jul 2014)


My Nokia 3310 is pretty fast too, can you add it to the list ?

posted on 04 Feb 2015, 10:11 2

17. macawmatt (Posts: 32; Member since: 20 Feb 2012)


Ran this on my Droid Turbo since we didn't get any love, ran it three times and averaged it.

Random Read - 15.85MB/s
Random Write - 12.34MB/s
Seq. Read - 206.03MB/s
Seq. Write - 62.51MB/s

Not bad, smoked the new Moto X and Nexus 6 in every regard, maybe this is Moto's best phone right now.

posted on 04 Feb 2015, 21:32

51. joey_sfb (Posts: 5339; Member since: 29 Mar 2012)


Your Droid Turbo is indeed fast. May i know its a 16GB, 32GB or 64GB model?

posted on 05 Feb 2015, 01:54

54. Victor.H (Posts: 651; Member since: 27 May 2011)


Awesome, thanks! And there's the difference with encryption again.

posted on 04 Feb 2015, 10:15 3

19. medicci37 (Posts: 1271; Member since: 19 Nov 2011)


Great choice for a article PA! This is something I've been curious about for a long time

posted on 04 Feb 2015, 10:21 1

22. wargreymon (Posts: 671; Member since: 05 Nov 2013)


What firmware version did you guys use on the G3? I got these scores on D855 20H:
Random Read - 21.91MB/s
Random Write - 8.68MB/s
Seq. Read - 240.70MB/s
Seq. Write - 31.84MB/s

posted on 05 Feb 2015, 01:54

53. Victor.H (Posts: 651; Member since: 27 May 2011)


Triple-checked on the LG G3, and the results are consistent with what's in the article. What's your version? I ran it on LGD855 with 4.4.2 and 3.4.0 kernel.

posted on 05 Feb 2015, 02:15

56. wargreymon (Posts: 671; Member since: 05 Nov 2013)


20H is latest Lollipop.

posted on 04 Feb 2015, 10:39 3

25. josephnero (Posts: 555; Member since: 16 Nov 2011)


Z3 FTW

posted on 04 Feb 2015, 10:57

27. isprobi (Posts: 628; Member since: 30 May 2011)


I find it interesting that Sony phones usually get rated at or near the top in most tests and discussions that I see regardless of what is being examined yet they have a hard time selling phones. I think the key to their survival is getting their phones into carrier stores whatever it takes.

posted on 04 Feb 2015, 11:34

29. JumpinJackROMFlash (Posts: 448; Member since: 10 Dec 2014)


I get far better numbers with my Note 4 than any of your phones. Maybe update your firmware to the latest?

posted on 04 Feb 2015, 12:07 1

30. strudelz100 (Posts: 644; Member since: 20 Aug 2014)


And I/O can be very different mattering on which I/O Scheduler you pick.

Possibly the easiest way to improve device performance is in fact changing the Scheduler. Most OEM's that use purchase Qualcomm's dalvik settings (Everyone but Google) go with "Read over write" or "ROW". Great for avoiding lag, but other schedulers have much better write speeds.

Its no coincidence that Nexus 6 has VERY low storage speed due to default Encryption. And its also no surprise that Samsung does well in the write portion of the test: their software tends to lag in my personal daily use.

posted on 04 Feb 2015, 12:26 1

33. numbertwo (Posts: 14; Member since: 20 Mar 2013)


My Z2 (Envy 2.2 rom, Andropluskernel v21).

Random Read - 15.05 MB/s
Random Write - 8.11 MB/s
Seq. Read - 78.1 MB/s
Seq. Write - 44.18 MB/s

posted on 04 Feb 2015, 12:49

38. TechieXP1969 (Posts: 8353; Member since: 25 Sep 2013)


I would love to see how the iPhone and phones like the Lumia or any Windows Phone stack up to any of these Android phones. I am willing to bet Apple uses very cheap slow flash storage.

posted on 04 Feb 2015, 14:34

45. isprobi (Posts: 628; Member since: 30 May 2011)


You forgot BlackBerry 10 phones like the Passport.

posted on 04 Feb 2015, 14:17

44. mike2959 (Posts: 508; Member since: 08 Oct 2011)


Haters gonna hate. By the way all the phone OEM's are cheat. It's just our roided up guy beat your roided guy.

posted on 04 Feb 2015, 16:39

48. Rain017 (Posts: 60; Member since: 11 Jul 2014)


nah... this is boring well yeah samsung is fast but touchwiz makes it slow it loads a bit slower at times innshort the experience is not as consistent as what M8 offers yeah hardware is not enough basis for a speedy performance
optimization thats all it is.. Sense 6.0 Rocks!!

posted on 04 Feb 2015, 21:38 1

52. joey_sfb (Posts: 5339; Member since: 29 Mar 2012)


Galaxy Tab S 8.4 LTE default TW KK rom (16G model)

SEQ Read = 187.38 MB/s
SEQ Write = 39.84 MB/s

Random Read = 17.14 MB/s
Random Write = 6.41 MB/s

posted on 05 Feb 2015, 04:16

57. aleiva (Posts: 1; Member since: 05 Feb 2015)


OnePlus One with Stock CM 4.4.4 (16GB model)

Random Read = 15.03 MB/s
Random Write = 8.71 MB/s

SEQ Read = 199.32 MB/s
SEQ Write = 38.59 MB/s

posted on 05 Feb 2015, 05:42

58. WallStreet (Posts: 256; Member since: 21 Sep 2014)


Awesome idea to test this functionality as it really resembles everyday usage. Maybe you can test more phones and put up the results- like Xperia Z3 Compact and BB Z30!
Congrats to PhonaArena for this article and keep'em comin'!
I think this is one of the most important aspects of a phone's functionality next to battery life.

posted on 05 Feb 2015, 06:17

59. Badelhas (Posts: 9; Member since: 29 Sep 2014)


Results on my HTC One M8 are twice the ones presented. What gives?

Want to comment? Please login or register.

Latest stories