Samsung Galaxy S6 (Exynos 7420) vs HTC One M9 (Snapdragon 810) vs iPhone 6 (Apple A8) performance review


Three phones will battle for the attention of most users throughout most of 2015: the Apple iPhone 6 with its record-beating sales, the completely reimagined Samsung Galaxy S6, and the stylish HTC One M9.

Now that Samsung has finally realized the importance of designed and has released the Galaxy S6 in a completely overhauled aesthetic, we can finally say there is some sort of parity between flagships this year when it comes to their design appeal.

However, gone are the times of all-Snapdragon-toting devices, and this year, we have a true diversity of processors: the iPhone 6 touts the Apple A8 SoC, the Galaxy S6 features a Samsung-made Exynos 7420, and the One M9 sports a Snapdragon 810 chip.

What are the differences between the three, and which one works best? We take a look at the CPU and GPU performance, as well as the effect they have on battery life in this Galaxy S6 vs One M9 vs iPhone 6 performance review.

CPU


Rather than speaking about the system chip on the whole, we take a look first at its one crucial part: the CPU performance. It’s common knowledge that the CPU takes the heavy load in most of daily operations including browsing (surfing Facebook too), while the GPU starts to take that heavy load in parallel operations, most notably games and video encoding. What’s not so common is an understanding of the finer details of single-threaded vs multi-threaded apps, or put in simpler terms: how much having multiple cores helps (or does not help) in different apps.

There is no simple answer to that: Android handles all UI interactions in the main thread of an app, which alone suggests using at least two threads for the vast majority of apps, but it’s much harder to judge the actual usefulness of an octa-core setup. One educated guess is that most common apps will not reap the benefits of the octa-core chip fully, while some specialized demanding apps and games will do see a boost in performance. Given the elevated importance of single-core performance, when looking at benchmarks, we tend to prioritize the single-core score of devices (without neglecting the multi-core result, naturally). This is an important differentiation because some particular benchmarks like AnTuTu make excellent use of multiple cores, which may not be the case in real-life usage. Rather than looking at synthetic benchmarks, it’s worth keeping an eye on more granular ones that allow better understanding of CPU performance.


First, let’s recap the specs, though. The Galaxy S6 runs on an Exynos 7420 SoC with an octa-core big.LITTLE setup consisting of four high-performance Cortex-A57 cores running at up to 2.1GHz, and four Cortex-A53 battery-efficient cores running at up to 1.5GHz. The One M9, on the other hand, sports a Qualcomm Snapdragon 810 with a similar big.LITTLE A57/A53 solution running at 2GHz (A57, slightly underclocked) and 1.5GHz (A53). The iPhone 6 is the odd one here with its Apple A8 chip that features ‘only’ two CPU cores running at ‘only’ 1.4GHz. We’re stressing only since pundits are often quick to accuse Apple for being behind the times, without taking in consideration that the Cyclone v2 cores are both larger in size and complexity, and are able to run at their nominal clock speeds for prolonged periods of time, while the clock speeds in the former two devices are ‘turbo’ clock speeds that cannot be sustained effectively.

With this in mind, let’s turn to the actual benchmark scores, the comment is right below them.

Geekbench 3 single-core Higher is better
Samsung Galaxy S6 1440
HTC One M9 1209
Apple iPhone 6 1630
Samsung Galaxy S5 944
Samsung Galaxy Note 4 1112.67
HTC One (M8) 888
Apple iPhone 5s 1323
LG G3 950
Geekbench 3 multi-core Higher is better
Samsung Galaxy S6 5127
HTC One M9 3738
Apple iPhone 6 2927
Samsung Galaxy S5 2900
Samsung Galaxy Note 4 3259.67
HTC One (M8) 2613
Apple iPhone 5s 2240
LG G3 2545
Basemark OS II Higher is better
Samsung Galaxy S6 1767
HTC One M9 1413
Apple iPhone 6 1239
Samsung Galaxy S5 1054
Samsung Galaxy Note 4 1038.67
HTC One (M8) 1071
LG G3 951
Sunspider Lower is better
Samsung Galaxy S6 354.5
HTC One M9 721.3
Apple iPhone 6 353.4
Samsung Galaxy S5 777.3
Samsung Galaxy Note 4 1087.87
HTC One (M8) 693.1
Apple iPhone 5s 415.7
LG G3 947.2


Two things catch the eye the most: one, the iPhone 6 is still the best performing phone when it comes to single-thread performance, and two, the Galaxy S6 is clearly the best-performing Android phone (that comes close to the iPhone 6 in single-core CPU performance). Also, despite the similarity between the CPUs of the Galaxy S6 and the One M9, CPU performance is still noticeably higher on the S6’s Exynos chip.

The biggest improvement in CPU performance is clearly achieved by Samsung: a whopping 50%+ jump in single-core CPU performance. Interestingly, despite both the S6 and M9 using stock Cortex A57 cores, the M9 is outperformed by a hefty margin of more than 18%. It's important to know that HTC has pushed an update to its One M9 just before the launch of the phone, throttling performance of the chip in order to lower down what seemed like very disturbing overheating problems. Recent testing by AnandTech has revealed the full extent of that throttling, and you can see for yourself how the One M9 caps its processor further down with more use.

GPU


The next important chapter to overall great performance is the GPU, of course. Taking over in games, a good GPU guarantees that you’ll be able to play the latest titles on your phone with ease, and gives you faster processing speeds in some visually-demanding apps.

First, let’s recap the specs. The Galaxy S6 ships with an ARM Mali-T760 MP8 graphics chip, the HTC One M9 uses the Qualcomm Adreno 430, and finally, the iPhone 6 sports an Imagination Technology PowerVR GX6450.

Before diving in the benchmarks, here also a few remarks are due. First comes the issue of screen resolution. You’ll see on-screen and off-screen benchmark scores (offscreen tests are used to neutralize differences in resolution and equalize all devices to a 1080 x 1920-pixel resolution).

When it comes to resolution, the differences between these three phones could not be larger: the iPhone 6 sports a 750 x 1334-pixel 4.7” display, the One M9 - a 1080 x 1920-pixel 5” screen, and the Galaxy S6 - a 1440 x 2560-pixel 5.1” display.

How much is that a factor for games? A higher resolution naturally means an increased load for the GPU, and hence, lower frame rates. How much does game scaling actually affect GPU performance is an open question, but so far indications are that higher resolutions do take a heavy toll on performance. With this in mind, let’s take a look at the actual graphics benchmarks, split in on-screen and off-screen scores.

GFXBench T-Rex HD on-screen Higher is better
Samsung Galaxy S6 37
HTC One M9 49
Apple iPhone 6 48.9
GFXBench T-Rex HD off-screen Higher is better
Samsung Galaxy S6 46
HTC One M9 50
Apple iPhone 6 42.3
GFXBench Manhattan 3.1 on-screen Higher is better
Samsung Galaxy S6 16
HTC One M9 24
Apple iPhone 6 25.8
GFXBench Manhattan off-screen Higher is better
Samsung Galaxy S6 19
HTC One M9 24
Apple iPhone 6 17.8


Effect on battery life


Finally, when speaking about CPU and GPU performance, we're inevitably faced with the question of how performance relates to battery life. With all the countless reports about the Snapdragon 810 overheating, there was reason for worry. Luckily, HTC has pushed an important software update just days before the launch of the One M9, capping its performance, but also securing it from overheating.

In order to see the effect of the new system chips on battery life, we turn to our battery test that has all phones pre-calibrated at 200 nits (a brightness level comfortable for indoor viewing), and running a custom script that simulates typical smartphone use. Keep in mind that the result represent battery life with no off-screen time. Keep in mind that some phones (the iPhone 6, in particular) does a stellar job with stand-by time, and this result is not factored in.

With all this in mind, it's interesting to see that there is a slight regression in battery life on the Galaxy S6 and One M9 (in comparison with their predecessors), while the iPhone 6 is a modest improvement over the iPhone 5s. At the same time, the Galaxy S6 remains the best scoring device in terms of battery life, while the One M9 follows and the iPhone 6 is at the third position. This is an interesting result mostly because the Galaxy S6 has a Quad HD display (which puts an extra strain on battery), while the others have lower resolutions, and yet still, it manages to score above them. It's important to also note that the Exynos 7420 on the Galaxy S6 is manufactured using the most advanced, 14nm process, while the rest are made using less efficient manufacturing processes.
Battery life (hours) Higher is better
Samsung Galaxy S6 7h 14 min (Good)
HTC One M9 6h 25 min (Average)
Apple iPhone 6 5h 22 min (Poor)
Samsung Galaxy S5 7h 38 min (Good)
HTC One (M8) 7h 12 min (Good)
Apple iPhone 5s 5h 2 min (Poor)
Charging time (minutes) Lower is better
Samsung Galaxy S6 78
HTC One M9 106
Apple iPhone 6 147
Samsung Galaxy S5 122
HTC One (M8) 207

Conclusion


All in all, we're most impressed with the improvements in performance achieved by the Galaxy S6's Exynos 7420 system chip. It is still a runner-up to the iPhone 6 when it comes to the hugely important single-core performance score, but it's getting very close, and with all the improvements to TouchWiz, it no longer feels terribly laggy - quite the opposite. 

When it comes to grpahics, though, the Adreno 430 in the One M9 sweeps the floor with its overwhelming advantage over the Mali chip in the S6 and the PowerVR GPU on the iPhone 6 (those two score on par, with the S6 being a bit ahead).

So which of the three should you pick if you're looking for the best performance? None will disappoint in daily use where software optimizations matter a lot (the One M9 is in our subjective view the fastest), and when it comes to gaming it is again the One M9 that pulls ahead. For intense browsing and all other heavier CPU tasks, though, the Galaxy S6 and iPhone 6 stand out.

FEATURED VIDEO

78 Comments

1. bucky

Posts: 3795; Member since: Sep 30, 2009

so to summarize... Iphone - best per core performance GS6 - best multi core performance HTC - best gpu Each phone has its strengths. Question with the HTC GPU...will it hold this performance for a while or downclock shortly afterwards?

3. TyrionLannister unregistered

These benchmarks are not correct by a long shot. S6 scores higher in GPU benchmarks too. See AnandTech or official GFXBench results if you don't believe me.

5. Neros

Posts: 1016; Member since: Dec 19, 2014

Doesn't matter that andreno beats mali in benchmarkz because mali will beat it in real-life apps. Both gpu and cpu got declocked with the update. Go look for a speed test between m8 and m9. M8 destroys m9 easily. SD810 is a joke. https://m.youtube.com/watch?v=fNN0I4zB5Wk

8. shaineql

Posts: 522; Member since: Apr 28, 2014

Rofled at M9 after update nerf , get rekt :D

17. Supraman21

Posts: 467; Member since: Jun 09, 2010

From reading your previous comments you seem to know your way around processors. Legitimate question..why can't Samsung design a chip that can't beat apples single core performance when its almost a year older?

23. Dr.Hye

Posts: 95; Member since: Apr 16, 2014

Samsung is working on one and should be released with the note 5 or by 2016 flagship. It is called mongoose

25. rubyonrails3

Posts: 375; Member since: Oct 01, 2014

will be glad to see How Note 5 Cores stack up against A9 processor. After seeing poor performance on M9, I feel now only iPhone can compete in term of raw power with Samsung mobiles. thing is competition between iPhone and Galaxy(even Note series) got tough now. iPhone and Galaxy mobiles are geat in camera great in performance and benchmarks now great designs too useful features too like fingerprint scanner I wish apple kick sumsung butt with iPhone 7 or 6S whatever it will be.

56. Techielover

Posts: 105; Member since: Feb 20, 2015

We don't know if the A9 will be cyclone gen 3 or something new all together.

26. arch_angel

Posts: 1651; Member since: Feb 20, 2015

Because Apple as been using the same dual core set up from the beginning only focusing on upgrading the dual core technology in there phones so they excel at making super high performing dual cores. The more cores the lower the single core performance is usually. It's quit impressive that Samsung has managed to get such a high single core performance out of the 4 cortex A57 cores Being tested in the geekbench test. Seems likely that Samsung well pass Apple in single performance soon despite using more cores than Apple

28. Taters

Posts: 6474; Member since: Jan 28, 2013

Um I am 100% sure that Samsung has surpassed Apple a long time ago in all aspects of hardware. The difference? Android is so open that you can't cheat a single lick without an Apple employee who is experienced with software finding out with a few standard benchmark tests. With IOS, it so closed that they could cheat like hell at benchmarks and no one without special knowledge could find out.

41. arch_angel

Posts: 1651; Member since: Feb 20, 2015

I'm talking about the single core performance which Apple has been top dog in for a while now.

29. elitewolverine

Posts: 5192; Member since: Oct 28, 2013

It's a little more complicated, heck sometimes it can even come down to patents. Chip designs are the like the area 51 of companies and why most people will delegate it out. It is not easy to design a chip, even intel borrows patents for x64 which amd has, probably the only reason amd is still alive imo. At any rate, a large single core can increase in complexity, allowing it to perform better single operations ever clock cycle. After intel stopped chasing the mhz rate, and was loosing to lower clocked but more 'brute' forced amd's they redesigned their chips. Even now dual core intel extremes are killer for games because most games wont use quad or 8 cores on a cpu. I would probably point to an article that could explain the in/outs more than i could.

30. TyrionLannister unregistered

PA ran the 32-bit GeekBench here. In reality, the S6 on 64-bit geekbench score about 1510 which is within 10% of Apple A8. A8 is merely 6 months old and is based on gen.2 cyclone cores which are custom cores. Cyclone are 6-way superscalar v/s 3-way superscalar of Cortex A57 used in the Exynos. That means that it can do double the calculations per clock, also known as IPC(instructions per-clock). It's a very huge deal that a chip with A57 cores can reach within 10% of cyclone gen.2. To put things in perspective: At 20 nm HKMG. A cortex A57 core takes about 3.8 mm^2 and a Cyclone core takes about 6 mm^2. You can't expect cortex A57 to match the incredible muscle of the cyclone custom cores. Anyway, it's not all gloomy. Cortex A57 consume less power compared to A8 at about the same performance and at same clock, they consume considerably less power. Also, the denver cores used in tegra K1 is a 7-way superscalar chip which has higher IPC than even the cyclone. It scores 2200 on Geekbench single core. Also, a custom core from Samsung is soon going to be released codenamed 'mongoose' which is rumored to muster a score of 2240 on a single core. Either way, single core score is not as necessary on Android as Java is extremely good with multi-threading and most of the times, 3 or more threads are active. Hope this answered your question.

33. Supraman21

Posts: 467; Member since: Jun 09, 2010

Definitely! Thanks. When you say threads, does that mean cores? From my understanding most Android applications only require two cores?

38. TyrionLannister unregistered

Multi-threaded means a process that runs on multiple cores. So yes, your assumption was correct.

39. Napalm_3nema

Posts: 2236; Member since: Jun 14, 2013

It would be a big deal, but you aren't factoring in the clock rate. 2/2.1 GHz is a huge jump over 1.4 GHz.

40. TyrionLannister unregistered

Still, it's only 1.5x high and doesn't overcome the 2x IPC of Apple A8. The chip should be about 2.3-2.4 GHz to match the A8

44. BobbyBuster

Posts: 854; Member since: Jan 13, 2015

Name me a single 64bit native app for Android that's not a benchmark. You can't since there isn't any.

52. TyrionLannister unregistered

Dude, 85% of all android apps are 64-bit. Google it.

65. BobbyBuster

Posts: 854; Member since: Jan 13, 2015

85% of ALL yes, including all the craps. And close to 0% of those I and most people actually use. Check the apps you are using, and you will be surprised to see almost none of them runs in 64bit. You talk way too much more than you know. STFU.

70. TyrionLannister unregistered

ROFL. All the apps I use are Java coded including all my system apps. That means all my apps can be 64-bit if I use an AArch64 based kernel and AP. You talk way too much more than you know. STFU.

50. Techielover

Posts: 105; Member since: Feb 20, 2015

But would "mongoose" be able to sustain that high single core score over multiple runs from a benchmark.The A8 are especially able to sustain their scores over time. We don't know if "mongoose" is dual core,tri core or quad core(anything more is not gonna happen).If it's clocked at 2.2 samsung gonna have to piggyback on their 14nm process to keep power consumption down. Isn't multi threading possibly on even a single core processor or am I confusing this with multiprocessing.

53. TyrionLannister unregistered

I have no idea. It depends on so many factors.

55. Techielover

Posts: 105; Member since: Feb 20, 2015

I can't wait for "mongoose".We will truly see if it lives up to its hype if the note 5(lets hope it's ready by then so the s7 could have gen 2 or 1.5)comes with UHD. Off topic:Hope the note 5 or s7 comes with 14nm LPP

71. Salazzi

Posts: 537; Member since: Feb 17, 2014

multi-threading can be done on a single core level as well.

73. sunspider

Posts: 4; Member since: Jan 26, 2015

Finally we have a smart guy here to answer clearly how is the different between A8 core (cyclone) and Exynos 7420 core (cortex A57). Some stupid ifan alway said that A8 is just 2core compare to 8core of Exynos. But they dont know that the core of A8 is bigger and consume more power than cortex A57 core in Exynos. If Apple make a 4 core or 8 core version of apple A8 SoC (with the same cyclone core as now) and you the iphone user will use your stupid iphone with the charger always plug in.

35. j2001m

Posts: 3061; Member since: Apr 28, 2014

Because it as 2 big cores that's all ther is to it as iOS works best like that, if they put there CPUs into any Google OS phome it will lag badm there score is not very good when you look at it only having 2 cores

59. vincelongman

Posts: 5761; Member since: Feb 10, 2013

Because the 7420 is a octa core, with 4 "big" A57 cores and 4 "small" A53 cores While the A8 is a dual core with 2 "huge" Cyclone cores Basically there's no way Samsung (or Apple or Intel or Nvidia) can put 4 "huge" cores in a phone (or tablet) without overheating Or having to underclock, which would reduces its single core performance anyway That's also why the iPad Air 2's A8X has 3 Cyclone cores, not 4 And that's why Intel's Core M and Core U series are dual cores as well

61. ph00ny

Posts: 2069; Member since: May 26, 2011

Because single core performance isn't their top priority. Same reason why they're shoving in 8 cores It's like comparing small block chevy engines which are light and small in package size but big in displacement. People often say things like why does it need *insert size* liters to output only this much power when *insert manufacturer* does it with *insert smaller displacement* to put out same power?

6. Neros

Posts: 1016; Member since: Dec 19, 2014

Best gpu only in benchmarks.

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.
FCC OKs Cingular's purchase of AT&T Wireless