Samsung Galaxy S4 Twice As Fast As iPhone 5

[Shocky]

Not a bit fan of this benchmark but it does show impressive numbers, HTC One does fairly well.

Wish Samsung would hurry up with the Exynos Octa versions.

The benchmark analysis in question only tested the Samsung Galaxy S4 with the Qualcomm Snapdragon 600 SoC, and according to Poole, Primate Labs can’t say how the Galaxy S4 with the new Exynos 5 Octa SoC will perform.

However, considering the fact that the Galaxy S3 with Exynos 4412 at 1.4 GHz (4 cores) is proved faster than the other variant with the Qualcomm Snapdragon MSM8960 at 1.5 GHz (2 cores), Poole believes that the Geekbench score will be similar in terms of the Galaxy S4, sporting 8-core Exynos processor, as well.

Samsung Galaxy S4 Twice As Fast As iPhone 5, Even Galaxy S3 Proves Faster Than Apple

 

[breadnatty08]

Keep in mind the iPhone 5 is over 6 months old so it's not a huge surprise.
Doesn't it have the same GPU though, but clocked higher?

 

[Shocky]

Keep in mind the iPhone 5 is over 6 months old so it's not a huge surprise.
Doesn't it have the same GPU though, but clocked higher?

Similar GPU, PowerVR SGX543MP3@325MHz in the iPhone 5 and SGX544MP3 in the Exynos powered S4.

Still don't know what the clockspeeds of the S4's GPU will be, so don't know how they compare yet, rumours suggest 533MHz but it's probably fud.

http://www.itproportal.com/2013/03/17/samsung-galaxy-s4-gpu-specs-powervr-and-adreno-ahoy/

 

[jackdubl]

What this really shows me is the exynos seems to be a lesser processor than a qualcomm with equal cores. So US customers may be okay with the qualcomm.

 

[Shocky]

What this really shows me is the exynos seems to be a lesser processor than a qualcomm with equal cores. So US customers may be okay with the qualcomm.

Yes, that was was the case the older Exynos 4412 as that used the older Cortex A9 CPU.

But that is not the case with Exynos 5410 used the Galaxy S4 because that uses the newer Cortex A15 CPU. The Qualcomm CPU Krait has similarities to the Cortex A15 but it remains to be seen how they compare.

But we suspect the Cortex A15 will outperform Krait (Snapdragon CPU) even at a lower clockspeeds, so it's the other way around now.

 

[jackdubl]

Yes, that was was the case the older Exynos 4412 as that used the older Cortex A9 CPU.

But that is not the case with Exynos 5410 used the Galaxy S4 because that uses the newer Cortex A15 CPU. The Qualcomm CPU Krait has similarities to the Cortex A15 but it remains to be seen how they compare.

But we suspect the Cortex A15 will outperform Krait (Snapdragon CPU) even at a lower clockspeeds, so it's the other way around now.

What's the Note 2 have in it? I was running benchmarks on one in the store along with my S3 and was shocked that it was only about 100-150 better on quadrant than my dual core S3.

 

[Shocky]

What's the Note 2 have in it? I was running benchmarks on one in the store along with my S3 and was shocked that it was only about 100-150 better on quadrant than my dual core S3.

Exynos 4412, 4xCortex A9@1.6GHz. Pretty much the same as the international S3 but with a minor speed increase and 2GB of memory.

Should score around 6000 in quadrant.

 

[Hadron]

The N4 and HTC One scores there seem to reflect cpu differences pretty accurately: after scaling to same clock the Snapdragon 600 should be about 15% faster than the S4 Pro, and that's what we see here.

The S4 score relative to the One is about 5% higher than the clock difference.

Be interesting to see what the dual-quad Exynos does, but real world performance is what matters, so I'm actually more interested in how it balances power consumption and responsiveness than the benchmark scores.

 

[rushmore]

Keep in mind the iPhone 5 is over 6 months old so it's not a huge surprise.
Doesn't it have the same GPU though, but clocked higher?

Benchmark smenchmarks

All I know is my Excite, TF300 and S3 all show improved benchmarks with each update, yet all of them seem no better with emulators like MAMEreloaded 1.39.

In spite of the A6X in my iPad 4 being similar for benchmarks, it is far faster with MAME and pushes a lot more pixels. Ditto with the same EA and Gameloft games. Not real apples to apples (no pun intended), but the A6X performs above and beyond synthetic benchmarks.

If you want to see true performance of the cpu in the chipset, test the game Dead or Alive + with MAMEreloaded. It is a choppy mess on every device except the iPad 4. It is not arcade smooth on the iPad 4, but it is playable. Not so on my other devices.

The Snap 600 should rock, but point is actual results matter- not benchmarks. I need to test MAMEreloaded on a Snap 600 device.

 

[CafeKampuchia]

Keep in mind the iPhone 5 is over 6 months old so it's not a huge surprise.
Doesn't it have the same GPU though, but clocked higher?

I'm sure the honest folks at Cupertino still believe their A6 is faster just like they still believe their Retina display is superior to Android's latest

 

[Shocky]

Be interesting to see what the dual-quad Exynos does, but real world performance is what matters, so I'm actually more interested in how it balances power consumption and responsiveness than the benchmark scores.

Well we already have the Nexus 10 for that, it's clocked a little higher than Quad Core version but it's almost able to match Snapdragon 600 with just two cores.

HTC One - 2535 (4 Cores Snapdragon 600 1.7GHz)
Nexus 10 - 2466 (2 Cores Exynos 5250 1.7GHz)

So with 4 A15 cores I think 4000+ should be possible, should be efficient with the 4 A7 Cores. I wonder if power saving mode will just switch the to A7's now.

Android Benchmarks - Geekbench Browser

 

[deslock]

Benchmarks can be good indicators of performance, but keep in mind that they are only indicators. Also the article's title is misleading since the S3 that we get in the USA is at the bottom of the list with a 1410 score.

And, there's a lot of variation in test results.

In this Geekbench2 test from last October, the scores were:

1834 Note2
1655 S3 (Exynos)
1644 iPhone5​

And from the longer list of results:

2535 HTC One
1923 Note2
1717 S3 (Exynos)
1711 S3 (Exynos - 4G)
1673 S3 (Exynos - Korea)
1606 S3 (Exynos - LTE)
1411 S3 (Snapdragon - N America)
1378 S3 (Snapdragon - Sprint)
1350 S3 (Snapdragon - TMobile)
1343 S3 (Snapdragon - Verizon)​

Interpolating the results, the Snapdragon-equipped S4 ought to be ~55-66% faster than the Note2 and ~124-136% faster than the Snapdragon-equipped S3.

How that translates to user experience with different screen resolutions and OS versions is unknown, but it seems promising.

 

[EarlyMon]

When selecting benchmarks, it's important to choose those that portray the results that you want to see.

The Nexus 10 scored 13,750.

More comparisons to the S4 Pro used in the Nexus 4 here -

AnTuTu Test Suite: Google Nexus 10 Review - News - AnTuTu Benchmark -- Know Your Android Better

Cheers!

 

[Rxpert83]

I wish benchmarks would die a fiery death

 

[Shocky]

When selecting benchmarks, it's important to choose those that portray the results that you want to see.

The Nexus 10 scored 13,750.

More comparisons to the S4 Pro used in the Nexus 4 here -

AnTuTu Test Suite: Google Nexus 10 Review - News - AnTuTu Benchmark -- Know Your Android Better

Cheers!

It's not about what results you want, it's about having the knowledge to understand what the results are showing you. Without that they are meaningless.

AnTuTu needs to broken down to understand the scores, but based on those results the Nexus 10 does really well for a dual core device running a 2560*1600 resolution.

The single core benchmark was interesting, that's 1.5GHz vs 1.7GHz I presume.

As for the Galaxy S4 result, no idea where they got that from but check out the xda benchmark thread of the S4. It scores higher than that, 25900 in one of the runs.

Wheres the Tegra 4 result from? is this graph made up?

http://forum.xda-developers.com/showthread.php?t=2191834

 

[EarlyMon]

I pulled that graph at random from a google search, top result was here - AnTuTu Benchmark: Samsung Galaxy S 4 vs HTC One vs NVIDIA Tegra 4 | Android and Me

And it is important to know what the results are showing you, I agree.

As most of your apps are made in Java and running the Dalvik Virtual Machine, you might want to check out CF Bench.

As I've often said - a benchmark is an app that runs inside the Dalvik that makes various function calls based on the compiler and various system library calls using various threading schemes.

Some of which may correlate to apps you use, if the benchmark functions match those used in your apps.

As for cpu performance - see CF-Bench, where native and Java results are shown. Nicely, chainfire makes no pretense about his final score, or indeed the benchmark itself, meaning anything, it's just a way he chose to represent a final result. Oddly though, I've found the Java scores to be a fair indicator of the overall difference in feel for general purpose apps on phones I've used.

There's also the Vellamo browser benchmark (don't have an opinion on their metal scores), and I find that handy as well. While it's been derided as coming from Qualcomm and therefore being biased - it simply calls your default browser and then scores how it performs. Change your browser, change the score. I find it effective at giving an idea of actual browser performance on a phone and browsing differences between two phones - provided they were both tested using the same browser set as default.

As we've discussed before, I'm not a gamer and therefore can only refer to the fact that games make a wealth of OpenGL calls, any or all of which may or may not be indicative of something.

Using some ultra-high-stress games as suggested by rushmore and then seeing how they feel and perform seems like a pretty good idea to me. More subjective, but perhaps more accurate. I recall seeing various games run side by side on youtube between two different phones with similar benchmarks of their day and the actual performance difference - even to me, a non-gamer, was as plain as broad daylight.

 

[Shocky]

About what I would expect, taking the clockspeed difference into account the Nexus 10 is still looking good considering it's Dual Core vs Quad Core.

CF-Bench

Nexus 10

Native - 17377
Java - 6439
Overall - 10814

Nexus 4

Native - 23074
Java - 7427
Overall - 13685

 

[EarlyMon]

Indeed, that's an excellent showing.

Is either one noticeably different from the other in day to day use?

 

[Shocky]

Indeed, that's an excellent showing.

Is either one noticeably different from the other in day to day use?

Probably, most apps don't take advantage of additional cores

I'd prefer a Dual Core Cortex A15@2GHz or higher over 1.6GHz Quad Core.

Also the more I like at the Nexus 10 benchmarks the more I question why they went with the PowerVR GPU as Mali T604 does a fantastic job.

 

[Gearu]

It's just like putting a car on a Dyno and only looking at the final horsepower/kilowatts, it's all for nothing if the car doesn't handle well or is slow getting there.

 

[ceva321]

seriously? phones are fast now...in real life who cares?

 

[treb1797]

Yahoo!

I thought folks on this thread might find this of interest.

 

[rushmore]

With the exception of the speed reason, we can already say the other nine for the S3

At least as far as the CPU is concerned, the iPhone 5 runs circles around my S3. No contest.