Database Info

G2x Benchmark Comparison

This thread is meant to be a comparison of changes over time for the G2x and not really a 'me too' benchmarks thread. The idea is to see unequivocally how various changes or versions impact our hardware over time. Benchmarking is something I enjoy and take very seriously. At this time I do not plan to run any custom ROMs and I will be holding off on OC/UV kernels until they have matured.
Feel free to suggest additional tests but please be realistic in what you're asking for.
Completed Tests
G2x 2.2 Stock
G2x 2.2 Rooted + Frozen Apps
G2x 2.3 Stock + Frozen Apps
G2x 2.3 Rooted
G2x 2.3 Stock + faux123 SV Kernel
G2x 2.3 Stock + faux123 OC Kernel @ 1.1
G2x 2.3 Stock + faux123 OC Kernel @ 1.4
G2x 2.3 Stock + faux123 OC Kernel @ 0.5
G2x ROM Benchmarks
G2x CM 7.1 RC vs Nightly 124 (2.3.5) vs Nightly 154 (2.3.7) Benchmarks
Planned Tests
???
Testing Methedology
All tests are run on freshly booted phones
Airplane mode is enabled for tests not involving the internet
First runs are ignored for caching purposes
Each test is run 10 times and averaged
Erroneous data points are ignored (values that +/- the norm of other tests excessively)
Revisions
2011.05.05 - Initial Release: G2x Stock vs G2x Rooted + Frozen Apps
2011.07.26 - Second Release: G2x 2.2 Stock vs G2x 2.2 Rooted vs G2x 2.3 Stock
2011.08.30 - G2x 2.3 Rooted vs Faux123 Kernels
2011.08.30 - G2x 2.3 Faux123 Kernels @ 500 MHz vs G1 @ 614 MHz
2011.09.07 - G2x ROM benchmarks

2011.05.05 - G2x Stock vs G2x Rooted + Frozen Apps
Higher Resolution Images
This test demonstrates the rooting and then using an application to freeze apps that are running. I did not freeze MyAccount or Tegra Zone since I use them.
AppPack
logmein.rescue
logmein.rescuesecurity
EA Games
NFS Shift
Nova
Polaris Office
SmartShare
T-Mobile TV
TeleNav
Video Chat (qik)
Wi-Fi Calling
Zinio Reader
As can been seen from the attached graph the there was a noticeable bump in performance in all tests performed on the phone. SmartBench 2011 showed a larger gap than Quadrant because a couple of its tests are SMP aware and probably benefited more from the extra overhead on the two cores. The gain of ~250 in Quadrant for comparison is the value a stock G1/MT3G gets by itself in Quadrant. SunSpider performs extensive small javascript tests that I did not expect to see much gain from since it is more implementation bound than CPU bound where SmartBench however performs much more CPU intensive visual browser tests.
2011.07.26 - G2x 2.2 Stock vs G2x 2.2 Rooted vs G2x 2.3 Stock
Higher Resolution Images
I've updated my G2x via the LG update software to the GB 2.3 that was released for it. I did not unroot/unfreeze prior so I lost root but my frozen apps stayed frozen post update. As can be seen from the graph the Gingerbread tests show it performing within 2% of the results of the previous FroYo 2.2 rooted phone. I will not be able to perform a true stock test until I re-root the phone and unfreeze applications. The SunSpider test shows a decent 5% gain in speed which is not unexpected since browser performance was a touted feature of 2.3 on phones.
The lack of overall performance gain though was expected for this update before I ran the tests.
FroYo was the performance release version of the 2.x series of versions where 2.3 was a feature release. The addition of new features will typically garner a drop in overall performance unless additional time is spent focusing solely on performance increases. Google has already stated that the next phone oriented performance release will be Ice Cream Sandwich (4.0?) which will merge in the multi-core enhancements of the Honeycomb 3.x series.
This release of 2.3 is still running the 2.6.32.9 kernel version # as the previous 2.2 version was. This DOES NOT mean it is the same kernel, but it does indicate a lack of major changes, just primarily bug fixes and tweaks were done. The kernel controls task switching and various other base level processes, without a significant change at this level there will not be significant changes at the upper levels.
2011.08.30 - G2x 2.3 Rooted vs Faux123 Kernels
Higher Resolution Image
After running my updated 2.2 Rooted + Frozen to 2.3 Unrooted + Frozen I felt there was something wrong with it as some things felt slightly off. My belief is the freezing process removes the permissions from an application required to read/write/execute it which in turn may have prevented specific items from being updated during the LG Updater process. I decided to do the GB root process and flashed the stock rooted GB file onto my phone which being a ROM flash and not an updater over wrote the frozen applications. Afterwards with everything running the phone just behaved more as I expected than the previously frozen version. Benchmarks attached show this as everything increased (except for SunSpider).
A note on SunSpider I was reporting results wrong previously as I forgot lower scores in SunSpider mean better results. Which in terms of the 5% variance is still not much of a change.
Now being rooted again I decided to start working with kernels to see how they changed things.
Faux123 SV (Stock Voltage & Frequencies) - This kernel showed higher values in some tests but lower in others. The main disadvantage is the memory compression in my opinion as it actually sacrifices bandwidth for capacity. The thing is though even with our ~380mb of RAM it's still more than enough for daily use. If you use the System Stats Live Wallpaper with memory listed you'll see that even with the most taxing applications this phone never reaches capacity. The decrease in bandwidth is noticeable with larger applications like games as they take longer to load and to switch between large maps and other data sets.
Faux123 OC/UV @ 1.1 - The slight bump in frquency provides the extra benefit to push the memory bandwidth past the hurdle created by compcache. This kernel provides an all around (except in 3D applications) performance increase making it very worth while even at this slight MHz bump.
Faux123 OC/UV @ 1.4 - Everything else before this was in single digit % of increase. The bump to 1.4 GHz on both cores provides significant increases as the graph shows. If your phone can run stable at these speeds then I say go for it. For me 1.4 GHz was as high as I could go and stay stable, 1.55 GHz would crash my phone on boot.
2011.08.30 - G2x 2.3 Faux123 Kernels @ 500 MHz vs G1 @ 614 MHz
Having been an Android user from the start with a T-Mobile G1 & MT3G I still actively use that phone for testing and playing around with. When I actively used it we were always trying to find various tricks just to get smidgens more performance from it. So I decided to underclock the G2x down to 500 MHz and overclocked the G1 to 614 MHz running CS-DOS (GB 2.3.4) with everything tweaked for the fun of it.
This comparison shows the difference in terms of an ARMv6 versus an ARMv7 Cortex9 architecture and is just a fun informative bit. While being 100 MHz slower the G2x still was able to outperform the tweaked G1 by the following.
Quadrant - 72% faster
SmartBench - 79% faster
BrowserMark - 27% faster
SunSpider - 61% faster
2011.09.07 - G2x ROM benchmarks
High Res Image
I've completed my various ROM comparison testing and below are my numerical findings. I averaged the overall results and as listed in the order below are the highest to lowest scoring ROMs in terms of performance. In terms of numerical, graphical, & I/O performance they're all very close to each other. However it becomes very clear that either the stock ROM's web browser has been modified or the Android 2.3.4/2.3.5 web browser has had some significant performance optimization done to it as BrowserMark & SunSpider both return values nearly twice as fast as the 2.3.3 based versions. I believe that CompCache settings on the Faux123 kernel used in the Faux123 ROM result in its lower CF-Bench scores as it impacts memory bandwidth. With Miui it has a different/modified web browser in place and it shows as it has the worst of the browser benchmark scores.
CM7.1-RC1
CM7 Nightly 124 (2011.08.31)
Eagles Blood 1.08
Faux123 1.3.1
G2x 2.3 Root
Stock Tweaked 1.2.2
Miui 1.8.26
Weapon G2x 2.4 (0823)
2011.09.30 - G2x CM benchmarks
With the release of G2x Nightly 154 the codebase is now Gingerbread 2.3.7 so I decided to do a CM benchmark comparison to see how it handled compared to the others. Of not since the 14x series of nightlies EXT4 was also enabled by default.
As 2.3.6/2.3.7 were mainly Sprint Nexus S updates along with Google Wallet feature additions I am not surprised from the results to see the numbers are fairly close to each other. The 2.3.7 eeked out slightly ahead in CF-Bench but lagged behind in SunSpider by about 500 milliseconds making it's score larger (ie slower) than 2.3.5 however those values are still much better than 2.3.3 based browsers.

Performing benchmarks with the newly installed GB version on my G2x now, results to follow. I was rather surprised to see that all of the apps I had frozen while rooted are still frozen now post installation even though my phone is no longer rooted.

my quadrant score
my quadrant score on a g2x ..

This isn't meant to be a showoff type thread but a scientific analysis. If you want to contribute I suggest following the guidelines I set out in the OP along with providing detailed information about your setup and settings.

I've updated the first 2 posts with the same information.
2011.07.26 - G2x 2.2 Stock vs G2x 2.2 Rooted vs G2x 2.3 Stock
Higher Resolution Images
I've updated my G2x via the LG update software to the GB 2.3 that was released for it. I did not unroot/unfreeze prior so I lost root but my frozen apps stayed frozen post update. As can be seen from the graph the Gingerbread tests show it performing within 2% of the results of the previous FroYo 2.2 rooted phone. I will not be able to perform a true stock test until I re-root the phone and unfreeze applications. The SunSpider test shows a decent 5% gain in speed which is not unexpected since browser performance was a touted feature of 2.3 on phones.
The lack of overall performance gain though was expected for this update before I ran the tests.
FroYo was the performance release version of the 2.x series of versions where 2.3 was a feature release. The addition of new features will typically garner a drop in overall performance unless additional time is spent focusing solely on performance increases. Google has already stated that the next phone oriented performance release will be Ice Cream Sandwich (4.0?) which will merge in the multi-core enhancements of the Honeycomb 3.x series.
This release of 2.3 is still running the 2.6.32.9 kernel version # as the previous 2.2 version was. This DOES NOT mean it is the same kernel, but it does indicate a lack of major changes, just primarily bug fixes and tweaks were done. The kernel controls task switching and various other base level processes, without a significant change at this level there will not be significant changes at the upper levels.

I've updated the first 2 posts with the same information.
2011.08.30 - G2x 2.3 Rooted vs Faux123 Kernels
Higher Resolution Image
After running my updated 2.2 Rooted + Frozen to 2.3 Unrooted + Frozen I felt there was something wrong with it as some things felt slightly off. My belief is the freezing process removes the permissions from an application required to read/write/execute it which in turn may have prevented specific items from being updated during the LG Updater process. I decided to do the GB root process and flashed the stock rooted GB file onto my phone which being a ROM flash and not an updater over wrote the frozen applications. Afterwards with everything running the phone just behaved more as I expected than the previously frozen version. Benchmarks attached show this as everything increased (except for SunSpider).
A note on SunSpider I was reporting results wrong previously as I forgot lower scores in SunSpider mean better results. Which in terms of the 5% variance is still not much of a change.
Now being rooted again I decided to start working with kernels to see how they changed things.
Faux123 SV (Stock Voltage & Frequencies) - This kernel showed higher values in some tests but lower in others. The main disadvantage is the memory compression in my opinion as it actually sacrifices bandwidth for capacity. The thing is though even with our ~380mb of RAM it's still more than enough for daily use. If you use the System Stats Live Wallpaper with memory listed you'll see that even with the most taxing applications this phone never reaches capacity. The decrease in bandwidth is noticeable with larger applications like games as they take longer to load and to switch between large maps and other data sets.
Faux123 OC/UV @ 1.1 - The slight bump in frquency provides the extra benefit to push the memory bandwidth past the hurdle created by compcache. This kernel provides an all around (except in 3D applications) performance increase making it very worth while even at this slight MHz bump.
Faux123 OC/UV @ 1.4 - Everything else before this was in single digit % of increase. The bump to 1.4 GHz on both cores provides significant increases as the graph shows. If your phone can run stable at these speeds then I say go for it. For me 1.4 GHz was as high as I could go and stay stable, 1.55 GHz would crash my phone on boot.

2011.08.30 - G2x 2.3 Faux123 Kernels @ 500 MHz vs G1 @ 614 MHz
High Res Image
Having been an Android user from the start with a T-Mobile G1 & MT3G I still actively use that phone for testing and playing around with. When I actively used it we were always trying to find various tricks just to get smidgens more performance from it. So I decided to underclock the G2x down to 500 MHz and overclocked the G1 to 614 MHz running CS-DOS (GB 2.3.4) with everything tweaked for the fun of it.
This comparison shows the difference in terms of an ARMv6 versus an ARMv7 Cortex9 architecture and is just a fun informative bit. While being 100 MHz slower the G2x still was able to outperform the tweaked G1 by the following.
Quadrant - 72% faster
SmartBench - 79% faster
BrowserMark - 27% faster
SunSpider - 61% faster

I find this thread very interesting and informative. Thanks for this
Sent From My G2x EB 1.07

Thanks!
I've started performing ROM benchmarks now. This is my intended set of ROMs to evaluate using the same testing patterns. I chose ROMs that are actively being worked on and used.
CM7.1-RC1
CM7 Nightly 124 (2011.08.31)
Eagles Blood 1.08
Faux123 1.3.1
Miui 1.8.26
Stock Tweaked 1.2.2
Weapon G2x 2.4 (0823)

2011.09.07 - G2x ROM benchmarks
High Res Image
I've completed my various ROM comparison testing and below are my numerical findings. I averaged the overall results and as listed in the order below are the highest to lowest scoring ROMs in terms of performance. In terms of numerical, graphical, & I/O performance they're all very close to each other. However it becomes very clear that either the stock ROM's web browser has been modified or the Android 2.3.4/2.3.5 web browser has had some significant performance optimization done to it as BrowserMark & SunSpider both return values nearly twice as fast as the 2.3.3 based versions. I believe that CompCache settings on the Faux123 kernel used in the Faux123 ROM result in its lower CF-Bench scores as it impacts memory bandwidth. With Miui it has a different/modified web browser in place and it shows as it has the worst of the browser benchmark scores.
CM7.1-RC1
CM7 Nightly 124 (2011.08.31)
Eagles Blood 1.08
Faux123 1.3.1
G2x 2.3 Root
Stock Tweaked 1.2.2
Miui 1.8.26
Weapon G2x 2.4 (0823)

2011.09.30 - G2x CM benchmarks
With the release of G2x Nightly 154 the codebase is now Gingerbread 2.3.7 so I decided to do a CM benchmark comparison to see how it handled compared to the others. Of not since the 14x series of nightlies EXT4 was also enabled by default.
As 2.3.6/2.3.7 were mainly Sprint Nexus S updates along with Google Wallet feature additions I am not surprised from the results to see the numbers are fairly close to each other. The 2.3.7 eeked out slightly ahead in CF-Bench but lagged behind in SunSpider by about 500 milliseconds making it's score larger (ie slower) than 2.3.5 however those values are still much better than 2.3.3 based browsers.

bilalrashid said:
my quadrant score on a g2x ..
Click to expand...
Click to collapse
Yeah! Run that script that inflates your score! Quadrant so useless. Informative post.

zetsumeikuro said:
Yeah! Run that script that inflates your score! Quadrant so useless. Informative post.
Click to expand...
Click to collapse
As a comparison between various users and phones I agree, but it still functions well at least for testing performance gains or losses on the same phone.

Eaglesblood 2.4 , Dragon 1.51, o/c'ed @1.4g
Sent from my LG-P999 using xda premium

mustangtim49 said:
Sent from my LG-P999 using xda premium
Click to expand...
Click to collapse
Post details such as cpu speed, etc.

More than that, the data lacks relevance. You have no baseline, no definition of testing process, just one benchmark from one application that does not correlate to the ones I've already presented so there is no way to compare these numbers to any others provided.

I'm running tsugi 2.3.4 oyasumi 2.6.32.9 gb kernel running @ 1400 MHz quadrant average 3784 fresh boot ten tests. Antutu reboot ten test average 6420
Sent from my LG-P999 using xda premium

I would recommend using different colors. The All-Orange makes it hard to compare the key to its corresponding graph.

But orange is my favorite color!
I'm going to add in AnTuTu benchmark to my testing set as it seems to be fairly popular now. I'm also going to do some additional tests soon including the latest nightly kang builds of 7.2 RC0 along with some updated Faux kernels. I'm rather excited about nVidia finally making the Tegra2 resources available so hopefully upcoming CM 7.x builds & CM 9 builds will have full true hardware support.

[Q] Transformer , ICS and overclocking ?

I've been wondering ..
I've got ICS on my transformer and i've got root .. I saw some video of benchmarks for the transformer with HC and then with ICS .. and ICS was getting alot better score (35-40% or some).
I know some people where overclocking with HC and some custom kernels, but i doubt there is yet a custom kernel for the ICS..
But is anyone here that had overclocked Transformer with HC, that can tell me how much diffrence OC'ing made ? 10%? 20%? none at all ?
Also was is stable ?
I would like to go to maybe 1.2ghz.. 1.3ghz even .. But im not sure if it is worth the trouble.

I had my TF overclocked to 1.5ghz 24/7 (using Interactive Governor) on Krakd Rom and it was extremely stable, never rebooted or froze and I can say it made a pretty big difference in speed/responsiveness. I would say around 30% increase IMO.
It varies between tablets though so you may find your tablet is unable to OC at 1.5 ghz and remain stable.
The browser in HC @ 1.5ghz was faster for me than it is now on ICS stock clock (Revolution HD) so I can't wait for an OC kernel to be released for ICS to hopefully get the browser running a bit quicker.

kitchenfork said:
I had my TF overclocked to 1.5ghz 24/7 (using Interactive Governor) on Krakd Rom and it was extremely stable, never rebooted or froze and I can say it made a pretty big difference in speed/responsiveness. I would say around 30% increase IMO.
It varies between tablets though so you may find your tablet is unable to OC at 1.5 ghz and remain stable.
The browser in HC @ 1.5ghz was faster for me than it is now on ICS stock clock (Revolution HD) so I can't wait for an OC kernel to be released for ICS to hopefully get the browser running a bit quicker.
Click to expand...
Click to collapse
Alost exactly the same for me except i seem to find the browser in ICS only a tiny bit faster and then only on some sites and then only whilst doing the initial render. After that render I find ICS to be much faster, even than OC HC.
I'm sticking with stock unrooted ICS for now and won't be doing anything different until somebody creates an OC kernel for it.

[Q] Is the performance of ICS ROMs slow for everyone else too?

I have tried a couple of AOSP-like ICS ROMs (HyperNonSense, Virtuous Inquisition 4.0) and I am not very happy with the performance. Pressing the back button often causes a small stutter or lag, and browsing (although fast) has lag on occasion or stutters.
It is not that the experience is 'bad', but it is not as fluid and fast as one would expect it to be from a dual-core phone with the kind of specs included within.
To give you a better idea, my Samsung Galaxy S (i9000), running the BlueLighting ROM (ICS 4.0.3) with a 'mild' OC to 1.2 GhZ runs buttery smooth, and beats my Sensation XE (ICS 4.0.3, tested at 1.2 GhZ 'underclock, 1.5 Ghz 'stock' and 1.7 Ghz 'OC') every time and in every situation: Browsing pictures, using stock browser, browsing applications, contacts, pressing the back button etc.
Why does a phone with a single-core CPU, clocked lower, and what seems to be inferior specs outperform a phone that should be doing much better?
I love AOSP experience, and I would ideally avoid any Sense-ROMs (although Sense-lite are ok, I suppose).
Can anyone tell me what their mileage is, with which ICS ROM, and whether you think there is a ROM I should try?
Thanks!

I also had the Virtous Inquisition Rom on my Sensation and i have to say it works very well.
Maybe you have to try another aosp rom like the Cyanogen mod9 from ultimate sensation,it also works very good for me

Thanks I will give it a try and let you know!
Also, can anyone suggest a good ICS kernel? I have had trouble finding a stable kernel. Some can work fine at speeds above 1.6GhZ, but some will not even boot -- perhaps they are not for ICS. I will have to look around. I am surprised there is (no offence!) little variety and option on offer in regards to kernels -- particularly for those interested in the more exotic CPU profiles.

I have also used both Hyper NonSense (1.5 currently, no longer being developed but is stable) and early Virtuous Inquisition releases. For me, both were quite fast - Hyper NonSense more so, but I haven't used a recent release of VI. In both cases, I stuck with the stock kernel distributed with the ROM.
FWIW, Hyper NonSense has also been giving me EXCELLENT battery life. It's a shame to see development stop on it.

Your Experience - Best performing rom?

Ok so I am currently running CleanRom 2.7 Oc'd to 1.8 and without Fsync disabled. I also use the dock for battery & email purposes.
With this combo I honesty expected alot more from this tablet. It just seems a bit laggy. With a 1.8ghz Quad core processor & a streamlined OS, this thing should be FLYING! What am I missing? My GS3 on CM10 runs circles around this tablet. I know it has an extra gig of ram, but I wouldn't think there would be such a giant difference.
So, in your experience - what is the fastest most efficient rom with out lag? CM10? Another Cleanrom? Other Rom?
Please give me your opinion & any other suggestions. Thanks

EyeAndroid said:
My GS3 on CM10 runs circles around this tablet.
Click to expand...
Click to collapse
We've been over this a thousand times on these forums.
The TF700 has a resolution of 1920 x 1200
The SGS3 has a resolution of 1280 x 720
More pixels means more lag. Regardless of the rest of the hardware. You want a faster tablet, get a low-resolution one.

EyeAndroid said:
Ok so I am currently running CleanRom 2.7 Oc'd to 1.8 and without Fsync disabled. I also use the dock for battery & email purposes.
With this combo I honesty expected alot more from this tablet. It just seems a bit laggy. With a 1.8ghz Quad core processor & a streamlined OS, this thing should be FLYING! What am I missing? My GS3 on CM10 runs circles around this tablet. I know it has an extra gig of ram, but I wouldn't think there would be such a giant difference.
So, in your experience - what is the fastest most efficient rom with out lag? CM10? Another Cleanrom? Other Rom?
Please give me your opinion & any other suggestions. Thanks
Click to expand...
Click to collapse
If I were you, I'd go to CRomI 3.1 Odex with disabled fsync (seriously: there are no risks 'cause the stock kernel is as stable as a rock).
I too am an SIII user (with 1 gig of RAM) but I really do think the tf700 isn't slow at all.
Maybe you forgot to update your bootloader when going form ICS to JB?
I am getting near 7000 on Quadrant (which is not really a good benchmark app, but it gives an idea of what scores you'll be able to get with the tf700.

I can only compare the original ROM to CleanROM and CROMi haven't tried other ROMs.
No dock here either so I wouldn't have input there.
My experience with the stock ROM was a heartache.
Jelly Bean just didn't come thru as promised far as I could see.
Since going thru the learning process to flash CleanROM\Clemsyns, then continuing on with CROMi there was no looking back.
This I consider a complete solution to my disappointment with the Infinity.
Almost returned the tablet to Amazon so glad I kept it now.
Not sure why some stock tabs run with such lagginess.
Oddly I fiddled with one in a store just 2 days ago and it was very smooth.
My recommendation would be an obvious one.
*Be advised once you start ROMing you may get hooked

Any Lagless JB Roms Out There?

Hey folks,
Can you cool kids recommend any lagless JB roms for the Xoom? I've flashed a few but they were unbearably laggy even with OC'ing. Any help you can provide will be appreciated.

I am using EOS4 Nightly 203 with ElementalX kernal v08 and it seems pretty good to me. I only OC to 1200.

Hi,
I use CyanogenMod 10.1 Nightlies. Smooth for the majority of the time, I haven't overclocked.
The reason Google stopped supporting our Xooms at ICS was partly due to its lack of hardware capability so it's never gonna be perfect!