Develop Bugs

[Benchmark] 3D/OpenGL Performance Across ROMs [28/07/10]

Hi.
I'm have now rewritten the Spreadsheet and will shortly be integrating my CPU benchmarks and GPU benchmarks into the same sheet (not really that interesting or anything but it's there if you want to see exactly how things are worked out) but it explains why the progress has been slow recently. I have now taken this thread over from Arkymedes - and a HUGE thanks to him for creating the spreadsheet and doing most of the benchmarks which are still being referenced in them. He has done a brilliant job!
This page will continue to keep the tests strictly to GPU performance. CPU performance can be viewed here.
(like Arkymedes) I will take in consideration the clock used in each ROM and try to use the maximum allowed by the kernel using SetCPU. HOWEVER, all rom's will be tested at 691 MHz for consistency and any other frequencies will be detailed.
About JIT, unfortunately, all Arkmedes tests show that it doesn't affect 3D performance YET, since all this apps are not properly optimized and didn't take JIT into account. However, this may change with Froyo builds so keep an eye out.
All the ROM's clocked @ 691MHz "Performance" under SetCPU, unless specified otherwise. All tests are conducted 5 times and an average is taken. The tests are carried out in Aeroplane mode with no widgets running on the desktop. No apps are installed aside from the benchmark apps and Apps2SD is used (if enabled by the ROM). The latest version of the benchmark software is always used (they usually only add support for new phones in updates rather than changing benchmark parameters).
Benchmark apps:
Neocore - Qualcomm's own benchmark tool for 3D performance. Measured in FPS
Benchmark - Softweg's test suite for benchmarking. Only the GPU test done here to account for 2D performance. Measured by an internal score.
An3DBench - Tests based on jPCT, a free 3D engine for Java. 7 tests showing the capabilities of the engine. Measured by an internal score.
Note that all the tests mean: The higher the score, the better.
Anyway, here it goes
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}

A little suggestion: you should specify wether whe rom is sense or vanilla, and also the CPU clock used for the tests.

Can't wait!

that's a long couple of hours.......

Are you going to do VillainROM 10.2 with Sense + JIT enabled? If not I can try and do a benchmark.

Thread updated

Please Note:
I believe in the EzHero 2.7 stats, that jit is not enabled, with jit EzHero 2.7 is equal or greater than EzHero 2.0

maybe link each of the roms in the first thread for people to easily find...being quite new here i cant fins the Androbin-cr2-5.0.7...what ever that one is...

Your efforts are highly appreciated, good job.
But there is a problem: you didn't say which ROM is JIT enabled and which is not, because as you see CPU benchmarks have better results on JIT enabled ROMs, even though, as you can see, 3D is not affected.

My bench marks
My benchmarks with EzHero 2.7+ JIT:
Linpack- 4.886
Neocore w/ sound- 30.2FPS
Neocore w/o sound- 33.2FPS
Quadrant Standard- 280
Benchmark Pi- 7222 milliseconds
Very good job on the post, well done keeping us informed!

Hello!
What I'm trying to do here is to benchmark any version that I can find in these forums WITHOUT any further optmization. The benchmarkings are done OUT OF THE BOX
So, if I have to MANUALLY enable JIT or any OC AFTER the flashing, it will NOT show in these graphics!
I didn't do that because it is just impossible for me to keep track of every single optmization that can be done to every single one of this ROM's. I just don't have the time and resources for that.
So again, all this tests are done straight after the flashing of the downloaded ROM (logically, after some reboots and bla bla bla).
I really hope that you guys understand that. It's not a way to show which ROM is better, since there are MANY more factors that dictate that, like BT calls, GPS and so on. It's just a way to measure mostly 3D performance out of the box with the ROM's you provide.

sheridan2000 said:
maybe link each of the roms in the first thread for people to easily find...being quite new here i cant fins the Androbin-cr2-5.0.7...what ever that one is...
Click to expand...
Click to collapse
That's a good idea, but I guess the WIKI sticked in the first page should have done the job. It will be redundant but I can try to keep track of them.

Man sad day, all you have to do in EzHero to enable jit is go to the terminal app and type jit-on.
Still, job well done! Kudos!

xenoside said:
Man sad day, all you have to do in EzHero to enable jit is go to the terminal app and type jit-on.
Still, job well done! Kudos!
Click to expand...
Click to collapse
I will add EZHero 2.7 with JIT enabled after Italy x Paraguay

Arkymedes said:
I will add EZHero 2.7 with JIT enabled after Italy x Paraguay
Click to expand...
Click to collapse
World Cup always comes first my friend lol

its raining pretty hard in the game...
anyways, so 2.0 is JIT enabled and 2.7 is not in your graph? because huge difference in linpack scores with the older having a better score.
you should also mention Benchmark Pi - lower the score, the better.

-2y2k- said:
its raining pretty hard in the game...
anyways, so 2.0 is JIT enabled and 2.7 is not in your graph? because huge difference in linpack scores with the older having a better score.
you should also mention Benchmark Pi - lower the score, the better.
Click to expand...
Click to collapse
Yeah, 2.0 is jit enabled, but now 2.7 is optional due to complaints and split decisions. I posted the benchmarks with jit on the last post of the front page

im a Sense guy but its been awhile since I tried a non-sense one, your ROM is rrreally tempting right now lol

-2y2k- said:
im a Sense guy but its been awhile since I tried a non-sense one, your ROM is rrreally tempting right now lol
Click to expand...
Click to collapse
Well vanilla Eclair and EzHero are both great roms, heck even Androbin is tied with those two.

In the Quadrant Graph, androbin doesn't have his bar.

[Q] (SOLVED!) Overclock the Defy?

*SOLVED*
Milestone Overclock - a free download from the marketplace does now support the Defy. Quadrant shows promising figures for sure.
Just wondering whether anyone has gotten the Defy to Overclock?
I've been trying some Droid Overclock apps as well as purchasing Set CPU.
I've successfully rooted my phone which is a requirement for enabling overclocking features. Set CPU won't even start.
None of the apps I've tried have been working properly.
Any tips?

in developement
Look on the milestone overclock googlecode page, its on the todo list

good to know

Yepp, thanx a lot. Since the chipset has clear OC potential this will be very interesting. Especially if the Defy has an omap3630 core, we'll see.

dave79 said:
Yepp, thanx a lot. Since the chipset has clear OC potential this will be very interesting. Especially if the Defy has an omap3630 core, we'll see.
Click to expand...
Click to collapse
I'm pretty confident that it does indeed have a 3630: http://forum.xda-developers.com/showpost.php?p=9540564&postcount=34
Not only because of the screenshot from windows, but look at how close it is to the Droid X, running a 1ghz 3630.

Wow! and getting that benchmark score while running on 800MHz is quite impressive. I've got good hopes that a ported Milestone Overclock will unleash the raw power the phone truly possess.

And this is not the best defy can do. Actually 1117 has been reported from 2 users (one is me).

Im not sure how this phone is so amazing, and so cheap at the same time
umm yeah anyways, just wanted to say i got quadrant scored of around 1110 to 1118 ish i think- after task killing
Really hoping for some custom 2.2 roms in the near future, like some others said it will make this probably -one of- the best androids out there

xdaid said:
And this is not the best defy can do. Actually 1117 has been reported from 2 users (one is me).
Click to expand...
Click to collapse
My personal best is 1104

{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}

Whoa, nice one, Higgsy!

Sorensiim said:
Whoa, nice one, Higgsy!
Click to expand...
Click to collapse
weeeeell to be fair there was a little cheat there, that was with JIT enabled. But JIT by the standard method wasn't stable, at first it was fine but then after ab out an hour I started to get random force closes, so I reverted. Its a little insight to what can be achieved with JIT. IMO I'll take JIT over a 1GHZ overclock.

Could anyone of you guys share your benchmark results after running Milestone Overclock? I'm now running on 1.2GHz and it's stable

xdaid said:
And this is not the best defy can do. Actually 1117 has been reported from 2 users (one is me).
Click to expand...
Click to collapse
My first benchmark result was in mid 1000 but after a restart I got that score too.

To use JIT is better wait for froyo because have it in native form. There are a appl called JITTER in the market but I think that isn't reliable. (I don't know about other appl similar than)
Now I use the Milestone Overclock beta4 and configure the oc at 1000/56
Someone have better configuration to use? Can explain why?

Hey guys.. Overclocked my Defy with the Milestone Overclock. 1200/66
Got a 1438 benchmark on Quadrant Standard. Putting the vsel any lower got my Device to reboot.
After trying it out I switched back to 1000/52 because some users said that running on 66vsel could damage the CPU.
Main effect I had: Launcher runs very smooth, My battery life went down to 24 hours (37h before).
With 1000/52 I still get a benchmark around 1247
Works for me because I´m saving battery and still got enough power to have a smooth launcher.
But: Can anyone please tell me what the standard settings are ? *lol* Forgot them 800Mhz an vsel=???
Thx in advance.
Daniel (from germany)

Daniel.Stevens said:
Hey guys.. Overclocked my Defy with the Milestone Overclock. 1200/66
Got a 1438 benchmark on Quadrant Standard. Putting the vsel any lower got my Device to reboot.
After trying it out I switched back to 1000/52 because some users said that running on 66vsel could damage the CPU.
Main effect I had: Launcher runs very smooth, My battery life went down to 24 hours (37h before).
With 1000/52 I still get a benchmark around 1247
Works for me because I´m saving battery and still got enough power to have a smooth launcher.
But: Can anyone please tell me what the standard settings are ? *lol* Forgot them 800Mhz an vsel=???
Thx in advance.
Daniel (from germany)
Click to expand...
Click to collapse
Someone in the other thread was running 800/38, I haven't tried it because like you 1000/52 gives me better battery life than stock and more power on top.

Higgsy said:
Someone in the other thread was running 800/38, I haven't tried it because like you 1000/52 gives me better battery life than stock and more power on top.
Click to expand...
Click to collapse
Thanks for answering ;0)
I just wanted to know the original settings in case I wanted to restore them someday.
Or do I just have to click on "unload module" for that ?!? (I gues that´s it I was just to blind/dumb to see it *lol*)
Thx ;0)

Daniel.Stevens said:
Thanks for answering ;0)
I just wanted to know the original settings in case I wanted to restore them someday.
Or do I just have to click on "unload module" for that ?!? (I gues that´s it I was just to blind/dumb to see it *lol*)
Thx ;0)
Click to expand...
Click to collapse
Yeah just unload module and untick apply on boot and your back to stock

Im currently running 1100MHZ with a 58vsel, seems stable so far and got a quadrant benchmark of 1317. But i am wondering if this setting is safe for long term use, I need my phone to last till the end of this conract

[Q] Fastest Sense / Non-Sense Rom

I want to know what the absolute fastest rom is for the EVO...Sense or Non-Sense...don't care, and with everything working.
*Just to be clear......not asking what the best rom is. Not too sure I care about Quadrant scores either. Looking more at how it handles transitions, boot up, opening and closing apps, the whole enchilada.
If you leave a comment can you please also let me know what kernel you may have paired with it.
Thx.

cm6 or lithids GB builds for aosp, i like the cyanogen kernals and savage's too
mikfroyo 4.1 with netarchy's 4.2.2 for sense or ziggy's latest
u can overclock both roms for the speedfreak in ya
also use spare parts to speed up transitions

cmart4 said:
cm6 or lithids GB builds for aosp, i like the cyanogen kernals and savage's too
mikfroyo 4.1 with netarchy's 4.2.2 for sense or ziggy's latest
u can overclock both roms for the speedfreak in ya
also use spare parts to speed up transitions
Click to expand...
Click to collapse
Sorry, just edited my post because I wanted to include that the rom should have everything working.
I know CM6 now has that but what about mikfroyo?

Just head over to the development forum and check it out. Mikfroyo has everything working, and without a doubt is the fastest sense rom out there. Several other developers are using Amikam's base rom in their own work.

Hands down MikFroYo v4.1
I've been flashing ROMs for two years now. I have to say out of the box without doing any mods, MikFroYo v4.1 is the fastest ROM I've every used. Everything works, as long as you wipe everything and start from scratch. It is a Sense based ROM. I used to flash CM ROMs just to have a faster running Hero. Now with MikFroYo v4.1 on a EVO there is no loss in speed. Have fun and good luck choosing. I'm running MikFroYo v4.1 with launch pro and it is a beautiful thing.

blane73 said:
I've been flashing ROMs for two years now. I have to say out of the box without doing any mods, MikFroYo v4.1 is the fastest ROM I've every used. Everything works, as long as you wipe everything and start from scratch. It is a Sense based ROM. I used to flash CM ROMs just to have a faster running Hero. Now with MikFroYo v4.1 on a EVO there is no loss in speed. Have fun and good luck choosing. I'm running MikFroYo v4.1 with launch pro and it is a beautiful thing.
Click to expand...
Click to collapse
Can I ask what kernel you prefer with it?

CCallahan said:
Sorry, just edited my post because I wanted to include that the rom should have everything working.
I know CM6 now has that but what about mikfroyo?
Click to expand...
Click to collapse
mikfroyo will not let u down, it all works, plus the dev is constantly tweaking it even more; he supports his work bigtime and answers every question fast

Anyone using the latest SBS kernel with this? How does it run? Is battery life good?

Batt life is good. On a easy day with not to much action you can get 18-20 hrs. On a heavy day maybe 10-12. But thats what home, computer, and car charges are for.

Im considering Mikfroyo 4.1 too, but i was wondering if anyone knows of or can post a video of it. How similar is the UI to the stock Evo UI?
thanks

ctbenk said:
Im considering Mikfroyo 4.1 too, but i was wondering if anyone knows of or can post a video of it. How similar is the UI to the stock Evo UI?
thanks
Click to expand...
Click to collapse
Don't need a video. I just flashed 4.2 and there are very few changes to the stock UI. Looks basically the same with exception to a few additional installed apps, power notification drop down, maybe a style change on something. There are also some stock apps not included. But, is way faster than the stock one.
It is fast for a Sense rom but still can't believe an ASOP isn't the fastest available.

I've been getting around 2500 in quadrant and 39.5 consistently using CM 6.1.1 with netarchy's 7.6 SNAP BFS. I'm also running the iNexus 6.1.1 theme.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Sent from my PC36100 using XDA App

lawsonmh15 said:
I've been getting around 2500 in quadrant and 39.5 consistently using CM 6.1.1 with netarchy's 7.6 SNAP BFS. I'm also running the iNexus 6.1.1 theme.
Sent from my PC36100 using XDA App
Click to expand...
Click to collapse
That's good but I'm not too interested in benchmarks. Even Cyanogen has said they are trivial and not accurate most the time.
What I am interested in is user experience. How is CM6 when doing transitions, opening apps, etc? Mikfroyo right now, even with the sense, is pretty darn smooth and fast.
Or, maybe even throw Vaelpak in there. It seemed pretty quick when using. Any other seriously fast roms out there?
Would really like to hear an opinion from someone who has used CM and maybe Mikfroyo. Even if you don't like Sense (i don't) it can be removed from any rom.

yea i used quad on my evo when it was stock and was getting 1250-1300 consistantly, and my phone was getting slow, jerky, and unresponsive. Rooted and put mikfroyo on and im amazed at how much more smoothly it runs. Funny thing is it benchmarks at or lower then it did before, yet in real world stuff its obviously faster.

6.1.1 has been very smooth and I can honestly say that after enabling snap turbo, it is far and away the fastest tom i've used.

Fresh 3.5.0.1 w/ toast 4.1.9.1 was faster than mikfroyo 4.2 for me... CM 611 W/ Snap Turbo blows them both out of the water. Also, if benchmarks don't mean anything, nobody would develop them. If I was scoring an 8 on quad w a samsung, i would attempt to convince myself of that, but in reality, i'd just exchange it for an htc.

Nonetheless, for the rest of the weirdos who do use benchmarks to assist w choosing a rom or kernel, you may find the following beneficial: http://home.comcast.net/~ttechnology/compiled.html

lawsonmh15 said:
Fresh 3.5.0.1 w/ toast 4.1.9.1 was faster than mikfroyo 4.2 for me... CM 611 W/ Snap Turbo blows them both out of the water. Also, if benchmarks don't mean anything, nobody would develop them. If I was scoring an 8 on quad w a samsung, i would attempt to convince myself of that, but in reality, i'd just exchange it for an htc.
Click to expand...
Click to collapse
How was your battery life while on CM 611 compared to the other roms?

Comparable since i use cpu tuner as well and config'd it to max @ 400ish on screen off.

Rom benchmark comparison

This is my compilation of Antutu Benchmarks...I've done many others so expect more.
- All ROMS benchmarked at 768MHz except stock V...set to Airplane mode except Sun Spider
- Ran 3 of each benchmark...I never ran them after each other so no caching.
- Swap partitioned to 128Mb/512Mb fs .... I've shown that partitioning doesn't matter in my other post (check below)
- Phones only used the benching programs & a screenshot app
This took awhile, probably greater than 2 weeks as I restarted again and again when I found better ways to set up the phone, but I like to see actual numbers to back up reasoning...more to come!!!
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}

Swap size benchmark comparison
Here's my swap size experiment. Hope it answers just one persons thoughts on swap size.
http://forum.xda-developers.com/showthread.php?t=1105503

Newb here, where can I find Zerfies CM7? Did a forum search and I'm not seeing it. TIA

Thomas S said:
Newb here, where can I find Zerfies CM7? Did a forum search and I'm not seeing it. TIA
Click to expand...
Click to collapse
Over at android central under optimus s
Sent from my LG-VM670 using XDA Premium App

smokenbiskits said:
Here's my swap size experiment. Hope it answers just one persons thoughts on swap size.
http://forum.xda-developers.com/showthread.php?t=1105503
Click to expand...
Click to collapse
Is benchmarks really that important does it have real world usage?
Sent from my LG-VM670 using XDA Premium App

vegetajin8 said:
Over at android central under optimus s
Sent from my LG-VM670 using XDA Premium App
Click to expand...
Click to collapse
What issues might pop up when using a ROM for an Optimus S phone instead of our Optimus V? I know the buttons are swapped. Anything else?

482123 said:
What issues might pop up when using a ROM for an Optimus S phone instead of our Optimus V? I know the buttons are swapped. Anything else?
Click to expand...
Click to collapse
There's a sticky at android central from mmarz in the optimus v section that will explain how to port an s rom to the v.
Sent from my VM670 using XDA Premium App

it's a comparative analysis, if it's better in benchmarks then it will generally run faster in real life

That is really cool to see ^.^
Thanks !!!!

Thomas S said:
Newb here, where can I find Zerfies CM7? Did a forum search and I'm not seeing it. TIA
Click to expand...
Click to collapse
You don't want that. Better off getting aospCMod.

Hey what filesystem are you using when doing these benchmarks? Also on the stock ROM benchmark on quadrant advanced, are you using the stock unrooted kernel? because I have a stock unrooted v that has never had anything flashed to it and im getting like 480's.

Unreliable
You can't judge a phone when it comes to anything based off of CM7. CM7's stage fright is unstable and basically, skips some of the initial tests and marks them as completed. This causes highly inflated scores. All the 2.2 or non CM7 based ROM's will post more accurate and relevant scores. I have tried almost all of these ROM's and Bumble Bee is the winner for me. Off of real performance speeds, not Quadrant Scores..

nikeman513 said:
You can't judge a phone when it comes to anything based off of CM7. CM7's stage fright is unstable and basically, skips some of the initial tests and marks them as completed. This causes highly inflated scores. All the 2.2 or non CM7 based ROM's will post more accurate and relevant scores. I have tried almost all of these ROM's and Bumble Bee is the winner for me. Off of real performance speeds, not Quadrant Scores..
Click to expand...
Click to collapse
True, but that only applies to benchmarks that use stagefright enabled tests, such as Quadrant. He ran his tests on several different benchmarking tools. Besides, we all know benchmarks don't matter, right?

nikeman513 said:
You can't judge a phone when it comes to anything based off of CM7. CM7's stage fright is unstable and basically, skips some of the initial tests and marks them as completed. This causes highly inflated scores. All the 2.2 or non CM7 based ROM's will post more accurate and relevant scores. I have tried almost all of these ROM's and Bumble Bee is the winner for me. Off of real performance speeds, not Quadrant Scores..
Click to expand...
Click to collapse
I seem to have better performance out of cm7. I play crusade of destiny every now and then. It lags a bit on bumblebee and other 2.2 rpm's, but its pretty smooth on cm7.
Sent from my VM670 using tapatalk

thank you so much for doing this work man. This is exactly the kind of critical information that a newer user like me needs. In fact I consider this one of the more critically important posts / collection of data that I have come across in my searches for the ROM that will best fit my needs.
I've been torn between a few ROM's (Stock V, Harmonia, Bumblee, Gingerbread based ROM) and this data you've provided will allow me to pick a ROM based on 1) the features I think I will need 2) it's performance level relative to other ROM's.
That's not to say some developers might not leapfrog each other in performance, but it gives me a good idea of who is on a path that I also wanna walk on. And right now that's looking like Harmonia to me (maybe Bumblebee as a trial). But I LOVE the absolutely stripped down nature of Harmonia 1.3.1 on my LG OV.
again, thank you so much for this info.

Interesting. I'd love to see you test Jerry's or KemikalElite's ROMs. On a straight IHO build ondemand @ 245-729MHz I'm getting into the 1500s on AnTuTu, low 900s on Quadrant[1], high 9s on Linpack single threaded.
1: Quadrant seems to be rather dependent upon minimum CPU speed here, as bumping that up a bit got me into the 1100s.

Quadrant scores should be completely thrown out. It's easy to optimize code for the benchmark. Acer released an official patch that did just that for some of their Android phones.
Smartbench may be a better choice. I also would like to see one of the I.H.O. builds tested against the other ROMs.

Vishwar said:
You don't want that. Better off getting aospCMod.
Click to expand...
Click to collapse
Why doesn't he want that I was thinking about getting the same 1
Sent from my LG-VM670 using XDA Premium App

cjtulowiecki said:
Why doesn't he want that I was thinking about getting the same 1
Sent from my LG-VM670 using XDA Premium App
Click to expand...
Click to collapse
At this point no one should be downloading AOSP due to its lack of support and general lack of being a good ROM. Bumblebee (2.2, most stable, super fast, great battery life, etc, but stock), Harmonia (2.2 with a lil more features), Ath3nos (most stable 2.3/CM7 so far) and Inferior Human Organs (2.3, CM7.1, still some issues but wide open source, actual bug logging, etc.)

DustbinK said:
At this point no one should be downloading AOSP due to its lack of support and general lack of being a good ROM. Bumblebee (2.2, most stable, super fast, great battery life, etc, but stock), Harmonia (2.2 with a lil more features), Ath3nos (most stable 2.3/CM7 so far) and Inferior Human Organs (2.3, CM7.1, still some issues but wide open source, actual bug logging, etc.)
Click to expand...
Click to collapse
I agree with most of this, except whether Ath3nos is most stable, it's based on an earlier source of IHO, and IHO had a lot more fixes since then. I haven't had any issues with IHO (except the documented ones already, a shrinking list), maybe because it's better documented it seems easier to pick at the details. It should be THE Gingerbread/CM7.1 to use for Optimus V right now.

[Kitkat] Defy(+) CM11 Performance Tests

Hi guys,
Following the development of CM11 for the Defy(+) by @Quarx, I decided to share the evolution of performance "Nightly".
The results were obtained from the average of 3 tests for each "Nightly", using two app for measuring performance: Quadrant and AnTuTu.
Configuration for testing:
CM11 Nightly (by @Quarx)
Defy+ (MB526)
CPU (OC): 300/20 600/35 900/50 1200/64
CPU (SC): 300/33 600/48 800/58 1000/62
Governor: Interactive
I/O: Deadline
Installation conditions: No clean install (version over version)
TWRP > Advanced Wipe > Dalvik Cache/Cache
Install > Nightly/GApps > Wipe Cache/Dalvik > Reboot System
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Added test on SC (Stock Clock) at the request of @Blechd0se
* Random reboots occurred frequently in the early versions. For me no longer occurs since version 2014.01.24 (except version 2014.01.28, very problematic);
* Crashes occur on Google Search (maybe because I'm using GEL) and Aero Control in the first 4-5 reboots (it seems necessary to stabilize the unclean installs). After, no more crashes!
* @Quarx and @Blechd0se, the greatest variability between consecutive tests in both apps is about IO (reaching 40-60%), by showing up unstable;
BUG @Quarx (and @okij):
Camera: auto-focus doesn't work using size 3.7M (best aspect ratio - the same of screen);
Shortcuts from Apps moved to SD card disappear after reboot;
Battery Drain: 2014.02.08, 2014.02.09, 2014.02.10, 2014.02.11, 2014.02.12;
Battery charge bugs: 2014.02.18, 2014.02.19, 2014.02.20;
Additonal information:
_ Development-related forum (here);
_ Short Path to clean install (here);
_ Bug list by @okij (here);
_ CM11 by @Quarx (here);
_ BaNkS GApps (here) - use only Banks Minimal GApps;
_ For those interested, following link GEL (Google Experience Launcher);
Special thanks for the improvements in the kernel obtained by @Blechd0se!
Sorry for my bad english

Awesome work, keep with this !

Add
All tests were done with my device using all applications for daily use (added to BaNkS GApps, amounted 159 Apps), connected with the following services: Wi-Fi, Location, 2G/3G and Sync.
good:

Welcome @Saturnal here on xda and thanks for this nice debut, well done! If possible, update your thread (preferably the opening post) when new nightlies are released.
I just announced your thread here on the CM11 development thread, because many users are mainly reading the dev thread and I think that your thread deserves some attention.

Great work, keep up!
Could you provide some subjective experiance about less measurable performance as well? I think the most interesting and important would be wake up time during calls (for me at least). In this manner it seems to be worse over time on my device (I also don't do clean installs).

Great work! For reference I Can aport my numbers in antutu I get 4789 with a Defy Red lens in the last CM10.2

Thanks for the info! It is nice interesting to see the difference between the builds (and between the effect on the different benchmark-types).
How precisely does it make sense to compare these numbers with benchmarks on other Defy's with other installed apps?
If meaningful, I might do a benchmark of my own Defy.
I ask this because I find my Defy on Wiui to bug/crash/slow down a bit too much lately (after a few months of using it) and therefore I'm thinking of trying another ROM. As I read that Kitkat in general was lighter on RAM-usage, I tought of looking at that direction. (I'm not that interested in the latest GUI etc., rather looking for stability, efficiency and performance.)

Very interesting. For what it is worth, I get AnTuTu score 4893 with 0126, and a score of 5074 with 0205 nightly. The phone is overclocked to 1200 MHz.

okij said:
Welcome @Saturnal here on xda and thanks for this nice debut, well done! If possible, update your thread (preferably the opening post) when new nightlies are released.
I just announced your thread here on the CM11 development thread, because many users are mainly reading the dev thread and I think that your thread deserves some attention.
Click to expand...
Click to collapse
Thanks @okij and everyone, I'll update today.:good: Yesterday I had many problems to connect to the forum.

NPeete said:
Great work, keep up!
Could you provide some subjective experiance about less measurable performance as well? I think the most interesting and important would be wake up time during calls (for me at least). In this manner it seems to be worse over time on my device (I also don't do clean installs).
Click to expand...
Click to collapse
Sure! Is fairly smooth transition of screens (I'm using GEL). Keeps bug #31 (screen is half blacked out while getting incoming calls) and #26 (screen needs about 5 seconds to wakeup when a call is receive). Crashes occur on Google Search (maybe because GEL) and Aero Control in the first 4-5 reboots (it seems necessary to stabilize the unclean installs).

Johan_V said:
Very interesting. For what it is worth, I get AnTuTu score 4893 with 0126, and a score of 5074 with 0205 nightly. The phone is overclocked to 1200 MHz.
Click to expand...
Click to collapse
Thanks, but it's clear that I posted averages between 3 tests, so got higher values ​​than those shown in the chart Moreover, different test conditions lead to different values.
:good:Keep doing tests and posting to help all users with more references!

bblubb said:
Thanks for the info! It is nice interesting to see the difference between the builds (and between the effect on the different benchmark-types).
How precisely does it make sense to compare these numbers with benchmarks on other Defy's with other installed apps?
If meaningful, I might do a benchmark of my own Defy.
I ask this because I find my Defy on Wiui to bug/crash/slow down a bit too much lately (after a few months of using it) and therefore I'm thinking of trying another ROM. As I read that Kitkat in general was lighter on RAM-usage, I tought of looking at that direction. (I'm not that interested in the latest GUI etc., rather looking for stability, efficiency and performance.)
Click to expand...
Click to collapse
Makes sense to analyze the variation of the benchmarks on the same device in pre-established conditions. So we can see which version is more optimized for use!
If you think about changing ROM, I strongly recommend CM11 by @ Quarx!:good:

BadFranch said:
Great work! For reference I Can aport my numbers in antutu I get 4789 with a Defy Red lens in the last CM10.2
Click to expand...
Click to collapse
Have got 4853 (AnTuTu) and 1962 (Quadrant), but only posted the averages. Maybe it's time you experience the CM11!

Sent from my Defy+

Saturnal said:
View attachment 2561658
Sent from my Defy+
Click to expand...
Click to collapse
I have red lense Defy with GB kernel and latest CM7. No OC, 1000mhz
Antutu score 3900.
Now what is interesting is that you have everything better exceptdatabase I/O. I have 360 and you 215. Hm, Storage I/O is again on your side and nearly double. Mine 425 against your 922+. What is holding down CM11 in this department I wonder.

It is interesting how much results differ. With 0205 nightly I now get consistently over 5200 with AnTuTu. Of course benchmarks results will not translate to real world performance, but 0205 feels very smooth. Telephony problems are gone too, after the build.prop fix.
On the hand, quadrant scores are lower than Saturnal's, average is 1557.

ezantera said:
I have red lense Defy with GB kernel and latest CM7. No OC, 1000mhz
Antutu score 3900.
Now what is interesting is that you have everything better exceptdatabase I/O. I have 360 and you 215. Hm, Storage I/O is again on your side and nearly double. Mine 425 against your 922+. What is holding down CM11 in this department I wonder.
Click to expand...
Click to collapse
There is great variability in the score Storage I/O. Have found values ​​between 271-956. See a test using Stock Clock 1000MHz.

Johan_V said:
It is interesting how much results differ. With 0205 nightly I now get consistently over 5200 with AnTuTu. Of course benchmarks results will not translate to real world performance, but 0205 feels very smooth. Telephony problems are gone too, after the build.prop fix.
On the hand, quadrant scores are lower than Saturnal's, average is 1557.
Click to expand...
Click to collapse
It's an amazing score. If I get the best value of all tests pass little more than you found, reaching 5235 (AnTuTu).:highfive:

Very good work! Just a suggestion: could u add another benchmark app like basemark?
Enviado desde mi MI 2S mediante Tapatalk

espaciosalter20 said:
Very good work! Just a suggestion: could u add another benchmark app like basemark?
Enviado desde mi MI 2S mediante Tapatalk
Click to expand...
Click to collapse
Thanks! But, there's a great variety benchmarks in market. I lose a lot of time to obtain consistent results. Unfortunatly, I can't comply your request. Moreover, basemark is focused on graphics performance, while our target is the entire hardware performance combined with CM11.:good: