[APP] RealBench

Search This thread

yaocheng

Senior Member
Sep 26, 2007
69
0
JPC rooted with Ryanza lagfix 2.3

RealBench 1.4.1

Last Minute FPS: 55.00
Total Average FPS: 53.33
Longest Period without frames: 1.08sec
 

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
i think it will be better to test both read and write time, with different file size

Goal of this benchmark was to create a benchmark that tries to mimic real phone use. It maybe does it a bit too well though, because the results seem to be about as random as actually using the phone...

I think I might go thw other direction and make a SynthBench that tests everything individually.

EXT4 is a journaled filesystem..so the file has to be written and then the journal, that means the disk is accessed 2 times wile one write process
the EXT2 system is non-journaled so just the file has to be written, just one write access

it is possible to run ext4 without a journal, but i do not know which filesystem is faster

I understand that, but should EXT4 have well under half the performance of EXT2 for writing images to disk??
 

jodue

Senior Member
Dec 15, 2009
453
37
vienna
first of all, cool idea to make a more realistic benchmark as most others are just completely irrelevant for daily use. thank you.

i gave it a try (1.4.1) but it didn't terminate in 25 minutes...?! at the end of the list it just bounced over and over again?

i'm on JM2 + original lagfix by paul (/data/data on /dbdata/data)

i have some other ideas for a real benchmark:
* open intents (activities)
* open some rendering/javascript intense websites in a browserview
 

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
first of all, cool idea to make a more realistic benchmark as most others are just completely irrelevant for daily use. thank you.

i gave it a try (1.4.1) but it didn't terminate in 25 minutes...?! at the end of the list it just bounced over and over again?

i'm on JM2 + original lagfix by paul (/data/data on /dbdata/data)

i have some other ideas for a real benchmark:
* open intents (activities)
* open some rendering/javascript intense websites in a browserview

It never terminates, it runs till you stop it. You can see the score as it's going at the top. It's still really rough though. If I have the time, plan is to add a whole lot of different benchmarks, with a free google appspot server for collecting people's results, and then a score browser in the app so that you can compare your device with others.

Anybody who wants to help out with
1) Benchmarks in C code/java code
2) Ideas for benchmarks
3) Anything else you can think of

Go right ahead and post it in this thread. If you're interested in (1) and need a SVN/GIT to upload code to, just post in this thread and I'll set it up. :)
 

ykkfive

Inactive Recognized Developer
Jul 25, 2010
3,218
3,016
ykkfive.blogspot.com
It never terminates, it runs till you stop it. You can see the score as it's going at the top. It's still really rough though. If I have the time, plan is to add a whole lot of different benchmarks, with a free google appspot server for collecting people's results, and then a score browser in the app so that you can compare your device with others.

Anybody who wants to help out with
1) Benchmarks in C code/java code
2) Ideas for benchmarks
3) Anything else you can think of

Go right ahead and post it in this thread. If you're interested in (1) and need a SVN/GIT to upload code to, just post in this thread and I'll set it up. :)

i just have random read/write in my mind with file size choosen by the users, eg 1k, 512b, 10M, etc, just like others
and, probably the read write time for "each" partition
 

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
i just have random read/write in my mind with file size choosen by the users, eg 1k, 512b, 10M, etc, just like others
and, probably the read write time for "each" partition

Partitions other than /data need root... I guess the app should have some tests that are only available if your phone is rooted.
 

lsevald

Member
Aug 2, 2010
22
0
JM5 rooted with Voodoo beta 1

RealBench 1.4.1

Last Minute FPS: 53.17
Total Average FPS: 51.50
Longest Period without frames: 0.29sec
 

bajica

Senior Member
Apr 7, 2010
356
100
Osijek
RealBench 1.4.1

JM7 voodoo fix
C FPS:54
LM FPS:33,65
TA FPS: 32.85
LPwf: 1.39sec

ran it all the way, stopped scrolling, no more list :)
 

Segitz

Senior Member
Jun 23, 2007
78
12
Stuttgart/Konstanz
Bog standard (recently formatted) vanilla JM1 (new) with few apps installed (and taskkilled everything before benching) on international GT-i9000

lm_fps = 16.87, 8.50, 8.33, 10.68, 6.42
ta_fps = 13, 12.66, 11.25, 11.00, 9.98
lp w/o f = 2.90, ?, 3.65, 3.75, 3.75

Ran the test 5 minutes


As you can see... it WILDLY differs because it really matters WHEN you take your numbers... as a viable test, this is not yet working.


BUT... the wildly varying numbers also indicate exactly the problem... it SHOULD be much more leveled, but it isn't.
 

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
Bog standard (recently formatted) vanilla JM1 (new) with few apps installed (and taskkilled everything before benching) on international GT-i9000

lm_fps = 16.87, 8.50, 8.33, 10.68, 6.42
ta_fps = 13, 12.66, 11.25, 11.00, 9.98
lp w/o f = 2.90, ?, 3.65, 3.75, 3.75

Ran the test 5 minutes


As you can see... it WILDLY differs because it really matters WHEN you take your numbers... as a viable test, this is not yet working.


BUT... the wildly varying numbers also indicate exactly the problem... it SHOULD be much more leveled, but it isn't.

Yep -- the wildly varying numbers are not really fixable. This benchmark is trying to duplicate a phone under very heavy stress, such as when downloading an app in the background while scrolling through your facebook, etc.

I believe the app is doing a very good job at this though, and the numbers really do seem to correlate roughly to real world performance. As can be seen, voodoo has a very large effect on this benchmark, and so does OCLF/EXT2. OCLF/EXT2 actually produces higher FPS because of it writes much faster than voodoo. However, you will notice that OCLF occasionally pauses for up to 5 seconds - this is the buffer being overrun and EXT2 being forced to write to disk through RFS.

So I'm marking this benchmark as a success even at this stage. If you want to know if a change you have made to your phone will translate to better real world performance, this is probably the benchmark to use.

Of course, it still needs a lot more work, but it is already useful. :)
 

drunkeneti

Member
Aug 27, 2010
28
2
I understand that, but should EXT4 have well under half the performance of EXT2 for writing images to disk??

The performance difference is hugh but when the storage is slow in burst writing then it is possible....two years ago i bought a netbook with a ssd and it was incredible slow with a journaled filesystem..
Nevermind, your realbenchmark is a really cool idea...maybe some additional database querys to stress even further, but i do not really know how often the database is used in everyday use...

Sent from my GT-I9000 using XDA App
 

meanm50

Senior Member
Apr 14, 2009
156
9
PDX
Good work!

Droid Eris KaosFroyo V32
Last Minute FPS: 14.67
Longest periord w/o frames: .95sec

App F/C every 5 uses...only after about 30 or 40 seconds though.
 

LostAlone

Senior Member
Jul 15, 2010
164
0
Im running JM5 with Samset 1.9 and I'm seeing seriously random results.

Depending on my phones mood I get 18f/min and 0.09 second with no frames , or 0.12 f/min and 14 secs with no frames. Kinda weird.

Just as a thought, if you could add a force close button, it'd be appreciated. I find that if you try and close it when its stopped (ie not scrolling) then it can take a long time to get rid of it. I know thats when its at max load and such, but I'm finding that it won't do anything for far longer than just the 'time without frame' time. Either that or it needs to stop by itself sometime :p
 
Last edited:

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
Ryan you want me to toss this up on the market too?

Nah, it isn't really that useful yet. Eventually it'll get there. :)

I'm starting to think that when supercurio finishes up his lagfix, I'll make an EXT2 version and just not pull out the battery when the phone is on. :D

If it works on the iphone, it should work fine here.
 

RyanZA

Senior Member
Jan 21, 2006
2,023
784
JHB
Im running JM5 with Samset 1.9 and I'm seeing seriously random results.

Depending on my phones mood I get 18f/min and 0.09 second with no frames , or 0.12 f/min and 14 secs with no frames. Kinda weird.

Just as a thought, if you could add a force close button, it'd be appreciated. I find that if you try and close it when its stopped (ie not scrolling) then it can take a long time to get rid of it. I know thats when its at max load and such, but I'm finding that it won't do anything for far longer than just the 'time without frame' time. Either that or it needs to stop by itself sometime :p

There isn't a force close option for a reason! ;)
It makes about 35mb of files on disk, and when you close it, it removes those files.
Force close would leave that 35mb behind.

It should probably run for 2 minutes and then stop and give you a 'final score' I guess. :D