Benchmarking apps for N900 - How fast is your N900? - Page 6 - maemo.org

theonelaw | # 51 | 2010-08-10, 01:09 | Report

Originally Posted by jerryfreak View Post

can someone who understands the data a little better make a simple comparison between various kernel?. i havent been keeping on the bleeding edge with new ones as 800 the mhz kernel works fine for me. im curious if the newer 800 mhz kernels are superior.Ive kinda been not too focused on that cause i wasnt gonna mess with kernels til after pr 1.2 came out, any day now lulz

Yes,
second the motion.

Althought the benchmarks are quite informative,
an 'executive summary' would probably shed a bit more light on the subject
to clarify just what the situation is with the various configs.

something like this, perhaps, keeping it simple I hope.

0. baseline
Stock straight from the store shelf is 500 -600 works ok,
depending on your expectations and firmware update level.
Anybody done any benchmarks comparing PR 1.1 and 1.2 ?

1. Kernel changes
How do the various kernels compare with no settings changes?

2. Clock adjustment
Battery life versus performance.

3. Voltage adjustment
Stability vs battery life

4. Bleeding edges:
Performance, battery life, reliability. (you can have any two of these

)
The bleeding edge kernel with starvation shows that
the very best that can be hoped for as relatively stable
has about 50%-60% 'general' improvement in speed,
(dependent on the application type of course)

Quantifying the differences show whether the time and risk
is worth the achievable performance enhancement,
depending on what the reader wants to achieve.

Hope this helps.

sak500 | # 52 | 2010-08-10, 17:21 | Report

Works great @ 1Ghz CPU / 500DSP but dont knw how to copy and past the text. Dont want to type all the info. How are you guys doing it? DOes the result get saved in some folder?

Ok will post the last ones

Integer Index 22.78
FP Index 3.042

Memory Index 5.486
Integer Index 5.829
FP index 1.687

Last edited by sak500; 2010-08-10 at 17:34.

TiagoTiago | # 53 | 2010-08-31, 01:36 | Report

Originally Posted by jakiman View Post

Well, titan's kernel is limited at a maximum frequency of 1150Mhz.
So there's really no way for people to go faster and get higher scores. =P

Wouldn't people from cold weather get better scores than people in hotter weather?

TiagoTiago | # 54 | 2010-08-31, 01:41 | Report

Btw, would there be an alternative for this benchmarking that would show more than just numbers? LIke somthing with some flashy graphics and stuff

mail_e36 | # 55 | 2010-08-31, 18:04 | Report

Very nice thread, thanks Matan for posting the app for Maemo benchmarking.

maxximuscool | # 56 | 2010-10-11, 10:58 | Report

Mine stock 600Mhz

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST : Iterations/sec. : Old Index : New Index
: : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT : 246.48 : 6.32 : 2.08
STRING SORT : 42.829 : 19.14 : 2.96
BITFIELD : 8.2722e+07 : 14.19 : 2.96
FP EMULATION : 59.896 : 28.74 : 6.63
FOURIER : 1143 : 1.30 : 0.73
ASSIGNMENT : 3.901 : 14.84 : 3.85
IDEA : 659.83 : 10.09 : 3.00
HUFFMAN : 382.74 : 10.61 : 3.39
NEURAL NET : 1.1823 : 1.90 : 0.80
LU DECOMPOSITION : 44.409 : 2.30 : 1.66
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 13.421
FLOATING-POINT INDEX: 1.784
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU :
L2 Cache :
OS : Linux 2.6.28-omap1
C compiler : gcc version 4.2.1
libc :
MEMORY INDEX : 3.233
INTEGER INDEX : 3.439
FLOATING-POINT INDEX: 0.990
Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.
Nokia-N900:/home/opt/nbench#

ShaolinMilk | # 57 | 2010-11-06, 20:47 | Report

I keep on getting permission denied. I tried everything on here and its not working. Been trying for an hour now. It's frustrating the crap out of me!

Tigerite | # 58 | 2010-11-09, 11:30 | Report

A quick update with PR1.3 and BFS kernel, again running at 1.15Ghz starving config:

Code:

/home/user/nbench # kernel-config limits 1150 1150
the limits were set to [1150,1150]
/home/user/nbench # ./nbench

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          477.12  :      12.24  :       4.02
STRING SORT         :          83.407  :      37.27  :       5.77
BITFIELD            :      1.6086e+08  :      27.59  :       5.76
FP EMULATION        :          118.96  :      57.08  :      13.17
FOURIER             :          2238.2  :       2.55  :       1.43
ASSIGNMENT          :          7.6802  :      29.22  :       7.58
IDEA                :          1283.2  :      19.63  :       5.83
HUFFMAN             :          747.44  :      20.73  :       6.62
NEURAL NET          :          2.3392  :       3.76  :       1.58
LU DECOMPOSITION    :           87.69  :       4.54  :       3.28
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 26.228
FLOATING-POINT INDEX: 3.516
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU                 : 
L2 Cache            : 
OS                  : Linux 2.6.282.6.28-bfs1
C compiler          : gcc version 4.2.1
libc                : 
MEMORY INDEX        : 6.316
INTEGER INDEX       : 6.722
FLOATING-POINT INDEX: 1.950
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.
/home/user/nbench #

Pato | # 59 | 2010-11-23, 11:12 | Report

Linux Nokia-N900-51-1 2.6.28.10power46

/home/opt/nbench # ./nbench

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST : Iterations/sec. : Old Index : New Index
: : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT : 348.32 : 8.93 : 2.93
STRING SORT : 61.815 : 27.62 : 4.28
BITFIELD : 1.1808e+08 : 20.26 : 4.23
FP EMULATION : 87.84 : 42.15 : 9.73
FOURIER : 1649.8 : 1.88 : 1.05
ASSIGNMENT : 5.642 : 21.47 : 5.57
IDEA : 946.37 : 14.47 : 4.30
HUFFMAN : 552.05 : 15.31 : 4.89
NEURAL NET : 1.7235 : 2.77 : 1.16
LU DECOMPOSITION : 64.388 : 3.34 : 2.41
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 19.313
FLOATING-POINT INDEX: 2.588
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU :
L2 Cache :
OS : Linux 2.6.28.10power46
C compiler : gcc version 4.2.1
libc :
MEMORY INDEX : 4.653
INTEGER INDEX : 4.948
FLOATING-POINT INDEX: 1.435
Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.

Kamen | # 60 | 2010-11-23, 13:46 | Report

Originally Posted by TiagoTiago View Post

Wouldn't people from cold weather get better scores than people in hotter weather?

Kinda, if your from the Uk lol I think you could run your N900 @ 2Ghz lmao. Joke :| On a serious note so long as your willing to do it, I'm sure you'll be fine. **Do it at your on risk** I've ran mine @ 250 - 1000Mhz since a week after I bought the thing. That was before summer by the way, depending on how good the factory build of you exact phone is you can run it @1Ghz I'd like to say from my experiences and have no problems the 1st few time you max out the phone on a emulator you will smell that "Oh ****" smell of plastic. Leave it to cool down then you should be okay to have it run as a default script at boot up, I changed mine so it's continusly clocked @ 500 -1000Mhz, only because i hate the little initiation of the clock speed being cycled up from 125-250. I like it to be snappy and sharp/responsive.

I have benchmarked mine but it failed at the "NEURAL NET :" Stage. Not sure why, anyone ideas?