Active Topics

 



Notices


Reply
Thread Tools
Posts: 255 | Thanked: 109 times | Joined on Dec 2009
#21
letho 250-800 (1000) ULV @ 800

TEST : Iterations/sec. : Old Index : New Index
: : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT : 329.68 : 8.45 : 2.78
STRING SORT : 58.017 : 25.92 : 4.01
BITFIELD : 1.1159e+08 : 19.14 : 4.00
FP EMULATION : 82.32 : 39.50 : 9.11
FOURIER : 1544.2 : 1.76 : 0.99
ASSIGNMENT : 5.3255 : 20.26 : 5.26
IDEA : 887.59 : 13.58 : 4.03
HUFFMAN : 517.12 : 14.34 : 4.58
NEURAL NET : 1.6193 : 2.60 : 1.09
LU DECOMPOSITION : 60.551 : 3.14 : 2.27
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX : 18.170
FLOATING-POINT INDEX: 2.429
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 : 4.385
INTEGER INDEX : 4.649
FLOATING-POINT INDEX: 1.347
Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38

Last edited by baron von bubba; 2010-04-11 at 11:44. Reason: 800 not 1GHz
 
Posts: 1,427 | Thanked: 2,077 times | Joined on Aug 2009 @ Sydney
#22
Originally Posted by jaeezzy View Post
Actually., so how do one figure out whose is running as per the MHz? and is it doing any performance boost as per this benchmark?
Once we have some more numbers at similar speeds, we will be able to compare.

But you can clearly see that my 1000 and 1100 have the highest numbers by far. (=fastest)


Originally Posted by baron von bubba View Post
letho 250-800 (1000) ULV @1000
Are you sure you were running at 1000mhz? (that kernel by default is 800mhz) You need to make changes to run at 1000mhz.
If you check my 1000mhz scores, yours is significantly lower than mine. (and yours match up to 800mhz result above by jaeezzy)
 
Posts: 255 | Thanked: 109 times | Joined on Dec 2009
#23
yeah was just redoing and making sure had nothing else running etc.

keep getting "CPU:NNET--error in opening file!" now when i try and run the test. :0/
 
Posts: 1,427 | Thanked: 2,077 times | Joined on Aug 2009 @ Sydney
#24
Hmm. No idea what that error means sorry.

btw, I have just updated my post with 800mhz results also.
http://talk.maemo.org/showpost.php?p...9&postcount=17

Last edited by jakiman; 2010-04-11 at 11:46.
 
Posts: 255 | Thanked: 109 times | Joined on Dec 2009
#25
Originally Posted by jakiman View Post
Hmm. No idea what that error means sorry.

btw, I have just updated my post with 800mhz results also.
http://talk.maemo.org/showpost.php?p...9&postcount=17
my results seem pretty comparable with yours, the 1gig results i get before the error are a tiny bit higher.
 
Posts: 292 | Thanked: 131 times | Joined on Dec 2009
#26
Hummm. By reading those scores one can reach a few conclusions so far on tendencies:

1) The N900 processor can actually do more general instructions/cycle than earlier Intel Pentium and AMD K6.
2) Floating point is way behind. Maybe FP instructions are being emulated. I wonder if there is any way to recompile the benchmark with NEON support (also, will it help)?
3) Memory and integer performance seems to increase linearly with OC.

Now I wonder. How long would the N900 sustain a stress test? Let's say: running superPi for 48 hours? ;-)

Last edited by soeiro; 2010-04-11 at 12:49. Reason: typos
 
Posts: 946 | Thanked: 1,650 times | Joined on Oct 2009 @ Germany
#27
Originally Posted by soeiro View Post
Now I wonder. How long would the N900 sustain a stress test? Let's say: running superPi for 48 hours? ;-)
Feel free to fry...err..try yourself
 
Posts: 1,224 | Thanked: 1,763 times | Joined on Jul 2007
#28
Originally Posted by soeiro View Post
2) Floating point is way behind. Maybe FP instructions are being emulated. I wonder if there is any way to recompile the benchmark with NEON support (also, will it help)?
It is compiled with NEON. With emulated FPU it is much slower. Maybe VFP will give better results.
__________________
My repository

"N900 community support for the MeeGo-Harmattan" Is the new "Mer is Fremantle for N810".

No more Nokia devices for me.
 

The Following 2 Users Say Thank You to Matan For This Useful Post:
Posts: 1,751 | Thanked: 844 times | Joined on Feb 2010 @ Sweden
#29
hmm.. now, i am also getting the cpu:nnet--error

Will try to swap it and see if it helps.

Last edited by AlMehdi; 2010-04-11 at 16:03.
 
Posts: 1,224 | Thanked: 1,763 times | Joined on Jul 2007
#30
Originally Posted by Matan View Post
It is compiled with NEON. With emulated FPU it is much slower. Maybe VFP will give better results.

-mfpu=vfp gives the same results as -mfpu=neon. File is there, if anyone wants to try:

http://my.arava.co.il/~matan/770/n900/nbench.vfp
__________________
My repository

"N900 community support for the MeeGo-Harmattan" Is the new "Mer is Fremantle for N810".

No more Nokia devices for me.
 

The Following 2 Users Say Thank You to Matan For This Useful Post:
Reply

Tags
benchmark, burn in, overclock, stability, stress


 
Forum Jump


All times are GMT. The time now is 05:05.