Benchmark results - pandora first edition 256MB vs 1GHz

[Exophase]

You can't call the increase from one SoC to another one linear >_> What I mean is that both SoCs don't (usually) perform linearly with clock speed, but since DM3730 tends to do a little better at the same clock speed it can compensate.

 

[Linux-SWAT]

Let's start a flamewar, would you ?


;^)


Yeah, indeed, linear wasn't the word i intended.

 

[Grench]

lmbench took too long, so will run overnight... So here are some memory benchmarks from cachebench.

Very cool - now I keep looking at the date on my clock to see how much longer I have to wait for my 1Ghz Pandora - even though that wait is indeterminate other than, "soon".

 

[foft]

Adding both at 800MHz. The new one will not clock any lower than that. At least not with the gui widget.


(I also have lmbench results - will post in 20 mins or so!)


CC:


http://ssh.scrameta..../pandoracc2.jpg

GHz:


http://ssh.scrameta....pandoraghz2.jpg

CC800:


http://ssh.scrameta..../pandoracc8002.jpg

GHz800:


http://ssh.scrameta....pandoraghz8002.jpg

 

[foft]

OK, here is lmbench. Will see about getting better formatted ones.


I left my server on too for some perspective of what what pandora 2 has to beat(Intel® Core™2 Quad CPU Q6600 @ 2.40GHz)


Edit:Removed, look like nonsense:-(

 

[milkshake]

I don't understand any of this?

 

[foft]

I don't understand any of this?

cc600pand - old pandora at 600MHz


cc800pand - old pandora at 800MHz


ghz800pan - new 1GHz pandora at 800MHz


ghzpandor - new 1GHz pandora at 1000MHz

 

[foft]

how far/fast/high can the new 1ghz one go in terms of ghz?

Mine goes to 1150MHz. It is unstable at 1200MHz under load.

 

[Grench]

Basic float operations - times in nanoseconds - smaller is better

-----------------------------------------------------------------

Host                             OS  float  float  float  float

                                                 add    mul     div     bogo

--------- ------------- ------ ------ ------ ------

cc600pand  Linux 3.2.17   36.2   25.2  138.6  292.3

cc800pand  Linux 3.2.17   27.1   18.9  104.2  219.7

ghz800pan  Linux 3.2.17   36.2   25.2  138.9  291.1

ghzpandor  Linux 3.2.17   21.7   15.1   83.1  173.8

pvr.scram  Linux 2.6.38 1.2500 2.0800   15.8   15.4

Several of these 'don't feel right' or I'm misinterpreting them.


Does this mean that, for this operation, a 1Ghz Pandora running at 800Mhz is the equivalent of a CC Pandora running at 600Mhz?


Is this an effect of over/under clocking the memory?

 

[foft]

Doh. I think I swapped sd cards between tests and haven't fixed that. Will fix!


edit - no this still seems odd... Removed them. Oh well, time to play some games!