Why is my Opteron 165 faster at SuperPI when SP2004 is also running?

shaolin95 · 05-20-2006, 02:55 PM

Hi!
I tested this many times and its weird. If I run SuperPI and assign one core to do it my best score was 28.282 after many tries. The funny thing is that if I run superpi core 0 (for example) and at the same time SP2004 to core 1, I will get better super pi times consistently! This is very weird. My best has been 28.141 and I am able to hit that mark consistently for as long as sp2004 is running.
Any thoughts? Perhaps sort of the same as assigning high priority?

VulgarHandle · 05-20-2006, 03:11 PM

Quote:

Originally Posted by shaolin95

Hi!
I tested this many times and its weird. If I run SuperPI and assign one core to do it my best score was 28.282 after many tries. The funny thing is that if I run superpi core 0 (for example) and at the same time SP2004 to core 1, I will get better super pi times consistently! This is very weird. My best has been 28.141 and I am able to hit that mark consistently for as long as sp2004 is running.
Any thoughts? Perhaps sort of the same as assigning high priority?

lol, might wanna add this as a tweak in the extreme oc/pi section

r3w4 · 05-20-2006, 04:28 PM

Did you try assigning SuperPi 32M to one core and 1M on the other?

shaolin95 · 05-20-2006, 05:11 PM

Havent tried that yet. I know that running dual superpi doesnt make them faster though. I did try using realtime priority and it did help to hit the 28.141 so it seems to be creating the same effect that running sp2004 does. Kind of weird that it scores better when the other core is fully busy.
Can someone else try it just to validate this strange behavior.

cky2k6 · 05-20-2006, 05:47 PM

just tried that on my 170, and it scored aobut 2s slower with sp2004... could be that i dont have enough ram, as im currently playing with only 1 512 stick. nevermind, it was on blend, i set it to small ffts, it brought the superpi to the same level as with no sp2004, but not any lower.

shaolin95 · 05-20-2006, 06:15 PM

Maybe you are ram limited who knows.

krylon · 05-20-2006, 08:40 PM

That is really wild

±Pink=FloYd± · 05-21-2006, 04:12 PM

bench it with 3dmark or something and compare cpu scores with and without sp2004

Kaiser_Sose · 05-21-2006, 04:32 PM

How can you take the noise from that jet engine on you cpu

cky2k6 · 05-21-2006, 05:00 PM

well, what do you know. i was ram limited. i put my 2gb kit back in, and running sp2004 does have the same effect as running real time. i ran superpi a few times in a row to establish a margin of error, it was ranging between 34.76 to 34.67. when i ran sp2004 on core 1, the score went down to 34.203, definetely out of the margin of error.

shaolin95 · 05-21-2006, 06:27 PM

Quote:

How can you take the noise from that jet engine on you cpu

Honestly, you get used to it...besides I am a headphone gamer anyway and the benefits of not worrying about water leaks or dead pumps let me sleep better at night :-)

cky2k6, thanks for testing it. It only seems to work with some programs not all. I noticed that cause I was trying for a quick stability test and decided to try sp2004 and superpi 32mb. Previously my best time was 26mins + but that time I did 25mins so I was totally surprised.
Regards

biohead · 05-21-2006, 10:08 PM

when running only SuperPI, did you actually assign it to core 0 or didn't you touch the affinity? If you didn't, windows would try to balance the load on both the cores, thus making it slower in these sort ot tests.

shaolin95 · 05-21-2006, 10:38 PM

I use affinity all the time, thats why its weird at least imho.