Sniper wrote:
> I ran SuperPI w/ 1,000,000 decimal point(s) while RightMark sat to the side.
> CPU Core Speed never dropped below 3199.97 & went as high as 3200.05. In
> addition, it ran the SuperPI calculations in 31.328 secs.
>
> To me it appears that the proc is running as it should ... think if I
> completely turned off CoolNQuiet it would do better?
>
> Sniper
>
I found some results on hwbot. 31.3 is still a bit slow.
http://www.hwbot.org/listResults.do?...true&limit=100
26. SuperPi - 26.44 sec - Johny (warforum.cz) - (Athlon 64 6400+ X2 @ 3209mhz)
27. SuperPi - 27.16 sec - LifeTimeKid (LifeTimeClan) - (2x Athlon 64 6400+ X2 @ 3215mhz on air)
28. SuperPi - 27.3 sec - THRASHER2 (PC Apex) - (Athlon 64 6400+ X2 @ 3207mhz on air)
29. SuperPi - 27.73 sec - f3rr1s (PC Games Hardware) - (Athlon 64 6400+ X2 @ 3216mhz on air)
30. SuperPi - 28.12 sec - jonny13 (Freeocen.de) - (Athlon 64 6400+ X2 @ 3200mhz on air)
31. SuperPi - 28.2 sec - iman_22a (ShahrSakhtafzar Iran ) - (Athlon 64 6400+ X2 @ 3216mhz)
The guy with 26.44 is using DDDR2-800 5-5-5-18 2T dual channel (as
seen in his CPUZ). And so is the guy with the 28.2 second timing.
Weird. Maybe one of them has antivirus software running in
the background.
Is the memory properly installed for dual channel ?
(Usually, the slots to use are a matched color, for your matched sticks
of memory.)
Something else you can try as an experiment, is to set affinity
for just one core, for SuperPI before it runs.
1) Start SuperPI, so you see its initial window.
2) Press control-alt-delete, to start Task Manager
3) Find the SuperPI program in the list of running processes.
4) Right click, and select "Set affinity" using the item at the
bottom of the list.
5) Your setup should show "CPU0" and "CPU1" with a tick in
each of the boxes. Remove the tick from next to CPU0.
What that means, is for the rest of this test run,
SuperPI will run on CPU1.
6) Start the 1 million digit calculation and get a time.
If the time is faster, that says the movement of SuperPI
from CPU0 to CPU1 and back again (OS scheduler does that),
caused some performance loss. And sometimes it is an
interaction with Cool N' Quiet.
I did a number of runs with affinity not set, and with
affinity set so the process stayed on CPU1, and on
average it made no difference.
The power management on my motherboard is busted (BIOS
design decision) so my machine runs at top speed all
the time. There may be a setting in the BIOS to disable
CNQ, or you can use the Power control panel in WinXP
and change the power scheme to one that would prevent
CNQ from cranking it down.
Using RMClock, maybe you've already done enough experiments
to see how Cool N' Quiet works ? (I.e. RMClock graph lines
change when system is idle.)
Oh, another thing you can check. Bring up Device Manager,
and click the (+) next to "Computer". The value of the
text string underneath should be "ACPI Multiprocessor PC".
If it is something different, post back. If it says
"Uniprocessor", you could be running on one core.
Paul