Prime95 + Cool'n'Quiet problem

C

CitizenInsomniac

Limp Gawd
Joined
May 26, 2005
Messages
134
After finally reaching a stable OC speed I'm happy with (11x245 = 2.7 GHz) on my Athlon64 3500+ Venice and Asus A8V-E Deluxe mobo, I wanted to try the Cool'n'Quiet feature.
Instead of the stock AMD Cool'n'Quiet software, I decided to use the RMClock software because it allows me to set the min and max VID/FID values. I enabled C'n'Q in the BIOS, started RMClock and set the min values to their defaults (4.0x @ 0.8V), and max to my startup settings (11.0x @ 1.375V - my mobo adds 0.1V extra). It worked like a charm. Speed immediately dropped down to 980MHz (at 1.1V, not sure why, probably a smart feature). When I launched any application that taxed the CPU to 100%, RMClock immediately bumped the speed back to 2695MHz. So far, so good.

When I launched Prime95's small FFT torture test, however, something weird happened. Each test cycle, which usually takes 10-15 minutes to complete, just ended after 2 seconds. No errors, no signs that any bad calculations happened - just the tests zipping by nearly every 2 seconds. I ran it like that for about 1 hr and saw Prime95 go through something like 700 small FFT tests. WTF?
 
Almost every person I talked to said that Cool N Quiet with any o/c can cause a lot of problems.

With no intention of stealing your thread, should you be using Cool N Quiet even thought you decided to OC? Should I?
 
I've heard problems with cool and quiet and overclocking.

If you want cool and quiet, don't overclock.
 
But does anyone know why overclocking and C'n'Q don't mix? As far as I understand, C'n'Q simply allows dynamic FID/VID changes. How does this interfere with overclocking?
 
I'm sure it has somthing to do with changing the multiplyer...as to why having a higher FBS effects it, I don't know. Must be a cross problem between the CPU/Chipset/Driver ;)
 
When you have C'n'Q enabled, it changes the multiplier and voltage of the CPU dynamically, depending on load conditions. What it won't do is change the HTT speed, so, if you have it high, and you need a certain voltage to keep it stable at said HTT speed, anything that drops voltage is a bad thing. That's my theory anyway.
 
Ozymandias said:
When you have C'n'Q enabled, it changes the multiplier and voltage of the CPU dynamically, depending on load conditions. What it won't do is change the HTT speed, so, if you have it high, and you need a certain voltage to keep it stable at said HTT speed, anything that drops voltage is a bad thing. That's my theory anyway.
Click to expand...

If that were the case, then C'n'Q wouldn't work even at stock speeds. Dropping the voltage to 0.8V would hurt a 200MHz FSB as much as it would hurt an overclocked FSB then. Furthermore, it's actually possible to control the minimum voltage in C'n'Q so this seems like a non-issue.

One thing I should probably clarify is that there are 3 separate technologies here: C'n'Q the CPU feature (as enabled in the BIOS), C'n'Q the Windows driver (as available from the AMD website) and RMClock (a 3rd party utility from Rightmark). I think everyone is assuming I'm using the C'n'Q Windows driver here. I'm not. The problem with the AMD driver is that it assumes default voltage and FSB and doesn't let the user modify it. RMClock, on the other hand, allows you to specify min and max values to make sure that your CPU gets the right multiplier and voltage when hitting 100% load.
 
0mega said:
run superPI and see if it fooks your results
Click to expand...

OK, I will give that a try tonight.

BTW, just to clarify again: Prime95 didn't report any errors with C'n'Q+RMClock running. It just skipped from one test to the next very quickly, as if it wasn't even running them.
 
Prime95 has two sets of torture test data. One for machines faster than 1GHz and one for slower machines. I'll bet your C'n'Q setup is tricking prime95 into using the data for slow machines. If so, you'll see "Testing 400 iterations...." instead of "Testing (many thousand) iterations..."

This isn't a severe problem. You can workaround the situation by adding
"CpuOverride=1" and "CpuSpeed=2000" to the local.ini file.
 
prime95 said:
Prime95 has two sets of torture test data. One for machines faster than 1GHz and one for slower machines. I'll bet your C'n'Q setup is tricking prime95 into using the data for slow machines. If so, you'll see "Testing 400 iterations...." instead of "Testing (many thousand) iterations..."

This isn't a severe problem. You can workaround the situation by adding
"CpuOverride=1" and "CpuSpeed=2000" to the local.ini file.
Click to expand...

Ah, finally a knowledgable answer! Thank you, that fixed the problem!
 
You must log in or register to reply here.
Back
Top