cool n quiet on phenom II

[Veedo]

How high is your NB Freq? I noticed the same problem on my M4A79 Deluxe with CnQ enabled when the NB Frequency is at or over 2550mhz or so.

--Matt

this was the problem. i lowered my ht multiplier from 10 to 8, set freq to 1600, and changed my multi to 14x, while still using my 252fsb setting. this is giving me a little lower nb speed of 2268, while keeping my ram around 500mhz, which is what i wanted .

i then setup k10stat so that my max powerstate is 3400mhz, and lowest is 1700ish, with .98 volts. still going to play around with lowest settings, but it seems that power draw changes very little going any lower than 2000mhz and 1volt.

i created a shortcut to k10stat with the options
-loadpstate -nowindow and added it to the startup folder. i had to disable uac to keep from getting the blocked startup entry notification in vista.

so now my system boots at around 3500mhz, but when windows loads, cnq takes over and runs me at 3400 when at load.
difference in power draw at idle states is about 120 watts less with cnq enabled. keeps my systems power supply fan spinning slower, and is much quieter now! 120 watts is a pretty big difference, even more so for those using a higher vcore or clock speed. you can get the best of both worlds by doing this

[mattkosem]

this was the problem. i lowered my ht multiplier from 10 to 8, set freq to 1600, and changed my multi to 14x, while still using my 252fsb setting. this is giving me a little lower nb speed of 2268, while keeping my ram around 500mhz, which is what i wanted .

i then setup k10stat so that my max powerstate is 3400mhz, and lowest is 1700ish, with .98 volts. still going to play around with lowest settings, but it seems that power draw changes very little going any lower than 2000mhz and 1volt.

i created a shortcut to k10stat with the options
-loadpstate -nowindow and added it to the startup folder. i had to disable uac to keep from getting the blocked startup entry notification in vista.

so now my system boots at around 3500mhz, but when windows loads, cnq takes over and runs me at 3400 when at load.
difference in power draw at idle states is about 120 watts less with cnq enabled. keeps my systems power supply fan spinning slower, and is much quieter now! 120 watts is a pretty big difference, even more so for those using a higher vcore or clock speed. you can get the best of both worlds by doing this

I figured it would be that. It sucks though huh? That extra NB clock makes a BIG difference.

--Matt

[Veedo]

I figured it would be that. It sucks though huh? That extra NB clock makes a BIG difference.

--Matt

really? do you have some benchmarks? this is my first phenom, so im still learning. i may run some benches and compare sometime, but i wont have time this weekend. thanks for the info

[kink]

Yes, CNQ can normally (?) be used with FSB overclock, but...

The Phenom IIs have 4 power (performance) states. Each pState has an associated CPU multiplier and voltage. When you bump the FSB, you risk causing an instability at any one of the 4 pStates. Sounds like you're stable at pState=0 (max performance), but probably not at pState=3 (max power save).

At a FSB 252 and pState=3, the CPU frequency is 1008MHz (up from 800MHz), but, the default voltage specified in the pState table still applies, which could be a problem.

Unfortunately, I don't know of a good way to permanently alter the voltages in the pState table.

I've just developed a small tool addressing exactly issues like these, here's the thread: http://www.xtremesystems.org/forums/...d.php?t=220185

[mattkosem]

really? do you have some benchmarks? this is my first phenom, so im still learning. i may run some benches and compare sometime, but i wont have time this weekend. thanks for the info

Sure. It's makes a very substantial difference. Here are some quick comparisons.

--Matt

[Sintel]

I had the same problem, enabling CNQ slowed the system to a crawl. If you're interested, my solution is posted here:

http://home.comcast.net/~pmc650/site...4cfc46a7a02692

Anyone tried this tool or any other 3rd party tool? Is AMD aware of the problem? The review sites completely missed it.

Just noticed this yesterday myself: quick superpi 1M bench on Windows XP scores all over the place (between 24 and 29 seconds) but if I set affinity to 1 cpu, 21s no problem.