I been running my AMD 64 for a while and just recently learned how to enable Cool N Quiet.

I was wondering if it does indeed save power or not? Isn't the power supply pulling the same amount of power from the wall regardless of how much power the CPU draws? The Cooler Master PSU I bought a while back claims to run at a true 450 watts.

System Specs:
Coolermaster true power 450 watts
AMD Socket 939 3000+
DFI Ultra D Mobo

No, when your PSU says 450w on the label it does not mean it always draws 450w constantly, nor does it mean that that is the max it will draw from the wall. The 450w means that is the max the internal components can draw from it, if you're drawing max power from it you'll find it sucks more than 450w from the wall.

So yes, CnQ will save alot of power.

Think about it in this (simplistic) way:
Your house has a power circuit which runs to your wall sockets. If you connect a two bar electric heater to the socket and switch both bars on you will draw say 2kW, switch one bar off it'll drop to 1kW and the disc in your meter will spin slower.
Switch off the heater completely and the meter will stop because no power is being drawn from the circuit.

It's the same with CnQ (to an extent). At idle it'll drop the CPU frequency back to 'one bar' and when you need some oomph, it'll ramp back it back up to 'two bars'.

Sadly, CnQ doesn't actually get "back to two bars". If you've got a PC that's faster than you need anyway, this doesn't matter - otherwise maybe check out rmclock
http://cpu.rightmark.org/products/rmclock.shtml
Bob

What? Need quote on that.

Really, where does that come from?

I have an x2 4800+ Toledo (E6 revision) here under CnQ which (when viewed with CPU-Z 142.3 ) sits at 1004.6MHz with 5x multiplier at idle. Loaded up it rises to 2411.2MHz with 12x multiplier as it should.

If anything I've gained 11.2 MHz over it's rated spec....

AFAIK the issue is mostly that QnQ is very "reluctant" to return to 100% levels. Of course, if you're using something like SiSoft Sandra CPU Burn then that'll likely push it there. If you're doing something more cycle-use-bursty it likely won't.
If you'd like to see what happens on your own machine, there are diagnostics tools at rightmark (and their version of CnQ seems to have grown out of both the diagnostic tool and a dissatisfaction with what CnQ was doing).
It's free software (well, there's a cheap-ish upgrade to a "pro" version that adds so few features that it's more of a way to make a donation).
What are you using for a diagnostic?

All that said, this is not something for the technically inexperienced. It typically requires finding/reading a CPU mfgr's spec for the particular chip you have and then deciding what multiplier/voltage combination you want to use for each level of CPU use (the AMD docs, at least, have a range for each at each level, and my experience is that the typical chip will be a bit flaky at the lowest voltage levels for each multiplier, although you might "get lucky" - the same issue as with overclocking). My experience is also that it's most difficult to test at the "mid" levels - things like Sandra push usage to the max. Some chips that do fine with an even-lower-than-spec voltage at their lowest multiplier need mid-spec-range voltages at mid and/or top multipliers. Basically a PITA to get "cranked down" as low as you can go.
Also, this does not work for overclocking - except possibly for the AMD "black box" (unlocked multiplier) CPUs, which I haven't tried.
And don't expect any "magic" - this won't let you run an 89W processor at 25W

Anyway, if you're looking for a "quote" just go check out rightmark, try the free stuff (which is likely all you need). I'm just an end-user, have put this stuff on a couple of AMD based machines, it seems to work - and it seems to be an appropriate solution for machines that both run 24x7 and run a server OS that doesn't deal gracefully with being suspended (e.g. win 2k server).
There's lots more quotes on the forums there, from folks who have used this product in a lot more situations than I have. I haven't touched my settings (or upgraded the product) on any of the PCs I have this working on for at least six months, so go there for more current info - rightmark CPU forum

Oh - as I vaguely remember, that "find processor specs" step was supposed to be an automated process at install. I had"oddball" processors (basically whatever the fastest 25W processor was at the time of purchase), so maybe I'm overstating the difficulty.

Sorry Bob but I think you're off on a tangent some how.

We are talking about AMD's Cool n Quiet which works just fine.

http://www.amd.com/us-en/Processors/Pro ... 72,00.html

Erm - well you've spotted that I sometimes can't type - meant to type CnQ not QnQ.
We're talking about the AMD driver that adjusts multiplier and voltage, yes?

Go look at the rightmark stuff. They explain things better than I do - and you can then decide if it really has something to do with CnQ or not.

Bob

Read the Rightmark stuff, then tested my A64 x2 3800+ EE windsor core. The AMD driver worked well. Must be an issue only with old A64 drivers...

The RM tools just seem to provide the same information as CPU-Z and provide a manual control function version of AMD's own C'n'Q. Is that about right?

As I've said before, at idle my processor gets ramped down and under load it ramps back up to full speed as it should.
I really don't think I have a problem that needs this cure?

I think it's a function of what you're doing with the particular PC, as you can decide when things ramp up and down.

I think the issue that bobkoure is referring to is that the CPU must detect that it is being loaded and then switch voltage and multiplier and this has a time lag before the CPU reaches full speed.

For applications that constantly load the CPU, the speed won't be affacted but ones that load the CPU in bursts can result in slow down due to the mode switching delay. I experienced this when using a Sempron based PC to watch digital television - with high def channels, there were continual picture breakups with CnQ enabled and these went away when CnQ was turned off.