Tuning Cool'n'Quiet: Maximize Power And Performance, Part 1

[dunklegend]

I wish they would have included the Athlon X4 620, which being a cheap four cores processor, is in high demand.

But it was a nice article.

 

[edzknows]

-ClkCtrl:#(0-4) - Enable Clock Control (activates clock/core control).
0:No Control
1:UnGanged.
2:Ganged (Load of Highest core).
3:Ganged (Average load of all cores).
4:Ganged (Load of Lowest core).

Whats the default? if none is set?
Does kstat need to be running to set these? or can you just run it once, set it up and not run it later?

 

[mende21]

I think it needs to be running or it stays at the stock settings. I added the options I wanted to the command line and put a shortcut in the startup folder. I use the unganged option.

 

[crushert]

I'm sorry to say this, but something is very wrong with your test setup that has led you to multiple faulty conclusions. The statements you made from those conclusions are inappropriate in the review world. Cool-N-Quiet is not broken and is still working. If it was really broken, you would hear a roar in the online review community, but you don't. No, more than likely, your specific test system is having problems, but don't assume that single sample size means CnQ broken for everyone. Enabling power management does save power. The fact that you made a statement so contradictory to that concept should have been a reason to question your test results and look for another explanation before declaring Cool-n-Quiet as broke. This is an example of go back and "try again".

 

[crushert]

I'm sorry to say this, but something is very wrong with your test setup that has led you to multiple faulty conclusions. The statements you made from those conclusions are inappropriate in the review world. Cool-N-Quiet is not broken and is still working. If it was really broken, you would hear a roar in the online review community, but you don't. No, more than likely, your specific test system is having problems, but don't assume that single sample size means CnQ broken for everyone. Enabling power management does save power. The fact that you made a statement so contradictory to that concept should have been a reason to question your test results and look for another explanation before declaring Cool-n-Quiet as broke. This is an example of go back and "try again".

 

[crushert]

I'm sorry to say this, but something is very wrong with your test setup that has led you to multiple faulty conclusions. The statements you made from those conclusions are inappropriate in the review world. Cool-N-Quiet is not broken and is still working. If it was really broken, you would hear a roar in the online review community, but you don't. No, more than likely, your specific test system is having problems, but don't assume that single sample size means CnQ broken for everyone. Enabling power management does save power. The fact that you made a statement so contradictory to that concept should have been a reason to question your test results and look for another explanation before declaring Cool-n-Quiet as broke. This is an example of go back and "try again".

 

[crushert]

I'm sorry to say this, but something is very wrong with your test setup that has led you to multiple faulty conclusions. The statements you made from those conclusions are inappropriate in the review world. Cool-N-Quiet is not broken and is still working. If it was really broken, you would hear a roar in the online review community, but you don't. No, more than likely, your specific test system is having problems, but don't assume that single sample size means CnQ broken for everyone. Enabling power management does save power. The fact that you made a statement so contradictory to that concept should have been a reason to question your test results and look for another explanation before declaring Cool-n-Quiet as broke. This is an example of go back and "try again".

 

[kokin]

[citation][nom]crushert[/nom]I'm sorry to say this, but something is very wrong with your test setup that has led you to multiple faulty conclusions. The statements you made from those conclusions are inappropriate in the review world. Cool-N-Quiet is not broken and is still working. If it was really broken, you would hear a roar in the online review community, but you don't. No, more than likely, your specific test system is having problems, but don't assume that single sample size means CnQ broken for everyone. Enabling power management does save power. The fact that you made a statement so contradictory to that concept should have been a reason to question your test results and look for another explanation before declaring Cool-n-Quiet as broke. This is an example of go back and "try again".[/citation]
You just quadruple posted, but I think you missed the point with the article. Cool-n-Quiet works all fine and dandy, but it's not optimized so to speak and this article helps people fine-tune and lower their settings.

 

[dimitrik]

Fabulous article! But can you please specify which Phenom II 945 you used? The 95W version or the 125W version? There is a big difference and I'm seeing some fairly different numbers to your on my 945.
Thanks!

 

[voicu83]

nice job BUT: did you guyz know that Gigabyte realeased some new versions of BIOS that shutdown unneeded cores to save power under no load? at least for the GA-MA785G-UD3H I am using. it's called "CPU Core Control". i'dd really like to see how low can a phenom II x4 go on something like that but i don't have any measurement tools "sadface"

 

[fencer55]

Nice to know I can save some energy and reduce my carbon footprint a little!

 

[zodiacfml]

Glad I had a BE processor. You could have included a benchmark and power usage of any BE processor as it is really nice in terms of power management and overclocking.

Just today, I felt the temperature difference from the air coming out the the heatsink between stock setting and lowest multiplier/voltage setting at idle. It really does save power and lowers temperatures.

 

[osse]

Thanks for very useful article. Am building my GHTPC and the plan is using profiles for the diffrent task, for movies, music, youtube, lower voltage and speed to round 2-2,2 ghz stabel, and anonder gamingprofil, not overklokked, but at standard 955/965 frekvens for gaming, hopfully i can tweak a 5850 the same way, low voltage and frekvens for all other things than gaming. Goal is almost silent used as a HTPC, and accept a bit noise when gaming. So this articel gave me a lot of usful information

 

[bunga28]

"Special thanks to AMD for its input."

Thank you editor for using the correct grammar. Yes, I belong to the grammatical police and I am the top cop. Do you have a problem with that?

 

[kokin]

[citation][nom]bunga28[/nom]"Special thanks to AMD for its input."Thank you editor for using the correct grammar. Yes, I belong to the grammatical police and I am the top cop. Do you have a problem with that?[/citation]
Wouldn't it be dumb to say "Special thanks to AMD for it is input."? I thought "it's" always stood for "it is".

learning English
is fun!

I do have a problem with the grammatical police's top cop, cause he can't seem to do his job correctly. Thank you for being cocky and a hypocrite since both always seem to pair well together.

 

[clinamen_1]

Well, its is plural for it and does not stand for it is, for which you are correct.

On another note, if kstat operates like Overdrive, the settings are not firm in BIOS. I tested Overdrive on my board and found that when I went to Win7, BIOS ran in default, not with the settings applied in XP. So I ran it again and for some reason, when I went back to XP, I had problems. It appears overdrive expects to see default in BIOS from which it makes the adjustments. Do all your stuff in BIOS directly if you dual boot.

I don't like these utilities for AMD. Intel runs its utility more intelligently.

 

[zeterian]

Hi, good article. I assume the 945 processor is the 125 watt version. It would be interesting to see if the new 95 watt 945 reduced power consumption as much as advertised, as well as any effects on over/underclocking.

 

[diddo]

Nice article! Thanks for the information.

 

[xsever]

The print-link links to the wrong article.

Please fix it.

Thank you.

 

[NoCaDrummer]

you always have the option of buying one of AMD's energy-efficient low voltage models and paying upfront for power savings.

Yeah, that's if you can find one. Newegg and others don't seem to have the 45w Athlon II CPUs that were announced last year (and I check EVERY day). I have a 45w 2.5GHz Athlon x2 (which I really like!), but was hoping for something a little faster or with more cores.

 

[JohnnyLucky]

I'm slapping myself upside the head. This is an article about AMD's Cool'N'Quiet technology. I've got an Intel system! What was I thinking?

 

[beagle87]

@Razor

No. Switching between power states take time about 1 microsecon (about one million cycles). In fact, if you choose to have less power states, you can gain faster p-state switching for price of less power efficiency. By doing voltage tweaking, you don't loose performance, but you get cooler and quieter (tm? ) CPU.

 

[rvbeppler]

My Phenon II 940 Black Edition now is working very well at reduced voltages and higher clocks (Gigabyte GA-MA770-UD3 AM2+ Motherboard with 2x2GB OCZ DDR2 RAM):

Original:
3000MHz / 1.350V CPU / 1.175V NB
2300MHz / 1.250V CPU / 1.175V NB
1800MHz / 1.150V CPU / 1.175V NB
0800MHz / 1.000V CPU / 1.175V NB

Optimized:
3300MHz / 1.250V CPU / 1.050V NB (using FID 17, though this is an unlocked processor, no FSB change at all)
2400MHz / 1.100V CPU / 1.025V NB
1600MHz / 0.900V CPU / 1.000V NB
0800MHz / 0.700V CPU / 0.950V NB
(Each p-state was tested succesfully with Prime95 and IntelBurnTest)

Lower temps, lower noise, lower electricity bill and faster computer! It's wonderful!

Someone could make a software with an integrated burn-in stability test to find optimum values for each parameter (for example, kind of K10Stat + Prime95 + automatic values adjustment after found each crashing point...)

 

[Guest]

Well they still havent fixed it. 790gx a2+ and phenom II 940 runs the HT and NB at 1600mhz not 1800mhz. The new bios with ageas fix (sets the HT speed etc) still doesnt fix it. This is two biostar boards in row for me that dont properly run phenoms. First is a phenom I 9850 now a phenom II 940. I shouldnt have to overclock to stock on a "supported" cpu and loose cool and quiet. Oh and ya the new bios disables CnQ when setting custom P-states. The old one did not. Thanks biostar.