Phenom II and C'n'Q 3.0 - Reviewers were wrong :p:

[Lightman]

As mentioned in many review articles regarding Phenom II Cool'n'Quiet 3.0 solved problem with rather big performance hit when utilizing energy saving option. New solution was to simply clock all cores at the same frequency forced by highest clocked core p-state.
Initially I was quite happy with this solution as it worked very well for me in Vista x64. When I switched to XP x64 I've noticed that C'n'Q is working like was the case for Phenom I! Each core is clocked depending on utilization, separate from the others. I've reinstalled AMD CPU Driver for XP/2003 Server but same thing happens.
This lead me to belief that what was claimed to be C'n'Q 3.0 feature really is only Vista patch. Question is: can this patch work same way on Phenom I?
Can we get same patch for older OS like XP?
Can AMD include option in BIOS to select either old or new model of multiplier switching independent from OS?

What do you make of this guys and girls?

[Mechromancer]

I think the way Windows handles threads is the core of the problem (no pun intended). Microsoft needs to patch windows to allow programs and processes to pick a core and stick with it until it the next time it is run. Lock the affinity MS!!!!

[Nedjo]

Vista has different way of handling the p-states... if I'm correct Nehalem's Turbo feature only works under Vista 'cos of that...

[Rammsteiner]

Ugh, Vista... vista...

I want to get Vista working but Im going to hold it off untill at least CCC 9.1 is out At this point, I think MS should be sued for war criminality, ethnic purification against my drivers

[soundood]

ok i have never used it as it messed up my clocks and crashed and all that, thats behind us now.

ok, so i havent been paying attention because of this, but have just done a crash course on C&Q v3.0.

this all sounds great, improvements made to utilize the 45 nm production blah blah blah,

but is it any better? can we OC to 3.5 gig, and still use it? maybe C&Q dropping it to 1.2 gig instaed of the rated 800MHz?

if so i am in, my new crosshair II is on the way, and i am thinking this hybrid power might be worth trying, with C&Q as well this could be a cool solution to energy saving (nothing new i know, but it is for me lol)

but my spidey senses are tingling, i know it cant be that easy, it would be....you know....tooooo easy.

anyone got C&Q 3.0 running with an overclock?

and thanks lightman, i knew nothing of a new C&Q

[Lightman]

ok i have never used it as it messed up my clocks and crashed and all that, thats behind us now.

ok, so i havent been paying attention because of this, but have just done a crash course on C&Q v3.0.

this all sounds great, improvements made to utilize the 45 nm production blah blah blah,

but is it any better? can we OC to 3.5 gig, and still use it? maybe C&Q dropping it to 1.2 gig instaed of the rated 800MHz?

if so i am in, my new crosshair II is on the way, and i am thinking this hybrid power might be worth trying, with C&Q as well this could be a cool solution to energy saving (nothing new i know, but it is for me lol)

but my spidey senses are tingling, i know it cant be that easy, it would be....you know....tooooo easy.

anyone got C&Q 3.0 running with an overclock?

and thanks lightman, i knew nothing of a new C&Q

Yes, I was able to run C'n'Q up to 3.3GHz/880MHz max/min p-state just by increasing HTT to 220MHz. I will try to push a bit more, but at the moment my NB is set to 11x and going above 2.4GHz on it with stock volts hangs my comp .
Problem is: only way so far on most mobos is to clock via HTT and let C'n'Q manage multis and voltages in a range 4x/1.0V to max. multi/stock volts. As soon as you reaches max. stable frequency on any of the settings that's your C'n'Q OC limit.
I know that my chip will do 1.4GHz/1.0V easily so that gives me theoretical 350MHz HTT limit on the idle setting, but 3.6GHz is the max. on the load with stock volts which in my case will require 240MHz HTT. So I'm looking at 960MHz low p-state and 3.6GHz full throttle

I will try my theory soon. I need to lower NB multi to 10x in BIOS and set memory to DDR 800.

Give me

[soundood]

I know that my chip will do 1.4GHz/1.0V easily so that gives me theoretical 350MHz HTT limit on the idle setting, but 3.6GHz is the max. on the load with stock volts which in my case will require 240MHz HTT. So I'm looking at 960MHz low p-state and 3.6GHz full throttle

I will try my theory soon. I need to lower NB multi to 10x in BIOS and set memory to DDR 800.

this sounds pretty good, around a gig in low power mode and 3.6 gig in full throttle, to be honest i was thinking 3.6 gig for a 24/7 OC, if you know what i mean.

now i just need to hear that hybrid power mode works with a GTX260 and a crosshair II and i am very happy

[Lightman]

this sounds pretty good, around a gig in low power mode and 3.6 gig in full throttle, to be honest i was thinking 3.6 gig for a 24/7 OC, if you know what i mean.

now i just need to hear that hybrid power mode works with a GTX260 and a crosshair II and i am very happy

It works fine! I have just set HTT to 235MHz from BIOS and everything is OK.
For my 24/7 setting I will only need to tweak memory for lower latencies and I'm sorted.
BTW if your mobo is increasing vCore by overvolting p-state, then you are only limited by the max. HTT frequency you can reach. For example my GBT is adding voltage on top of what CPU sets, so if I add 0.1V in BIOS AOD still reads 1.35V but real is 1.46V, same goes for lowest p-state which instead of 1.0V will be 1.1V!
You are loosing a bit of energy efficiency but who cares we are EXTREME

[soundood]

this is very good news lightman, i allways liekd the theory of C&Q and used to emulate it with Crystal CPU, i am really looking forward to my new 'build now'

Crosshair II, 940BE, GTX260, with water cooling, all running a C&Q hybrid power config, with Windows 7 maybe? if not vista

thanks for testing this out lightman, gives one a clear idea what to achieve

[64NOMIS]

With Phenom II we doubled the number of p-states, with Phenom we only had Div 2 but like Phenom II frequency is modulated for each core individually. In addition 45nm and much better p-state regulation for the memory controller are what have dramatically llowered the idle and load power consumption.

Increasing the FSB should result in scaling of p-states, in other words, CNQ will work fine. However, if you touch the multiplier with a BE part all bets are off. The exception is if you use AMD OD which should allow you to employ CNQ p-states even for a multiplier modified system.

I have not tested all the permutations, and BIOS implementation is a major issue in determining outcomes.

[swaaye]

The exception is if you use AMD OD which should allow you to employ CNQ p-states even for a multiplier modified system.

I have not tested all the permutations, and BIOS implementation is a major issue in determining outcomes.

I tried AMD OD and it didn't recognize the PII 940 as an AMD CPU. It wouldn't load up at all, just gave a CPU unrecognized error.

[informal]

@swaaye
It could be BIOS issue.The board's BIOS must provide "hooks" for AOD to function properly.

[swaaye]

@swaaye
It could be BIOS issue.The board's BIOS must provide "hooks" for AOD to function properly.

Ah. I suppose that's probably the case then. I'm using a Gigabyte MA78GM-S2H.

[Lightman]

With Phenom II we doubled the number of p-states, with Phenom we only had Div 2 but like Phenom II frequency is modulated for each core individually. In addition 45nm and much better p-state regulation for the memory controller are what have dramatically llowered the idle and load power consumption.

Increasing the FSB should result in scaling of p-states, in other words, CNQ will work fine. However, if you touch the multiplier with a BE part all bets are off. The exception is if you use AMD OD which should allow you to employ CNQ p-states even for a multiplier modified system.

I have not tested all the permutations, and BIOS implementation is a major issue in determining outcomes.

Thanks for inside info Simon!
Now I've tried multis in AOD but they don't work. As soon as I change Power Profile to one using C'n'Q everything goes to default, including voltages.
That is on my GA-MA790FX-DQ6 with final F6 BIOS (not yet on GBT page).
I know some motherboards allow you to set C'n'Q lowest and highest multi from BIOS (if I remember properly it was MSI K9N or DFI for S939).

Oh, and do you know why Vista and XP are using p-states differently? (as explained in first post)

[BertM]

why dont u use k10stat lightman?
i changed cnq to 650mhz 0.75v for low and 3700mhz 1.425v for high.
even cnq down to 100mhz if i want.
but i asume u clock this way to oc the northbridge?

i have noticed that they still didnt add a positive voltage adjust for nb in the f6 bios. lame!!

[Lightman]

why dont u use k10stat lightman?
i changed cnq to 650mhz 0.75v for low and 3700mhz 1.425v for high.
even cnq down to 100mhz if i want.
but i asume u clock this way to oc the northbridge?

i have noticed that they still didnt add a positive voltage adjust for nb in the f6 bios. lame!!

Yes, for NB

Also I've got reply form GBT regarding that issue. It's posted in motherboard thread

[swaaye]

k10stat appears to be very nice

[Lightman]

k10stat appears to be very nice

It really works nice! Thanks for tip

[Titan7171]

Hey just wanted to post my results on C N Q and ask a question.

first off this thing really works good Im doing 254 bus speed idle is 1016mhz with the 4x multi at 0.992v (nice for surfing the net) then jumps up to 3.8ghz with the 15x multi at 1.55v when I start a program or game.

The only thing is the HT link is set at 1016mhz and the nb is at 2286mhz I tried setting the ht and nb multi but CNQ disabled when I did that.

[Mara]

Hey just wanted to post my results on C N Q and ask a question.

first off this thing really works good Im doing 254 bus speed idle is 1016mhz with the 4x multi at 0.992v (nice for surfing the net) then jumps up to 3.8ghz with the 15x multi at 1.55v when I start a program or game.

The only thing is the HT link is set at 1016mhz and the nb is at 2286mhz I tried setting the ht and nb multi but CNQ disabled when I did that.

My past experience show that for Cool'n'quiet to work, I can not change voltages or multipliers... only HTT speed can be changed. (Since C'n'Q does not adjust the HTT.) If I manually set either voltage or multipliers, C'n'Q doesn't work anymore. (That's on Gigabyte board in my sig.)

I did compare the power consumption from the wall on my PC by having my 940 at 3.5GHz (stock Vcore), and manually downclocking it to 800MHz (0.8V Vcore). The difference in Watts was only ~20W, thus I found it not much benefit to bother with.

[Titan7171]

My past experience show that for Cool'n'quiet to work, I can not change voltages or multipliers... only HTT speed can be changed. (Since C'n'Q does not adjust the HTT.) If I manually set either voltage or multipliers, C'n'Q doesn't work anymore. (That's on Gigabyte board in my sig.)

I did compare the power consumption from the wall on my PC by having my 940 at 3.5GHz (stock Vcore), and manually downclocking it to 800MHz (0.8V Vcore). The difference in Watts was only ~20W, thus I found it not much benefit to bother with.

I see, in my case though Im running 1.55v for 3.8ghz 24/7 so dropping from 1.55v 24/7 to 0.992v is a bigger difference.

lets say a computer is on for 8 hours

5 hours on idle for websurfing at 1.55v and at load for 3 at 1.55v gaming

now with CNQ 5 of those hours are at 0.992v and the other 3 are at 1.55v

it has to add up over a month

[Mara]

I see, in my case though Im running 1.55v for 3.8ghz 24/7 so dropping from 1.55v 24/7 to 0.992v is a bigger difference.

lets say a computer is on for 8 hours

5 hours on idle for websurfing at 1.55v and at load for 3 at 1.55v gaming

now with CNQ 5 of those hours are at 0.992v and the other 3 are at 1.55v

it has to add up over a month

I agree 100% that if you can get C'n'Q working while you have increased the Vcore that high at max performance setting, it makes big difference at idle.

[Brother Esau]

Ugh, Vista... vista...

I want to get Vista working but Im going to hold it off untill at least CCC 9.1 is out At this point, I think MS should be sued for war criminality, ethnic purification against my drivers

One can only hope Ram maybe Bush will get lynched too in the Process too if we get lucky

[bobruto]

bump!

Got 2 questions about k10stat:
1) How the hell do you guys use it?
It says that clock control function is disabled in the p-state tab and also won't let me raise cpuVid for any of the p-states.
2) What language is its readme file written in? )

I'm trying to make overclock with raised voltages play nice with any CnQ replacement available for WinXP 32-bit.
CnQ disabled in BIOS, both PhenomMstTweaker and cnqsvcOC (which I altered to adjust CPU multipllier directly rather than calculate it from a rounded down desired frequency
like the original does) work very strangely: when it's time to switch to a lower p-state, they raise the voltage alright but won't raise the CPU multiplier.
So when cnqsvcOC goes from p-state 3 to 1 to 0 under load, I end up with p-state 0 cpuVid, but p-state 1 multipllier.
Same with PhenomMsrTweaker: even with its service disabled, when I try to switch p-states manually using 'activate' buttons, it will raise the voltage but not the multipllier.

Using Gigabyte GA-MA790X-UD4 with F3F BIOS, PII X3 710

[Lightman]

bump!

Got 2 questions about k10stat:
1) How the hell do you guys use it?
It says that clock control function is disabled in the p-state tab and also won't let me raise cpuVid for any of the p-states.
2) What language is its readme file written in? )

I'm trying to make overclock with raised voltages play nice with any CnQ replacement available for WinXP 32-bit.
CnQ disabled in BIOS, both PhenomMstTweaker and cnqsvcOC (which I altered to adjust CPU multipllier directly rather than calculate it from a rounded down desired frequency
like the original does) work very strangely: when it's time to switch to a lower p-state, they raise the voltage alright but won't raise the CPU multiplier.
So when cnqsvcOC goes from p-state 3 to 1 to 0 under load, I end up with p-state 0 cpuVid, but p-state 1 multipllier.
Same with PhenomMsrTweaker: even with its service disabled, when I try to switch p-states manually using 'activate' buttons, it will raise the voltage but not the multipllier.

Using Gigabyte GA-MA790X-UD4 with F3F BIOS, PII X3 710

Are you trying to increase multiplier above max. your CPU is supporting?
If not, it looks like BIOS issue. Is it working with C'n'Q enabled then?