Overclocking + Cool'n'Quiet: It can be done

CitizenInsomniac

Limp Gawd
Joined
May 26, 2005
Messages
134
I've heard an awful lot of "overclocking and cool'n'quiet don't work well together" on this board, but I've come to find that it's simply not true. I think what had mislead a lot of people is the fact that AMD's Cool'n'Quiet software sucks. It's poorly implemented and fails to expose settings for advanced users. C'n'Q the technology and C'n'Q the software/driver are two separate entities. It's possible to take advantage of C'n'Q on K8 processors without using AMD's software - while overclocking.

There are two utilities that I'm aware of that will allow you to do this: RMClock by Rightmark, and CrystalCPUID by Hiyohiyo. I'll focus on the former one because it's a bit more polished and user friendly.

The problem with AMD's software is that it only works with default CPU settings. So if you've overclocked your CPU by raising the FSB and lowering the multiplier, AMD's software won't work for you because when it tries to maximize the CPU speed it will set the multiplier to its higher default FID (and probably lower default VID), resulting in system instability. So if you're running a 11x200 1.4V chip at 10x260 @ 1.6V, you'll eventually end up with a CPU running at 11x260 @ 1.4V - and that won't go over too well.

The trick with RMClock is that it allows you to specify the minimum and maximum values for FID and VID in the section titled "P-State Transitions". Although many gurus will explain to you elaborately that FID and VID aren't exactly multiplier and voltage per se, I will for the sake of this discussion refer to FID as 'multiplier' and VID as 'voltage'. Here's what you need to do:

1. Install RMClock.
2. Open RMClock_Tweaks.reg in Notepad and change "AlternateDefaults" value to 1. This will make RMClock ignore the system default FID/VID values, particularly when exiting the program.
3. Execute RMClock_Tweaks.reg, thus importing the registry settings.
4. Run RMClock.
5. In the first tab, enable "Start minimized to tray", "Run automatically on startup" and "Restore CPU defaults on exit".
6. In "P-State Transitions" tab, set the max FID to your startup multiplier and your VID to your startup Vcore. Note that if you have any additional voltage set in the BIOS, such as +100mV or +10%, this will still get applied to your VID voltage. You can leave the min values as default for now.
7. Set both AC and Battery profiles to "Automatic Management".
8. Leave all other RMClock values as default.
9. Hit Apply and watch your CPU speed drop. :)

Things to watch for:
Make sure your CPU always has sufficient voltage to operate. If you know it needs 1.4875V to operate at max speed and only 1.475V and 1.5V are available in RMClock, set it to 1.5V - it's better to give it more juice than less. Same goes for minimum P-State though. Note: Looks like some CPUs will ignore very low VID requests. My Venice 3500+ refuses to set the VID to anything below 1.1V.

If you are using a memory divider (i.e. 6:5 aka 166), make sure your memory is never overclocked during P-State transitions. Remember that each CPU speed involves a different CPU:memory divider. Check out this A64 memory divider table. If you're running an overclocked FSB (and you probably are if you're reading this), this can be problematic. Because you're not running your memory 1:1, the actual memory speed is not the same for every FID step, as the table shows. Trouble is that some might cause your memory to run even faster than the established stable speed. Avoid those multiplier values.

Prime95's blend test seems to be the best stability test for RMClock because it doesn't always tax the CPU. However, in order to get Prime95 to work correctly, you must do this:
Open Prime95.ini in Notepad and add these values:
CpuOverride=1
CpuSpeed=2500

Naturally, you can change the latter value to whatever your CPU speed is. If Prime95 insists on running a small number of iterations for each test (i.e. 400) and tests execute way too quickly, try increasing the CpuSpeed value. This is necessary in order to prevent Prime95 from running the "slow computer" version of its tests.

If you want to test the stability of your computer at its lowest FID/VID setting, try setting RMClock's AC/battery management to "Minimal" and then running Prime95 small or large FFT for extended periods of time (16+ hrs). When you're done, return back to "Automatic Management".

Some motherboards will require you to enable the C'n'Q option in the BIOS before RMClock works, others won't. On my Asus A8V-E Deluxe RMClock works without C'n'Q enabled in the BIOS. If you have AMD Cool'n'Quiet software/driver installed, make sure you set your Power Management profile to "Always On" in order to disable it.

Finally, the download link to RMClock:
http://cpu.rightmark.org/download.shtml
 
Just to add the obvious: if you throw SpeedFan into the mix, these two apps make a killer combination. My overclocked Venice 3500+ idles at 1470MHz @ 1.1V, quiet as a mouse with its Zalman 7000 fan spinning at ~500RPM. The moment the CPU load jumps up, the full power of 2700MHz kicks in at 1.48V and the Zalman spins up to 2600 RPM. RMClock takes less than a second to kick in full blast, while SpeedFan takes about 1 minute to bring the fans to full speed (although you can make this faster if necessary).
 
New RMClock is out! You can download v1.6 at http://cpu.rightmark.org.

The new version addresses the issue I mentioned in the original post where certain FID combinations aren't good for your system because they can overclock your memory. Version 1.6 lets you define all P-States, so you can easily skip the bad multipliers.

Also, the original memory divider table I referenced requires too much math - try this table instead:
http://oskarwu.myweb.hinet.net/A64D1.jpg

In order to figure out the memory speed at each multiplier step, you'd calculate the CPU speed (FSB x multiplier, duh) and divide it with the number listed for that multiplier under the column corresponding to the memory "speed" (ratio actually) you're running. Ignore the first column that shows CPU frequency! It's irrelevant because it's based on a 200MHz FSB.
Say you're running your FSB at 245 MHz and memory at 6:5 ratio (aka 166, aka 333). The values would go like this:
4.0 x 245 = 980 / 5 = 196
5.0 x 245 = 1225 / 6 = 204.16
6.0 x 245 = 1470 / 8 = 183.75
etc.
So let's say you know for a fact that your memory can't run at your current CAS timings faster than 200 MHz. In that case 5.0x would be a bad combination and you'd want to avoid it.
 
thanks for posting this.. i have pointed out to quite a few people that this program exists, and it's really just a limitation in amd's software/drivers :D
 
No problem, I'm happy to contribute. I just wish more people would use it! :)

With the latest RMClock and Speedfan I have my 3500+ system idling at 980MHz @ 1.2V (my mobo adds +0.1V by default, otherwise it'd be 1.1V) with Zalman 7000 fan running at 500RPM.
 
Why not just set your power scheme to "Minimal Power Management"? It does the exact same thing :confused:

BTW, I dont have a "P-State Transitions" tab in RMClock, nor is the "Restore CPU defaults on exit" checkbox clickable.

One more thing, I see this in "RMClock_Tweaks.reg", but I cant figure out how to set it to "1" :eek:
AlternateDefaults"=dword:00000000
Other values that are by defalt a 1 have "00000064" set as their dword, does this mean I change: "AlternateDefaults"=dword:00000000" to "AlternateDefaults"=dword:00000064" to set it to 1 ???
 
Unknown-One said:
Why not just set your power scheme to "Minimal Power Management"? It does the exact same thing :confused:

Not without some sort of driver support. I don't think XP has built in support for P-State management.

Unknown-One said:
BTW, I dont have a "P-State Transitions" tab in RMClock, nor is the "Restore CPU defaults on exit" checkbox clickable.

Do you have an AMD CPU that supports Cool'n'Quiet (aka PowerNow)?

Unknown-One said:
One more thing, I see this in "RMClock_Tweaks.reg", but I cant figure out how to set it to "1" :eek: Other values that are by defalt a 1 have "00000064" set as their dword, does this mean I change: "AlternateDefaults"=dword:00000000" to "AlternateDefaults"=dword:00000064" to set it to 1 ???

Just change the last 0 to 1.
"AlternateDefaults"=dword:00000001
 
Just followed the instructions laid down, and it's sitting here at my same overclock 10x235...won't to down or lower voltage. :/ Anything you can think of? And yes, I enabled CnQ in the BIOS.
 
tsuehpsyde said:
Just followed the instructions laid down, and it's sitting here at my same overclock 10x235...won't to down or lower voltage. :/ Anything you can think of? And yes, I enabled CnQ in the BIOS.

Which CPU and motherboard do you have? Can you attach screenshots of your RMClock main tab and the PST tab?
 
from his sig:

Athlon64 3400+ CG Claw @ 2.35GHz 1.57v with XP-120 on MSI K8N Neo Platinum (BIOS v1.7)


owning a k8n neo plat and 3200+ clawhammer myself, i know c&q works on it.
 
Sorry to break in on the thread with my own question but since it's CnQ related thought it would be fine.

I seem to have broken Cool N Quiet. It's enabled in the bios and I have set the power management to 'Minimal' in the power options. Do I need to have all the options in the bios set to Auto for voltages? Not sure I have that set since it undervolts my memory. What else could I have done wrong? or could I have detrementally changed the 'Minimal' profile in some way?

Any ideas? Thanks

Edit: Unless its broken in bios 1011 final...
 
(cf)Eclipse said:
from his sig:

Athlon64 3400+ CG Claw @ 2.35GHz 1.57v with XP-120 on MSI K8N Neo Platinum (BIOS v1.7)

tsuehpsyde: Can you make sure the "Cool'n'Quiet Control" option in the Cell menu of your BIOS is set to "Auto"?

Do you have the AMD C'n'Q driver installed? If you do, can you make sure your Power profile is set to Always On in the Control Panel?

(cf)Eclipse said:
from his sig:
owning a k8n neo plat and 3200+ clawhammer myself, i know c&q works on it.

Any advice for him?
 
starfury6 said:
I seem to have broken Cool N Quiet. It's enabled in the bios and I have set the power management to 'Minimal' in the power options. Do I need to have all the options in the bios set to Auto for voltages? Not sure I have that set since it undervolts my memory. What else could I have done wrong? or could I have detrementally changed the 'Minimal' profile in some way?

Setting Power Management is necessary only if you're using the AMD C'n'Q driver. If you're using RMClock, the AMD driver will only get in the way, so make sure you set Power Management to "Always On".
 
This is like cool man heh heh. I'll be wasting some hours dialing this in.

I sure wish ACPI worked properly with this board. I'm hesitant to try the ACPI functions with suspend/standby causing my sensors to completly freak.
 
I was wondering about this myself. Up to this point I had just not used the CnQ feature, but now that I know I can... I will. :cool: Thanks!

- James
 
I have 2 rather minor complaints.
1 I wish RMclock accepted values above 1.45V
2. I wish it allowed my core accepted values less than 1.1V. Other than that I'm very pleased so far.
 
Autochthon said:
I have 2 rather minor complaints.
1 I wish RMclock accepted values above 1.45V
2. I wish it allowed my core accepted values less than 1.1V. Other than that I'm very pleased so far.

Neither are RMClock problems. They are properties of the CPU and motherboard. From the FAQ:

Q: CPU voltage indicated by RMClock greatly differs from the voltage reported by my vendor-supplied monitoring software. What's wrong?

A: Nothing. RMClock reports the CPU VID, i.e., just the requested CPU voltage level. It is up to the motherboard logic to decide whether this request should be handled or ignored. Most motherboards designed to properly support processor's power-saving features (e.g., AMD Cool`n'Quiet(tm) technology) should adjust the real CPU voltage so that it's close to the requested VID level. Other models may add some voltage delta (e.g., 0.1 or 0.2V), which is useful for overclocking. Finally, there are many motherboard models that can be set up (via BIOS settings) to supply constant voltage to the CPU, which is far from being good for proper power saving.


Q: Why can't I set CPU voltage above 1.55V (or even 1.45V) on my Athlon 64? BIOS setup offers the voltages up to 1.7V. Is there a way to unlock these "hidden" voltages?

A: RMClock manipulates with VID and not directly with a CPU voltage. Maximal VID value you can set is dictated by the CPU itself (as indicated in the "Maximal VID" field of the General tab). 1.55V is the absolute maximum VID for all Athlon 64 processors family; usually, maximal VID is even lower (e.g., 1.45V). There's no way to set current VID value higher than the maximal, as the CPU will choose maximal value instead (the same is true for FID, the bus clock multiplier). The voltage level you adjust in the BIOS is set in entirely different way, via direct manipulation with the CPU VRM and not involving VID changes. Please also note that the voltages you set manually in the BIOS setup cause the motherboard logic to ignore VID change requests, thus greatly diminishing power saving features of the AMD Cool`n'Quiet(tm) technology.
 
mattthemodder said:
Im at 803Mhz on 0.8V. Feels fast enough for Firefox and Winamp :p Temps are 30C on stock too

Nice one! Make sure to test that your system is stable at 0.8V - this can be kind of tricky since any kind of heavy computing will prompt RMClock to raise the voltage when in automatic management mode. Try temporarily switching RMClock's AC Profile to "Minimal" and then running Prime95 like that for 16+ hrs.
 
Please also note that the voltages you set manually in the BIOS setup cause the motherboard logic to ignore VID change requests, thus greatly diminishing power saving features of the AMD Cool`n'Quiet(tm) technology.

hmm does this mean that by setting my voltage to 1.45 * 110% in the bios that the 1.1V that RMclock monitor shows is incorrect. I'd assumed that while the Cpu was receiving 1.45V * 110% it was being modified by the CnQ technology down to 1.1V * 110% for use by the core.
 
Autochthon said:
hmm does this mean that by setting my voltage to 1.45 * 110% in the bios that the 1.1V that RMclock monitor shows is incorrect. I'd assumed that while the Cpu was receiving 1.45V * 110% it was being modified by the CnQ technology down to 1.1V * 110% for use by the core.

Actually, I'm not sure what that sentence is all about. You can try asking the author over at http://forum.rightmark.org/?id=6.

In my experince, VID requests will override the startup voltage defined in the BIOS. However, any "bonus" voltage you've defined in the BIOS will still apply. So yes, your assumption is correct - if RMClock requests 1.1V, you should be getting ~1.21V back.
 
After careful rereading of the readme file and some experiments. I have worked it out.

AS it states you CAN NOT have a manual voltage setting in the bios. All VID requests are ignored if bios CPU voltage is other than "auto". This complicates issues for those using the 5 volt rail for memory voltages from 3.3 - 4V.

I set voltage to auto and now my monitoring programs show my voltage changing as per VID requests. Basically I changed from 1.45V * 110% to auto * 110% and it's working now. As a result, cpu temps dropped a good 5C. Also set startup voltage to startup.
 
Nice one! Make sure to test that your system is stable at 0.8V - this can be kind of tricky since any kind of heavy computing will prompt RMClock to raise the voltage when in automatic management mode. Try temporarily switching RMClock's AC Profile to "Minimal" and then running Prime95 like that for 16+ hrs.
Will do. We really should have a little underclocking comp, the A64's are great at it!
 
Autochthon said:
AS it states you CAN NOT have a manual voltage setting in the bios. All VID requests are ignored if bios CPU voltage is other than "auto". This complicates issues for those using the 5 volt rail for memory voltages from 3.3 - 4V.

I set voltage to auto and now my monitoring programs show my voltage changing as per VID requests. Basically I changed from 1.45V * 110% to auto * 110% and it's working now. As a result, cpu temps dropped a good 5C. Also set startup voltage to startup.

Interesting. I only have one CPU voltage setting in the BIOS, the aptly named "Hammer Vid control" and I set that to a manul setting of 1.375. RMClock has no problem working with this setup. It sounds like you have two settings in the BIOS related to CPU voltage?
 
My DFI bios has 3 cpu settings. 1.startup voltage 2.Cpu voltage and 3.an additional cpu multiplier. The cpu voltage ranges from something that I don't recall ;) up to 1.55V and auto. The multipier is in increment of 104% 110% 113% 123% 126% 133% 136%. I may have forgotten some in there.

When set to something other than "auto" VID requests are ignored and though FID changes VID doesn't. I'll test a bit more later to make certain I'm not leading you astray.
 
Back
Top