K10stat: Voltage Scare

HoTsHoT · Feb 2, 2010

I have a 550 BE just unlocked to an X3, and my PC has this problem of restarting when jumping from CnQ voltage to normal OCed voltage.

My 550 BE runs at about 900 MHz on 0.75 Vcore, 1.25 CPUNB. Once it detects load (Prime95 Small FFT), K10stat ramps up the speed to 3.76GHz, 1.475 Vcore, 1.25 CPUNB. It used to be fine on dual, but it occurs more frequently on tri.

I raised the CnQ Vcore to 1.1v, and the problem goes away. Could this problem be caused due to a shock - where the power requirements sharply rise in a short period of time, causing the PC to hang? I could very well solve the problem with CnQ VCore at 1.1v, but I would like to save as much power as possible, my rig's on 24/7

Thanks.

Dolk · Feb 2, 2010

It doesn't surprise me that this happens. The unlocked cores are locked for a reason. The trigger you found is probably one of those reasons. I suggest that you try not to use CnQ at all with the unlocked core.

QuietIce · Feb 3, 2010

I don't think it's so much the unlocked core as CnQ v OC. There's a ramp-up in both speed and voltage that happens when CnQ kicks out - hundreds of small adjustments that are made over a short period of time. Obviously one of those small steps, probably in the overclocked region for speed, doesn't have enough voltage for the speed it's trying to run. Here's a rough example with only four equal steps and a fifth in-between step:

speed vCore
0900 - 0.750
1615 - 0.931
2330 - 1.112
3045 - 1.294
... 3402 - 1.384
3760 - 1.475

My guess is your hang is somewhere in that bracket between 3045 and 3760 GHz ...

HoTsHoT · Feb 3, 2010

Dolk said:
It doesn't surprise me that this happens. The unlocked cores are locked for a reason. The trigger you found is probably one of those reasons. I suggest that you try not to use CnQ at all with the unlocked core.

I had this problem on a dual before, it was a very rare occurrence but now on the X3 it happens more than ever. I did try attribute it to the unlocked core, but if it also happens on the dual then it's probably not the cause.

QuietIce said:
I don't think it's so much the unlocked core as CnQ v OC. There's a ramp-up in both speed and voltage that happens when CnQ kicks out - hundreds of small adjustments that are made over a short period of time. Obviously one of those small steps, probably in the overclocked region for speed, doesn't have enough voltage for the speed it's trying to run. Here's a rough example with only four equal steps and a fifth in-between step:

speed vCore
0900 - 0.750
1615 - 0.931
2330 - 1.112
3045 - 1.294
... 3402 - 1.384
3760 - 1.475

My guess is your hang is somewhere in that bracket between 3045 and 3760 GHz ...

Ah, I wasn't aware that CnQ (or K10Stat) ramps up in small steps rather than just jumping from the CnQ speed to the base speed.

So far in my K10Stat, I've configured P3 to be the CnQ state, the other states are in overclocked speed as I don't need too many states for no reason. I've tried playing with the ACC percentages - nothing. The only way was to increase the CnQ voltage.

My initial guess on why it would not work was due to the extra power requirements due to the unlocking of the core, therefore causing a restart every time due to insufficient voltage. Or, it could a sudden demand in power and thus the restart due to some form of current protection (or somewhat).

QuietIce · Feb 3, 2010

CnQ does for sure - don't know about K10Stat or anything else but they'd almost have to follow that same process or it would never work. You're talking about a piece of hardware running in the GHz, which means over 3 cycles every nanosecond. No way a power chip is going to ramp up that fast. Even if ramp-up is only 1/1000th of a second that's 3 million CPU cycles ...

AndreaACME · Feb 11, 2010

I have a question regarding K10Stat, which I'm using to keep OCing and UCing working together.

Example:
If I set a mild OC ( [email protected] ) by using multiplier FID=46 ( 6200MHz ) AND divider DID=1 ( div2 ) I get a predictable freeze as soon as the state is entered. If I set the same frequency using only the multiplier FID=15 ( 3100MHz ), DID=0 ( noDiv ) then it runs stable. So why don't I use only the multiplier? Because FID increments are too big in frequency.
The strange thing here is that for state P3 I actually UC using the same technique to reach as low as 800MHz FID=0 ( 1600MHz ), DID=1 ( div2 ) and it works fine.

Any ideas on this behaviour? Could this be to do with those intermediate ramping steps you were talking about?

Code:

CURRENT STABLE P_STATES

State - CoreSpeed - VCore   - VIMC
P0    - 3000MHz   - 1.2750v - 0.9750v
P1    - 2000MHz   - 1.0125v - 0.9500v
P2    - 1500MHz   - 0.9000v - 0.9500v
P3    -  800MHz   - 0.8500v - 0.9375v

Unganged (40%/20%, 50%/20%, 60%/20%) (100ms/1000ms)

QuietIce · Feb 11, 2010

That's possible. I've never read any details on how the P-states work exactly. Years ago I found a very good article about CnQ, which is where I got my info from. The P-states may move through each individual state from P3 to P0 - though I still can't see how they can avoid those baby-steps in between. It's possible that by using DID=0 you've added another "gear" so-to-speak for the routine to use by moving from DID=1 to DID=0 somewhere in the middle.

But that's all speculation on my part. I've not found a really good explanation of exactly how P-states progress from one to the other and in all honesty I haven't looked very far for it either. When CnQ came out I was interested but since I run SETI now the only modes my machines know are off (for maintenance) and 100% load ...

AndreaACME · Feb 20, 2010

Ok, just a quick update on my previous messages: tested some more on this "divider" problem, found out that I can actually use any multiplier AND / OR divider I wish and the way to do it is to NEVER change the values of multi/div of the P_state you're in. Simply force another P_state (and be sure to keep it there), then change all the settings and when finished simply enter the new P_state.

It all works very well now. Amazing ... I now have a profile where I'm overclocked to 3.5GHz on P0 and P3 goes right down to "deep-frozen" 400MHz !! I'm so happy now that I managed to achieve my goal of setting up a very flexible profile for all-purpose 24/7 use.

Regards.

HoTsHoT · Feb 20, 2010

AndreaACME said:
Ok, just a quick update on my previous messages: tested some more on this "divider" problem, found out that I can actually use any multiplier AND / OR divider I wish and the way to do it is to NEVER change the values of multi/div of the P_state you're in. Simply force another P_state (and be sure to keep it there), then change all the settings and when finished simply enter the new P_state.

It all works very well now. Amazing ... I now have a profile where I'm overclocked to 3.5GHz on P0 and P3 goes right down to "deep-frozen" 400MHz !! I'm so happy now that I managed to achieve my goal of setting up a very flexible profile for all-purpose 24/7 use.

Regards.

Could you give more clarification on how it's done? I'm currently running my PC on full speed 24/7, would love to save some watts this year

QuietIce · Feb 21, 2010

AndreaACME said:
Ok, just a quick update on my previous messages: tested some more on this "divider" problem, found out that I can actually use any multiplier AND / OR divider I wish and the way to do it is to NEVER change the values of multi/div of the P_state you're in. Simply force another P_state (and be sure to keep it there), then change all the settings and when finished simply enter the new P_state.

It all works very well now. Amazing ... I now have a profile where I'm overclocked to 3.5GHz on P0 and P3 goes right down to "deep-frozen" 400MHz !! I'm so happy now that I managed to achieve my goal of setting up a very flexible profile for all-purpose 24/7 use.

Regards.

Nice!

I agree with HoTsHoT, it would be nice if could go into more detail on how you did that so others can do it too ...

AndreaACME · Feb 21, 2010

Get a copy of K10stat from the internet ( go for a Google search: http://www.google.com/search?hl=en&q=allinurl:k10stat+download&btnG=Search&aq=f&aqi=&oq= )
Run it to configure your states temporarily and do all the necessary testing before deciding to put it in Windows auto-launch
Advice:

Always be aware of the present P_state you are in before making multiplier or divider changes. If necessary then: switch to another P_state, change the settings as desired and then change back to the new setting to test it. Changing voltage of current P_state is OK

Always leave the first profile as your 100% stable profile because K10stat will default to that when launched so don't "lock yourself out"; so mess around only on profiles 2 thru 5

All the frequencies that you'll see in K10stat are based on standard 200MHz FSB which means that if you overclock the bus, you will not see the true frequency values in K10stat.
For example, setting a 3000 (FID=14) on a 210MHz bus in K10stat will show you 3000 but with CPU_Z you will see the real frequency which is 3150MHz! So keep in mind this K10stat bug when you apply your voltages
For example, here are my current "DeepFreezeOC" profile settings:
Code:
```
|Tab: P-state|

Profile:  DeepFreezeOC

CnQ
State - AllCores  - CPU(V)  - NB(V)      up%   down%         
P0    - 3500MHz   - 1.4375v - 1.0000v    88    50
P1    - 3000MHz   - 1.2750v - 0.9750v    64    50
P2    - 2000MHz   - 1.0125v - 0.9500v    50    32
P3    -  400MHz   - 0.6875v - 0.9375v

Unganged (88%/50%, 64%/50%, 50%/32%) (1000ms/2000ms)
```
NOTE:

This profile was intended for normal 3GHz operation (in OS or Internet or Wordprocessing everyday use) with an OC kick-in only over 88% load (for heavy duty programmes or gaming use) and a powersaving deep-cool frequency of 400MHz (for idling). In other words it is a very powersaving oriented profile but with a sting in its tail. To obtain the various frequencies just play with the FID and DID and you'll get very accurate stepping. Adapt the profile to your needs/limits and note that these values work with MY CPU and all values are to be fine-tuned to YOURS!

After you stabilize your settings you can auto-launch it at Windows startup by creating a new task in TaskScheduler:

Code:

|Tab: General|
(ENABLE) Run only when user is logged on
(ENABLE) Run with highest privileges

|Tab: Action|
(CHOOSE) Start a program
Program/script: <DriveLetter>:\<PathOfInstallation>\K10STAT.exe
Add arguments:  -loadprofile:1 -StayOnTray -ClkCtrl:1 -NoWindow
Start in:       <DriveLetter>:\<PathOfInstallation>\

|Tab: Conditions|
(All DISABLED)

|Tab: Settings|
(ENABLE) Run task as soon as possible after scheduled start is missed
(CHOOSE) Do not start a new instance

Reboot or if you want it to start immediately right click the task and "Run" (be sure that no other instance is running already!)

Finding your stable P_states

Best way (and also the quickest IMHO) to find you stable states is to:

Decide which frequencies you will want to have as a MUST, for example: overclocked at 3.5GHz, normal 3GHz, frozen to 400MHz.
Ok, so you only have 4 states and you have decided three of them already so look for an intermediate fourth state: in this case ((P2-P3)/2)+P3 hence ((3000-400)/2)+400)=1.7Ghz.

In my profile I actually set it to 2GHz because I found that with my HW it performs well when I'm on the web looking at HD YouTube videos; 1.7GHz wasn't quite fluid enough and sometimes triggered the 3GHz state to compensate.
Open AOD (AMD Over Drive) and make sure K10stat is not running and AOD top right state indicator is a green disc with a red circle, set the first frequency goal and start with a higher voltage
Launch Prime95 in SmallFFT (make sure the Advanced > "Round off checking" is ENABLED) and wait 10-30sec. Note the voltage down and then push the voltage a notch lower and re-iterate until it freezes. The last successful value is your limit so consider testing for longer (hours) at 1 or 2 notches higher for stable operation

Once all the four frequencies and voltages have been found, you're ready to put them in K10stat's table of one of the profiles. Make sure AOD isn't running and that K10stat is in ClockControl mode (rightclick the icon in SysTray and "Enable Clock Control"). You're done!

This topic is so complex and vast that I hope I covered at least the essentials but for further knowledge I would suggest you read about:

K10stat guide / manual (-link not found-)
Guide to the Phenom II (http://www.overclockers.com/forums/showpost.php?p=5985621&postcount=2)
Heat dissipation methods (-link not found-)

Regards,
Andy

QuietIce · Feb 21, 2010

VERY nice - thanks for that detailed guide! :thup:

Bookmarked for future reference and referral ...

HoTsHoT · Feb 21, 2010

Excellent work there AndreaACME, it should serve as a great guide for K10stat users who have difficulty understanding the program. Great job!

Lothyr · Jul 26, 2012

K10Stat

Hi,

I realize this thread is roughly 2+ years old.

Can someone help me with K10Stat?
Preliminary experimenting has indicated K10Stat has solved a problem I was having, but I am not 100% sure I have a "good" K10Stat configuration...and I am still wondering how K10Stat affects BIOS settings ie. Does K10Stat Override a Bios Cpu FID? Why have any change in P-State at all? Except for Resets/Boots/Crashes/Power up/downs, why would I want a lower P-State? Sorry if this is ignorant question...Aren't these P-States mainly to save Power? In other words, if I don't care about how much juice my PC is consuming all the time, and I just want 1 constant frequency/power state (unless BIOS needs something specific for errors or w/e), is this possible, or maybe negative consequences?

I guess if someone replies, I'll post some questions

. It's an extremely useful Utility, but I might inadvertently screw things up as I am want to do...

P.S. I suppose I should at least explain the original problem a bit. Battlefield 3 was crashing/hanging/stuttering all the time. Very frustrating. I read about different solutions, etc. but PC still failed with BF3. Tried lots of Bios settings, tweaks, etc. Nothing was working, so I went back to stock/defaults on everything.
:fight:

Was able to rule out various possible problems though trial and error. Started monitoring everything I could via AIDA64...I noticed that even when I boosted NB vid in Bios, the voltage reported stock settings...I am using a NB = 2600MHz and I knew I needed more than stock voltage (around 1.175-1.2v). So to cut to the point, I "think" what may have been happening is that during BF3 gameplay the P-State lowered and triggered a lower NB vid, basically mucking up Northbridge in a situation where NB performance was needed, hence an "AppHang" or more commonly a "frozen" BF3 reporting it is still running in TaskManager, but no Video...something along those lines...

So my layman's approach after reading a bit was to "widen" the gap between P0 & P1 so that any State that BF3 drops in to or rises up to, is within that P0 to P1 range...basically I was trying to Force my PC to think it needed Full Power whenever running BF3...Will this work?

Thanks.

AndreaACME · Jul 26, 2012

Lothyr said:
Hi,

I realize this thread is roughly 2+ years old.

Can someone help me with K10Stat?
Preliminary experimenting has indicated K10Stat has solved a problem I was having, but I am not 100% sure I have a "good" K10Stat configuration...and I am still wondering how K10Stat affects BIOS settings ie. Does K10Stat Override a Bios Cpu FID?

YES, K10Stat does override those settings at login time (not before). This means that if you set your BIOS settings too low (undervolted) or too high (overheating) to boot then you have locked yourself out!
Always be conservative with changes, a bit at a time.

Why have any change in P-State at all? Except for Resets/Boots/Crashes/Power up/downs, why would I want a lower P-State?

Changes in P-States are desirable when you want to save power or overboost your system accordingly and dynamically with workload variations on the processor. I use mine to save energy and cool down the processor when it is not used much and also boost/overclock it when my load goes above high levels so that it is not overclocked all the time.

Sorry if this is ignorant question...Aren't these P-States mainly to save Power? In other words, if I don't care about how much juice my PC is consuming all the time, and I just want 1 constant frequency/power state (unless BIOS needs something specific for errors or w/e), is this possible, or maybe negative consequences?

YES, using K10Stat to make all P-States the same value is useless, you would be better off setting the Cool'n'Quiet OFF altogether and setting the desired FID in the BIOS; that will remain throughout.

I guess if someone replies, I'll post some questions . It's an extremely useful Utility, but I might inadvertently screw things up as I am want to do...

P.S. I suppose I should at least explain the original problem a bit. Battlefield 3 was crashing/hanging/stuttering all the time. Very frustrating. I read about different solutions, etc. but PC still failed with BF3. Tried lots of Bios settings, tweaks, etc. Nothing was working, so I went back to stock/defaults on everything.

Was able to rule out various possible problems though trial and error. Started monitoring everything I could via AIDA64...I noticed that even when I boosted NB vid in Bios, the voltage reported stock settings...I am using a NB = 2600MHz and I knew I needed more than stock voltage (around 1.175-1.2v). So to cut to the point, I "think" what may have been happening is that during BF3 gameplay the P-State lowered and triggered a lower NB vid, basically mucking up Northbridge in a situation where NB performance was needed, hence an "AppHang" or more commonly a "frozen" BF3 reporting it is still running in TaskManager, but no Video...something along those lines...

So my layman's approach after reading a bit was to "widen" the gap between P0 & P1 so that any State that BF3 drops in to or rises up to, is within that P0 to P1 range...basically I was trying to Force my PC to think it needed Full Power whenever running BF3...Will this work?

Thanks.

In general, I have never seen an application which crashes due to stock settings on a MB. If that happens it means that there are some other issues which are changing the operating conditions of the MB (i.e. a cluttered MB with blocked fans will rise in temperature causing randomic crashes and several other problems as a secondary effect).
I really don't think you will solve your problem by changing the P-State settings; maybe you will lower the temps in the MB which will help to stabilize the system but I wouldn't take this as a "solution" but only a workaround. You must do further technical scouting on your system to pinpoint a potential problem.

Use other debugging tools which READ values from sensors around your MB (like SpeedFan) and start logging them to disk so that you have data in the precise moment BF3 crashes. Check for temps and speeds and move on from there one step at the time until you get face to face with the primary problem.

Lothyr · Jul 26, 2012

YES, using K10Stat to make all P-States the same value is useless, you would be better off setting the Cool'n'Quiet OFF altogether and setting the desired FID in the BIOS; that will remain throughout.

Hmmm, maybe I am missing something.

BTW, I read most of your other posts: note to self: setup a stable default K10Stat...

The steps I take are:
1. Boot PC (check! lol)
2. Wait till everything loads up
3. Start K10State.exe, enable clock control, set Control Function to "UnGanged" {Not sure about that one, but Ganged seemed "Bad" lol}, Choose "Gaming" {My custom P-State, mainly P0 & P1 changes} and start, well, gaming..lol. I am pretty sure I have not set this up right - I'll give it a go today and see if I can improve settings...

I really don't think you will solve your problem by changing the P-State settings; maybe you will lower the temps in the MB which will help to stabilize the system but I wouldn't take this as a "solution" but only a workaround. You must do further technical scouting on your system to pinpoint a potential problem.

Noted. {sigh} back to research...In my own defence I am recording info to a spreadsheet and setting it up to be my guide/reference. I have a journal filled with every setting and combination I have tried thusfar...Basically I am trying to zone in on a "Performance Envelope". IOW, a cloud of values and settings which offer a stable Overclock on my system. I am not a computer engineer so it is taking me a while to zoom in on important features and I probably am ignorant of some important details...
Example: 472 pages of "BIOS and Kernel Developer’s Guide (BKDG) For AMD Family 10h Processors" seems really relevant, but somewhat obscure to me...

It's a puzzle for sure.

Use other debugging tools which READ values from sensors around your MB (like SpeedFan) and start logging them to disk so that you have data in the precise moment BF3 crashes. Check for temps and speeds and move on from there one step at the time until you get face to face with the primary problem.

Hmmm, I "think" this is what I have been doing...I use AIDA64's Sensor Panel w/ all sensors actively displayed - no logging though, except by eyeball {lol}.

My Caveman-like intellect deduced that NB VID was staying at 1.1v even though I had set BIOS to 1.2v...repeated several tests to monitor this value when everything else was within reasonable tolerances (Temperatures/Voltages/Utilization, etc.)
Therefore, I tried "forcing" a 1.2v on the NB in all situations w/ K10Stat other than say 50% CPU and below (browsing/video playback, general PC use)

BTW, you mentioned Cool'n'Quiet, I thought I was supposed to disable this by default when Overclocking? Same for C1E & SVM... Does Cool'n'Quiet need to be enabled w/K10Stat?

Thank you for your time and patience...

keny · Jul 26, 2012

@Lothyr, it looks like from you're description of you're problem that you are changing the wrong setting in you're bios.....you say you are upping youre NB to 2600 and bumping you're NB vid a notch, well the NB vid controls the motherboard NB (onboard gpu/fsb)and not the CPU NB which is the NB that you are overclocking, to stabilise the CPU NB you need to bump the CPU NB voltage up which inturn will cause a bit more heat through the CPU, do make sure you're cooler can cope (were talking small increase in heat not huge). Hope this helps you out :thup:

Lothyr · Jul 26, 2012

:

Aaargh! Ok, I will check this.

Thanks!

Lothyr · Jul 26, 2012

Somethings amiss.

Example:

Bios Settings:

Manual Mode Voltages:
CPU:1.5
CPU/NB: 1.2
CPU VDDA: auto
Dram: Auto
NB: 1.175
NB HT: auto
NB 1.8v: auto
SB: auto

But {theres always a but right?} AIDA64 shows these values (on overclock tab):

Thats when I looked in to K10Stat. When I plugged 1.2v in to NB VID for P0 & P1 the values in AIDA64 started showing...1.2v...see where I am going with this?

So something not right, am I right? :shrug:

keny · Jul 26, 2012

The bios rules as far as overclocking goes, software programmes are glitchy are unreliable, I only tend to use software to squeeze the last few MHz out of a overclock for benching only, I certainly would not use any sort of overclock software for a 24/7 system, trust me , if you set it in the bios and save before exiting, it will be set unless it fails during boot up and resorts back to stock to get the pc back up again, but that's not the case here, personally I would remove all the software mumbo jumbo you've got installed and put the CPU/NB at 1.3v and test you're overclock from there, I wouldn't worry about AIDA or other progs they just make things harder, and nobody needs that

K10stat: Voltage Scare

Registered

I once overclocked an Intel

Disabled

Registered

Disabled

Registered

Disabled

Registered

Registered

Disabled

Registered

Disabled

Registered

Registered

Registered

Registered

Member

Registered

Registered

Member

Similar threads