Undervolting a Phenom II X4 (HDXB97WFK4DGI)

There is no program to do that "without creating instability." You will have to test for stability each time you lower the voltage with a program like prime95 or OCCT.

To actually change the voltage, the best and most permanent would be with the bios. It is also easier than the next option I'm going to suggest.

K10stat is the program I use with my laptop because I do not have access to the voltage settings in the bios. It's a bit tricky to set up to change the settings on boot but once you get it to work it works well.
http://sites.google.com/site/k10stat/

If you want to change the settings in the bios, an option it to change the voltage settings on demand with K10stat while testing for stability so you know what you can set the voltage to in the bio safely. Test with Prime 95 for maybe 10 minutes each time. Once you find a settings, run prime95 for 8 hours to ensure stability.

I can see you going maybe to 1.2-1.3V.

Now I am considerung to lower Vcore as I would believe there is some safety margin.

I have no clue what you mean by that. Why do you want to undervolt? To reduce heat or to reduce power consumption?

Sounds like we wants to be able to turn down the CPU fan for a more silent build and save a little on his energy bill while he's at it. I've never been fond of this approach, as I prefer to overclock every bit of extra performance I paid for, but it has its uses I suppose.

There's actually an article on undervolting Phemon II's here on TH.
http://www.tomshardware.com/reviews/undervolt-cpu-pheno...

ASUS TURBO V is a great utility, THOUGH i wudnt go too far, as my x3 425 wasnt even stable with 1.12v

Hi,

the CPU HDXB97WFK4DGI is the business version which should have some energy saving potential.
Target: throttle the fans and save some € when using the PC as a server at 24/7.
As most of the time the CPU will run at 4x 800MHz the reduction of Vcore (idle) will have some impact. At 3.2GHz I have to accept some heat as the price for performance.

In the Bios I have not found the button to set the Vcore manually. Will try the ASUS turb V

Sunzi from Munich

If you ever decide to try K10stat, you can set the voltage for each P-state. Like I said 1.2-1.3V at best for at 3.2GHz, but for the lowest Pstate, you may even get as low as 0.8V

On my company notebook I am using RMclock 2.35 which throttles down the noisy fan significantly when dynamically reducing the Voltage from 1.2V to 0.9V.
However, at any new boot up I have to go into settings->power options->rmclock profile.
I guess K10stat will work similarily

Sunzi

Yes it does. But you can put a shortcut to it in the startup folder. I have it set up on my other computer. Under properties there are additional functions you have to add to ensure it reapplies the settings once you log in. I will copy and past them here tomorrow when I am on my other computer.

Hi Enzo Matrix,

thank you for your input.

Below article (unfortunately in German language) describes that the C3 stepping is an even better candidate for undervolting than the previous C2 stepping. Some user report that a Phenom II X4 965 has worked in P3 at 0.7V (instead of 0.912V) which I find quite remarkable.

"Phenom C3-Stepping mit geringerer Leistungsaufnahme verfügbar

Autor: Bjoern Bookmark and Share

AMD hat seine Phenom II CPUs überarbeitet und inzwischen sind drei neue CPUs im C3-Stepping verfügbar: Der Phenom II X4 945, 955 sowie 965 . Im C3-Stepping wurden kleinere Fehler behoben, aber auch die Leistungsaufnahme dürfte deutlich gesenkt worden sein.

PrintDas C3-Stepping erlaubt es unter anderem endlich auch bei einer Bestückung mit vier Speichermodulen diese mit voller Taktfrequenz anzusteuern, vorher wurde der Takt in diesem Fall auf 1067 MHz abgesenkt. Weiterhin sollen die CPUs mit geringeren Spannungen angesteuert werden können. Im Fall des Phenom II X4 965 hat dies sogar zu einer Absenkung der TDP von 140 auf 125 Watt geführt. Bei den beiden anderen Modellen ist die TDP gleich geblieben. Trotzdem dürfen sie bis zu 10 Watt weniger an Leistung aufnehmen, als ihre C2-Kollegen.

Der User mrtea berichtet in unserem Forum sogar, dass er seinen Phenom II X4 965 im Idle mit einer VID von nur 0,7 Volt stabil betreiben kann. Ein beneidenswertes Undervolting-Ergebnis, zumindest für Besitzer von Phenom-CPUs mit einem älteren Stepping."

Sunzi

My German's a little rusty.
But yes the C3 stepping does provide power improvements. But I thought you had the HDXB97WFK4DGI?

Yes, that is correct it is the DGM which is the C3 stepping.


Sunzi

Ok so if you decide to try K10stat at any point:
-change profile 3 to whatever you want the settings to be and click apply
-put a shortcut to the exe in the startup folder
- append " -lp:3 -ClkCtrl:2 -StayOnTray -nw" to the end of the target without the quotes and with the space at the beginning. This will make it automatically apply profile 3 when you log in and keep an icon in the system tray.
-You can right click on said icon anytime and change the profile 3 settings.

THX.

Will try.

Sunzi

Here are my observations using K10stat, CPU-Z, Prim95:

Starting with P-state 0 (3200 MHz) the Vid was gradually reduced starting from 1.3750V and using Prim95 to provoke full load (=heat)
When the µP is still "cold" Vid can be reduced down to 1.200V on the way down the µP seems to turn down one out of four cores before you will provoke a bluescreen at about 1.200V.
The hotter the CPU becomes the higher the bluescreen threshold gets which is due to the laws of physics. To run stable at least with the CPU I tested a voltage of 1.275Volt is recommended which will save about 6W of heat/noise.

On the other end at P-state 3 (800MHz) the CPU is surprisingly tolerant to run at lower than specified voltages. I had it running for quite a while at 0.688V without issues. To be on the safe side maybe 0.7V is a good number.

Again all observations I have made apply to the unit I have tested. Another unit maybe a litte worse or better but you get a decent ballpark number.

K10Stat would also allow you to undervolt the Northbridge (NB) but I have not made any extensive tests of how far you can or should get.

Maybe somebody has tried that one and is willing to share somexperience.

It is worth trying to get the system cooler/quieter/more energy efficient by approaching the individual limits og the CPU - I find

Cheers

Sunzi