Undervolting - VRM thermal issues

[cubusmedusa]

Hi,
for casual usage I usually have my CPU clocked to 2400MHz, HT disabled, everything else set to auto. Recently I undervolted the CPU just for the fun of it. I set offset to "-" and worked it down to -100mV running at ~0,9V. It worked fine. I thought that undervolting is good due to energy saving, better thermals, etc. I opened my pc case to work on my cabeling and accidentaly touched the VRM heatspreader of my P8Z68-V Pro Gen3 (it also has an array of VRM mosfets on the backside of the mobo). It was extremely hot on both sides. No more than 2 sec contact possible. I reset the voltage and VRM thermals are normal again. Is this a known issue or are the undervolters ignoring VRM thermals? The only reason that came to my mind was an increased swichting frequency of the VRMs due to low voltage that could have caused that issue.

Any ideas, comments?

Cheers,
Chris

Hardware: i7 2600K, Bequiet Shadow Rock TopFlow SR1, P8Z68-V Pro Gen3, 4x2GB Kingston Value RAM 1333MHz, Corsair TX850W, Samsung 840 EVO

 

[EarthDog]

I merged your double post which also took out your signature in the second post... Can you list your hardware again please? My apologies.

 

[cubusmedusa]

It wont appear unless I post a new message. Take this as a reply to your post. ;-)

 

[EarthDog]

I know, that is why I asked you to list it...

Anyway, that is an odd issue. They shouldn't be hot to the touch in most any situation, particularly at stock speeds, none the less underclocked. Perhaps it is because it was left on auto? Maybe set voltages manually.

Also, unless you are using the PC 24/7, you really are saving pennies per month with your underclocking and undervolting. Just enable power savings in windows and let it handle those duties. You also bought a quad with HT and disable it... makes no sense. Enable it and use it! After all, you paid for it.

 

[cubusmedusa]

You missunderstood. When everything is on auto (only clockspeed set to 2400) temps are fine. When I start undervolting VRM temps explode. I suspect drastically increased VRM switching frequencies to maintain stability. Thats the only reasonable explanation for hot VRMs despite low clockspeed and low voltages (at least for my reasoning )
I dont do it to save money. As I said, just for fun. But I would have never expected this and never heard of it. I am sure my mobo is working as it should, BIOS is up to date. I suspect that nobody has ever thought of VRM thermals while undervolting or other mobos can handle low voltages without affecting VRM switching.

Edit:
without that observation but even more so with that observation; I agree that undervolting makes absolutely no sense

Edit 2:
I tested full manual voltage instead of offset. I had to set it to .95 (quick and dirty), not lower, otherwise BSOD. With offset I reached .90. But with full manual voltage, VRM temps seem to behave.

 

[EarthDog]

I do not have an answer as to why that would happen. I doubt it is because of switching frequencies though as there is no voltage change there so for it to get hot like that without voltage changes, I find difficult to believe. I also don't think it would compensate for voltage, but just for clockspeed and bclk... which are stock or underclocked in the fist place. It should LOWER the switching speed for lower clocks if you think about it...

 

[cubusmedusa]

If I enter voltage manually (no offset) then there is no voltage change. As I reported in Edit2 temps seem to do well. But if I do the offset-method voltages and frequencies change all the time between max clockspeed and x16 and minimum voltage that I reach at x16 (1600 MHz) is 0.9V and VRM temps go up.
I think you are right that switching should go down with clock but consider the 0.9V. That might make the difference. I thought of the clocked (and not constant) supply of power to the cpu. And I can only imagine 2 reasons for increased VRM temps. 1 Current and 2 Switching frequencies.
Anyhow, interesting observation and lesson learned that undervolting does not make too much sense.

 

[EarthDog]

I did consider the .9v. What I said was that it makes no sense with lower voltage that switching frequencies go up as you theorized. Typically things like that are clockspeed related. You lowered the current, and the switching frequencies shouldn't move, particularly with LOWERING things. A higher switching frequency requires HIGHER voltage to work properly.


Leave everything on auto and set windows power management to let it do the throttling. You(your VRMs) should be ok like that.

 

[trents]

Might be a bios issue such that the under volting value you enter actually increases the voltage.