Undervolting older Intel mobile CPU

[green.pixel]

It's T6600 in Lenovo G550. I've tried ThrottleStop, but it seems there is no FIVR option in the UI. Latest Intel XTU installation says not supported by the processor type, for older 2.1 version it's unsupported platform.

Is there any other way to do it or some of these older mobile platforms don't support undervolting at all?

 

[green.pixel]

No one knows folks?

 

[orangpelupa]

how about manually diving into the bios?

uncfortunately it doesnt have modded bios https://www.bios-mods.com/forum/Thread-Lenovo-G550-BIOS-unlocking-request

 

[green.pixel]

Yeah, there is nothing you can do with about UV in the BIOS.

 

[Davros]

got an idea :
Control Panel -- Power Options -- Edit Plan Settings -- Change Advanced Power Settings -- Processor Power Management -- Maximum Processor State --
Setting -- ???%

 

[green.pixel]

I don't think that option reduces voltages, only clocks.

 

[Davros]

Yes but i'm assuming you want to reduce voltage to reduce heat or power draw
reducing clocks will achieve that and while maybe not changing voltage it will reduce amps and watts

 

[green.pixel]

I tried those Windows power setttings and they don't seem to work like they should. Whether it's 80 or 100% maximum processor state, the end result is the same, it doesn't lower the clock from default. Same for minimum, 1 or 10%, it doesn't go below 1,2GHz at which it is always in idle.

Then I trued adjusting Voltage ID and Set Multiplier in ThrottleStop, changed to 9x200Mhz and VID to min. available of 0,925V or -225mV, and it works. Temperatures dropped quite a bit and doesn't seem to throttle anymore. What I found interesting is that it is stable at those settings. Run Prime95 tests in Blend and Smallest FFTs variants for about an hour, and they didn't crash. It seems to me that either TS doesn't represent true voltage values or it really is overvolted at stock. Or I didn't test for long enough to show any instability.

 

[Zaphod]

No one knows folks?

It used to be possible to undervolt Core 2 Duo in Throttlestop, but not with the FIVR options. It uses (way) older tech so each voltage and multiplier combination needed to be set and tested for stability individually. As far as I can recall, it didn't play too nicely with SpeedStep/dynamic frequency either.

For notebooks I used RMClock in those days, it was rather easy to set up your own PID/VID table, but I don't know if it works with newer versions of Windows.

Edit: Penryn may be too new for RMClock, regardless. Given how slow it is, by today's standards, your best bet may be to try a fixed clock undervolt in Throttlestop, using the high performance power profile, and rely on the (then new) enhanced sleeps states for idle power saving.

Edit 2: Nevermind. I see that you got the manual PID/VID thing to work already.

 

[green.pixel]

RMClock shows higher VID and temperature compared to ThrottleStop and CoreTemp. Are the latter ones more accurate?

SpeedStep box is grayed out and the TS guide says that means the BIOS has locked this feature so it can not be turned on and off which is common for most laptops.

 

[Albuquerque]

Zaphod beat me to it: RMclock is likely the best answer here. Waaaaaay back in the day, I used it on several laptops with great success. Given the age of the hardware, I'd generally say RMClock is providing the more accurate readings as it was specifically built for CPUs in that particular era.

I think I still have a T7600 powered Dell E1505 somewhere around my home office. I wonder whatever happened to that thing? Now I wanna go dig it out "just to see"

 

[Zaphod]

RMClock shows higher VID and temperature compared to ThrottleStop and CoreTemp. Are the latter ones more accurate?

Check if RMClock correctly identifies the CPU. It stopped updating around that time, so It may think the CPU is a 65nm Merom Core 2 (which would explain the higher voltages).

 

[green.pixel]

Yeah you are right, it thinks it's just Genuine Intel(R) Core(TM) 2 Duo and Unknown Core. The last RMClock version is from february 2008, and this CPU is from 2009, so probably that's why.

And I was wrong above, it shows lower T compared to CoreTemp and TS which are nearly equal in readings.

 

[Davros]

how about cpu-z from 2013 (to identify it)
http://download.cpuid.com/cpu-z/cpu-z_1.63-32bits-en.zip

 

[green.pixel]

It is known, T6600.

https://ark.intel.com/content/www/u...ssor-t6600-2m-cache-2-20-ghz-800-mhz-fsb.html