I have mine (i3) set to run at 570mhz /20% screen brightness on battery and 1.5ghz, 90% screen when plugged in. I think I have experienced a record battery life for me the last 2 days. I dont know for sure, but definitely over 10 hours, maybe over 11. Is there a program to keep track of the time I'm not in sleep mode?
ProcThrottleMax, Powercfg setting
- Thread starter GreyFox7
- Start date
Iluminatus23
New Member
How did you limit your GPU?
Are you asking me? If so, I didn't but will be doing so in the future. To change GPU to max battery, type
powercfg -query
In the long list of junk, look for Intel HD Graphics and record the subgroup GUID (#1) and power management GUID (#2).
Now type
powercfg -setdcvalueindex SCHEME_CURRENT (#1) (#2) 0
powercfg -setactive SCHEME_CURRENT
By default, for some reason, when on AC the GPU is not set to maximum performance. You can fix that with
powercfg -setacvalueindex SCHEME_CURRENT (#1) (#2) 2
powercfg -setactive SCHEME_CURRENT
powercfg -setactive SCHEME_CURRENT
You can apply many changes at once to power configuration, and dont need the -setactive each time. Just make sure to type it before you exit command prompt.
I am conducting a battery trial at the moment. I will upload it and a complete log when I am finished. It may take a day or so, but I am logging my activities during the trial including exact times of sleep/wake should that happen. I am running at min frequency, I think 25% screen brightness and max power saving on the GPU
Update: At this point I have been unable to affect change in any way on the operation of the Graphics component, not even with XTU. All settings are completely ignored. However the very thing that needs to be controlled is in fact the method used once some period of Thermal Throttling is exceeded. I call this Panic Mode or Turtle mode during which the Graphics/CPU Frequency comes to the rescue albeit a little late.
Regardless of any settings through Powercfg setting the Graphics Power Plan or through XTU limiting the Graphics Frequency via the ratio the Graphics runs at FULL speed until the Graphics Frequency is mysteriously dropped after a period of Thermal Throttling. As soon as the Frequency drops from 989 mhz to 748 mhz the CPU/GPU temperatures drop and Thermal Throttling ends. It also appears that at that point everything is fine even though it continues to further constrain the Graphics Frequency.
The same thing actually happens with the CPU when running the CPU stress test however, through ProcThrottleMax you can control it and completely eliminate Thermal Throttling and the panic constraint. Unfortunately settings above 74 don't work and the CPU temp at 74 runs at about 69C so there is plenty of head room to go higher.
Note: in this graph Power Throttling (pink line) is abnormally high because I was attempting to use that to limit the Graphics to no avail. Note also that when Thermal Throttling is on power throttling is suspended.
The Frequency control is key and clearly the capability is there but its not available to be used prior to panic mode setting in.
There are two deficiencies IMO that need to be addressed.
1. The inability to set a cap on the Graphics Frequency - If there was a setting that allowed setting Max Graphics Frequency to 800-900 MHz the temperature could be controlled.
2. The odd limit of 74% on ProcThrottleMax setting above which it has absolutely NO affect. - ProcThrottleMax should be extended to work up to 85-90%. As long as the temps remain below 89/90C you will not see Thermal Throttling kick in.
I believe the combination of these two would allow the SP3 to operate in the 85% of max range on a sustained basis without being inconsistent and panic constrained. Which except for on benchmarks and possibly a little lower frame rates on some games would be largely unnoticeable. My rule of thumb is, it takes about a 20% difference to be user perceptible.
Regardless of any settings through Powercfg setting the Graphics Power Plan or through XTU limiting the Graphics Frequency via the ratio the Graphics runs at FULL speed until the Graphics Frequency is mysteriously dropped after a period of Thermal Throttling. As soon as the Frequency drops from 989 mhz to 748 mhz the CPU/GPU temperatures drop and Thermal Throttling ends. It also appears that at that point everything is fine even though it continues to further constrain the Graphics Frequency.
The same thing actually happens with the CPU when running the CPU stress test however, through ProcThrottleMax you can control it and completely eliminate Thermal Throttling and the panic constraint. Unfortunately settings above 74 don't work and the CPU temp at 74 runs at about 69C so there is plenty of head room to go higher.
Note: in this graph Power Throttling (pink line) is abnormally high because I was attempting to use that to limit the Graphics to no avail. Note also that when Thermal Throttling is on power throttling is suspended.
The Frequency control is key and clearly the capability is there but its not available to be used prior to panic mode setting in.
There are two deficiencies IMO that need to be addressed.
1. The inability to set a cap on the Graphics Frequency - If there was a setting that allowed setting Max Graphics Frequency to 800-900 MHz the temperature could be controlled.
2. The odd limit of 74% on ProcThrottleMax setting above which it has absolutely NO affect. - ProcThrottleMax should be extended to work up to 85-90%. As long as the temps remain below 89/90C you will not see Thermal Throttling kick in.
I believe the combination of these two would allow the SP3 to operate in the 85% of max range on a sustained basis without being inconsistent and panic constrained. Which except for on benchmarks and possibly a little lower frame rates on some games would be largely unnoticeable. My rule of thumb is, it takes about a 20% difference to be user perceptible.
although you can make these changes and see them in the Power Scheme. I see no affect being applied and the stress test graphs appear unchanged.
After I finish with the battery test, I will take a look. I am unable to run a 1080p youtube video smoothly with GPU max battery applied and min proc frequency. I will to see what differences there are, if any, should I change it to max performance
malberttoo
Well-Known Member
This Intel HD Graphics has a mind of its own I tell you... now its the number one (mainly only) offender on my Sleepreport. So not only can't I control it when its supposed to be running, its doing its own thing during sleep. I'm thinking I may restore my image from last Tuesday before things went sideways.
Cothek
Active Member
GreyFox, it's good to see that someone else is taking an active role in trying to reduce throttling on the SP3. I really want to keep this thing but the level of throttling makes it hard for me.
I actually started a thread similar to this on another forum and have been trying to reduce throttling with active cooling.
Trying to mitigate throttling with powercfg was my next step.
I will be experimenting with this and will post here.
I actually started a thread similar to this on another forum and have been trying to reduce throttling with active cooling.
Trying to mitigate throttling with powercfg was my next step.
I will be experimenting with this and will post here.
To be sure this is still throttling but will result in a stable consistent performance vs bursty choppy performance (alternating hare & tortoise). with the caveats listed above we cannot currently get there unless someone finds a solution that allows controlling the graphics clock frequency. I'm not sure if anyone from MS has seen this so I was considering submitting a write up to them.
If its all configurable through the Power Plan or Powercfg then users can choose, Bursty or Stable er Balanced or Sustained & Cool or perhaps All Day & Night Battery Life.
Cothek
Active Member
So I did a little experimenting with the throttle max and min and found something interesting.
I've been running siege battles in Mount and Blade: Warband as my "stress" test and here is a picture of the XTU graphs for one battle (game starts when processor and temp spike). Notice how quickly the processor ramps down to .8GHz - .9GHz (orange line) as well as the GPU freq:
Next I set the MIN throttle to %100, which doesn't really work but where the CPU would idle around .8GHz - 1GHz it idles around 1.7GHz - 2GHz. Then I ran a battle, and then another and another. This was after multiple battles:
The processor almost never dropped below 1.6GHz and the GPU was pretty pegged at 600MHz. The drop at the end was after the battle had ended and not sure what caused it. I tried several more battles to make sure this wasn't a fluke:
I'm going to try some longer tests but this is very interesting!
I've been running siege battles in Mount and Blade: Warband as my "stress" test and here is a picture of the XTU graphs for one battle (game starts when processor and temp spike). Notice how quickly the processor ramps down to .8GHz - .9GHz (orange line) as well as the GPU freq:
Next I set the MIN throttle to %100, which doesn't really work but where the CPU would idle around .8GHz - 1GHz it idles around 1.7GHz - 2GHz. Then I ran a battle, and then another and another. This was after multiple battles:
The processor almost never dropped below 1.6GHz and the GPU was pretty pegged at 600MHz. The drop at the end was after the battle had ended and not sure what caused it. I tried several more battles to make sure this wasn't a fluke:
I'm going to try some longer tests but this is very interesting!
