Power Saver using more power than High Performance

[gijoe50000]

So I recently noticed when messing around with the power profiles on my PC, in Windows 10, that the battery saver profile is using more cpu power than the high performance option.

On my 4790k the battery saver profile is always at 800mhz and the high profile is always at 4600mhz.

Now there's not MUCH difference, but the battery saver profile is using about 5W more than high performance. This doesn't make sense to me. I added a screenshot of an OpenHardwareMonitor plot to show it. In the screenshot the two black 'arrows' show roughly where the battery saver profile was active.

http://i.imgur.com/HPNXuLP.jpg

Is this normal?

Originally I had thought to set the profile to battery saver if I was leaving the PC turned on while I was away, if I was downloading something or whatever, but there doesn't seem to be much point.

I have most of the power options in the bios on auto C states etc, I'm thinking they must be better at power optimization than windows is, but it doesn't explain why the battery saver is using more power. It's weird.

Anyone else come across this? Is there something I'm missing here? Like some reason the cpu needs more power to be at 800mhz than at 4600mhz?

I know the Power Options in windows was designed mainly for people with laptops, who want to dim the screen and shut down hard drives after a certain amount of time etc..

 

[Ecky]

OpenHardwareMonitor will not be able to give accurate power readings; it estimates based on some sensors, but it's not a real power draw figure. If you were to use something like a Kill-a-watt to measure, you'd see it's wrong. That said, if your PC is idle and using the "balanced" profile, power usage should be within margin of error of running the power saving profile. I'd leave it on "balanced".

 

[gijoe50000]

I don't have a Kill-a-watt so I can't measure it. But even if OHM is not totally accurate, I find it strange that High Performance always seems to run at a lower cpu wattage. Margin of error doesn't seem to be an issue as the change in watts is consistent and repeatable.
Here's another screenshot with the cpu clock speeds at PS, HP and then balanced. It seems the cpu voltage is higher on when power saving is active.

http://i.imgur.com/f5q3p4R.png

 

[Ecky]

The voltage table is built into the CPU, and so shouldn't change based on Windows power profile. I'm leaning toward OHM giving a bunch of nonsense readings.

EDIT: I have a kill-a-watt, and can measure my system if you'd like. I'll also check on voltages.

 

[gijoe50000]

Thanks for the reply.

The voltage table is built into the CPU, and so shouldn't change based on Windows power profile. I'm leaning toward OHM giving a bunch of nonsense readings.

That might explain it, if the cpu doesn't 'know' that Windows has lowered the clock speed and it's supplying the same voltage regardless. I suppose that WOULD save power since it's just restricting the clock speed and lowering overall performance.

EDIT: I have a kill-a-watt, and can measure my system if you'd like. I'll also check on voltages.

It would be interesting to see the overall wattage difference alright..

 

[Ecky]

With my chip, power consumption measured at the wall:

"Balanced"
Idle: ~40.6w (1600mhz)
Power as estimated by CoreTemp: ~10.7w
Prime95: ~102w (3800mhz)
Power as estimated by CoreTemp: ~67.5w
Estimated power difference between idle and load: ~56.8w
Difference at the wall between idle and load: ~61.4w


"Power saver"
Idle: ~40.6w (1600mhz)
Power as estimated by CoreTemp: ~10.7w
Prime95: ~102w (3800mhz)
Power as estimated by CoreTemp: ~67.5w
Estimated power difference between idle and load: ~56.8w
Difference at the wall between idle and load: ~61.4w


"Power saver" w/maximum processor state at 5%:
Idle at the wall: ~40.6w (1600mhz)
Power as estimated by CoreTemp: ~10.7w
Prime95: ~52.6w (1600mhz)
Power as estimated by CoreTemp: ~22.4w
Estimated power difference between idle and load: ~11.7w
Difference at the wall between idle and load: ~12.0w


CoreTemp's estimated power is actually pretty close, it seems. When I checked a few years back, the numbers were wildly off. If we assume my power supply is ~90% efficient (which it should be with a platinum rating), 11.7w on 12v would be ~13w at the wall, and 56.8w on 12v would be ~63w at the wall.

 

[gijoe50000]

That's interesting..
By the looks of it, the power saving and balanced options are basically useless from a cpu point of view, since anything you want to accomplish at a lower clock speed will just take longer anyway. They might be handy if you want to customise your computer to sleep after a certain amount of time etc..

I'm assuming the bios takes care of all the cpu power savings, and the higher cpu wattage I noticed earlier was just windows running tasks to keep the clock speed down..

Thanks for running those experiments, it's always good when things start to make sense.. 🙂

 

[Ecky]

I actually have a lot of data, if you're interested.

As frequency rises, voltage needs grow:

As voltage and frequency rise, power usage goes up exponentially:

If you divide frequency by power usage, you can see that a CPU's efficiency is highest at its lowest frequency, because power usage drops faster than performance does:

However, if you take total system power consumption into account, a CPU is actually most efficient at roughly stock speeds for most of Intel's desktop CPUs:

The conclusions you can draw from this are:

1) If you load your CPU at 100% all the time, it's more efficient to leave it at stock speed than to underclock or overclock.

2) If you do work with your CPU and then immediately afterward, put your PC to sleep, it's more efficient to leave it at stock speed than to underclock or overclock.

3) If your PC is on all the time, and you don't really mind it being a little lower, you'll save power by keeping the frequency down.