I/O peripherals CPU power and battery notification icons

[-batab-]

Hello there,

1) I already murdered my CPU via ThrottleStop and Power Plan Explorer settings (12800HX and only 11W full load). However, processing any I/O output peripherals still requires a huge amount of power: I'm idling at 1-2W package power but moving the touchpad brings it to 6W.
As this seems a lot of power just to process I/O inputs, and since you implemented a lot of fixes during the years, do you think a fix for this would exist? Cutting another 2-3W of total power when doing super light stuff on battery would be awesome.

FYI: the only way I found to really limit this behavior is murdering the CPU even more by reducing the "Maximum Performance State" even more but this can get as bad as making the system almost unusable so this seems a little to much brute force for a very specific target.


2) Is it somehow possible to pull out a discharge power icon? Sometimes I forget I don't have my dGPU sleeping and having the discharge icon in the taskbar would be very handy.

 

[Assimilator]

Why do you care so much about 6 measly watts?

 

[-batab-]

Why do you care so much about 6 measly watts?

Well,
1) as with overclocking it's just fun;
2) it's optimization so it's fun again;
3) having 3-5W just for I/O processing on a power profile that uses 1-2W when idle and 11W when fully loaded are basically 30-50% of the whole power range. Battery life is definitely very sensitive to such power draw.

Plus I was just curious since I never dug into this stuff and looked weird.

 

[bug]

Why do you care so much about 6 measly watts?

I would argue not even 1W is "measly" when it comes to mobiles

 

[unclewebb]

I'm idling at 1-2W package power but moving the touchpad brings it to 6W.

This is a bug in the touchpad driver. Instead of sampling the touchpad a 1000 times per second, it is probably in an endless loop sampling the touchpad a 1,000,000 or a 100,000,000 times per second. My 4th Gen laptop had a similar bug. Just resting my big thumb on the touchpad and not even moving it sent the C0% straight up on the first thread of the first core.

This is a sign of old school programming from the era when a CPU only had a single core and no hyper threading. It took a lot of testing of different driver versions before I found one that was not quite as bad. Not great, just not quite as bad. Using an external mouse was the only real solution.

I think the core of some Windows drivers has not been updated in 25+ years. If it still sort of works then they just keep using it. No one seems to realize that a CPU can run 100 times faster compared to the good old days and there are also many cores and threads available to run code on. No need to lock any software to a single thread. The touchpad driver programmers need to have a closer look at the inefficient code they are using.

I would argue not even 1W is "measly" when it comes to mobiles

I totally agree. I think one of the reasons that Apple laptops often times have better battery run time compared to a similar Windows laptop is because of poorly written Windows drivers. A small difference in power can make a big difference in how long your laptop runs when unplugged.

 

[-batab-]

I would argue not even 1W is "measly" when it comes to mobiles

Actually, the same thing happens with mouse and keyboard. Pretty much any input device I could try. And it seems like it only happens when moving, if I rest on the touchpad it stays basically idle.

It's definitely something about polling or stuff, I was just wondering if there is something that can be done to limit such bad driver/Windows behavior. (Like for example what you implemented for stuff like Windows Defender boost)

 

[unclewebb]

I was just wondering if there is something that can be done

I do not have any magic tricks available to fix poorly written drivers. Maybe Microsoft could hire me and I could show them why sampling something 100,000,000 times per second in 2023 is a bad idea.

 

[thewan]

If windows were that efficient, ppl would be worshiping Microsoft instead of apple.
get a mac.

 

[-batab-]

I do not have any magic tricks available to fix poorly written drivers. Maybe Microsoft could hire me and I could show them why sampling something 100,000,000 times per second in 2023 is a bad idea.

Fair enough. To be honest you should have also been hired ages ago by Intel though. Having good hardware/software out of the box would be a good option.

If windows were that efficient, ppl would be worshiping Microsoft instead of apple.
get a mac.

Nah. As good as they are I just love tinkering with stuff. Plus I also need a lot of multicore power for simulations and some softwares don't even run on mac OS. That's not really an option.

 

[bug]

Actually, the same thing happens with mouse and keyboard. Pretty much any input device I could try. And it seems like it only happens when moving, if I rest on the touchpad it stays basically idle.

It's definitely something about polling or stuff, I was just wondering if there is something that can be done to limit such bad driver/Windows behavior. (Like for example what you implemented for stuff like Windows Defender boost)

If you're so inclined, you can do a comparison with Linux. Boot from a live media, install powertop and play with it a little. If you see similar problems with both OSes, it could be something fixable (or not) in the BIOS, instead of an OS-specific driver.