Defrag Tools #181 - System Power Report

Sign in to queue

Description

In this episode of Defrag Tools, Chad Beeder and Andrew Richards are joined by Paresh Maisuria from the Windows Kernel Power team and Zach Holmes from the Fundamentals team to talk about System Power Report, a new feature in Windows 10 Creators Update.

Related links:
Defrag Tools #168 - Powercfg Sleep Study (older version of this tool)
Defrag Tools #157 - Energy Estimation Engine (E3) (the framework used for estimating power usage)

Timeline:

[00:00] Welcome and introductions
[00:30] This is an updated and expanded version of a feature previously called Sleep Study. Now it covers everything related to power, not just details of modern standby states.
[02:55] You can still run it with powercfg /sleepstudy (for backwards compatibility) - but the new command is powercfg /systempowerreport, or powercfg /spr
[04:08] Opening up the generated report - lots more data than in the old Sleep Study report.
[05:32] Looking an active session: How much battery power was used, and by what? What was the screen brightness? Which apps used the most power?
[09:40] Why some power usage gets attributed to "Unknown"
[15:00] Unlike the old Sleep Study report, the System Power Report even gives useful info on traditional standby (S3) systems.
[16:40] Looking at a standby session: You can tell why a system went into standby, and why it woke up. Also lots of other stats, like how long it took to hibernate, etc.
[20:27] The report also contains an "expert tab" which contains data about the battery design capacity, current capacity, and health
[23:18] Bugchecks are also logged in the report (including the parameters).
[24:35] Still has all the details on a modern standby system (like in the old Sleep Study report). But enhanced. Now we have better instrumentation to track why a system got woken from standby.
[27:58] Of interest to OEMs and hardware engineers: We track power usage data for the SoC (System on a Chip) subsystems. Can give you the first indication of where to look further if power usage is too high.

Email us at defragtools@microsoft.com

Embed

Download

The Discussion

  • User profile image
    Zach Holmes

    Hello everyone,

    I wanted to clarify something I said at 9:15 - I said yellow implies a higher "brightness", but what I meant to say was a higher "power drain" and in that example I was also specifically referring to OLED panels because the R and G channel pixels both need to be lit. On LCD screens the color is more a function of the filter over the backlight so brightness carries the majority of the power impact, but in OLED, the number and intensity of each pixel required is a large influencer on power drain. Sorry for the confusing statement in the video.

  • User profile image
    Vinay

    Awesome work, Paresh! System Power report is delivering diagnostics to end user and with great value.

  • User profile image
    OzBobWa

    Time for an Edge Power Consumption episode?

    Considering: 

    https://channel9.msdn.com/events/WinHEC/WinHEC-Taipei-June-2017-Workshop/Designing-and-Optimizing-for-Long-Battery-Life-and-Responsive-Windows-Devices (WinHEC 2017 workshop)

    Microsoft's test: https://channel9.msdn.com/Blogs/CANITPRO/Microsoft-Edge-Experiment-Battery-Life (Canada IT Pros Jonathan Rozenblit, Anthony Bartolo)

    and public attempt: https://www.youtube.com/watch?v=q0112lYQBPE (Linus TechTips)

     

    What do you guys say?

  • User profile image
    ChadBeeder

    @OzBobWa: I don't know if we want to delve into that argument, it's not really within the scope of our show. But I would mention a few things:

    * After doing the battery rundown tests, in addition to just seeing how long it ran until the battery died, it would be more scientific to look at the System Power Report to make sure there weren't other background processes contributing too much to the battery usage.

    * To get even more scientifically accurate results from the report, it'd be best to use systems which have integrated power metering chips (we talked about this in more detail in our earlier episode on Energy Estimation Engine). Unfortunately, most devices on the market do not have these.

    * As there is video playback involved, the results may vary significantly across different hardware because hardware offload for GPU and audio may be a factor.

    * As mentioned in this episode, screen brightness is a big factor. I don't know if these other people who did these tests were forcing a consistent screen brightness across all of their tests. If not, their results could be misleading.

Add Your 2 Cents