Hacker News Reader

Comment by seanalltogether | Hacker News Reader

> Test Type: Idle desktop only (no applications or media playback, unless otherwise stated)

It's weird they didn't also include a simple web browser test that navigates a set of web links and scrolls the window occasionally. Just something very light at least, doesn't even have to be heavy like video playback.

View on HN

Not that weird. Idle desktop isolates the effects of the change to get a worst case scenario. Would be interesting to see a light activity test too though - see if you still get a noticeable difference.

Yeah this is not meaningful due to the unrealistic workload. Sad thing is, I bet a web browser test would still show the difference, as long as a page is kept static on the screen for more than a few seconds before moving on.

Power consumption is incredibly difficult to benchmark in a meaningful way because it is extremely dependent on all the devices in the system, all the software running, and most power optimizations are workload dependent. Tons of time went into this in the windows fundamentals team at Microsoft.

> We’re currently running the same test again on all three laptops to account for variance, but this time with a video playing to simulate a more active usage scenario. Once those results are in, we’ll update the relevant section with the new data.

What? "We are doing a second test to account for variance, but also changing the test setup" that doesn't make any sense.

They account for variance between different laptops. The test is changing the same way for all laptops. It makes sense.

Does that imply they didn't run the same test on the three different laptops before? Why are they presented on the same graph then?

No. Try parsing it like this:

>We’re currently running the same test again (on all three laptops to account for variance), _but this time with a video playing to simulate a more active usage scenario_

I think you'd normally include a comma before "on all" then. I'm not convinced that's what the author meant, but at least it makes more sense than what I read it as. Thanks for guiding me through that.

There's more discussion about it on the WAN show https://www.youtube.com/live/tkYiqvA7pmU around 33:42. They're competent people, so if some interpretation doesn't make sense, maybe it's not a good interpretation.

Would it not be a much better test to measure power draw for say 10 min with the seconds on and then off, for 10 cycles or so? Rather than waiting for the battery to run down. I guess it's more difficult for a laptop as your have to measure in between battery and laptop, but it should be easy for a desktop (maybe one with laptop CPU).

I agree. My guess is the way this may be implemented could keep the system from entering a lower energy state in some way or another, something which would be far less noticeable during normal usage.

Reply on news.ycombinator.com