Untitled topic • Hacker News Board

Re: undefined

#31

Post by SavantIdiot » Sun, Oct 03, 2021, 2:21 AM UTC

maleldil wrote: post #13

Is this the 'high quality' Apple fan boys like to talk about?

No, I think they're referring to the pre-installed ads on the start menu. Oh wait, that's Microsoft. Or maybe it is the utter failure of a control panel (Bluetooth, ACPI, Audio) that is still a crash-fest with Focal Fossa.

Let's face it, everything with a GUI now sucks. Hard.

Re: undefined

#32

Post by mthoms » Sun, Oct 03, 2021, 2:22 AM UTC

tmpz22 wrote: post #15

Tinfoil hat: they knew the M1 macs would be a big draw and got intentionally soft on optimizing 2019 intel models.

There's some truth to this but I don't think it was necessarily intentional. It must have been an "all hands on deck" situation getting M1/Rosetta ready.

I'm PO'ed either way. My 2019 16" has been plagued with this issue.

Re: undefined

#33

Post by dangrossman » Sun, Oct 03, 2021, 2:23 AM UTC

scottlamb wrote: post #29

I suspect there's another related reason for high kernel_task % CPU: your battery is very low and your charger is underpowered. It seems to throttle your CPU usage to avoid exhausting the battery while charging, and this throttling shows up in the same way as thermal throttling. I discovered this by accident when I plugged my MBP into my phone charger instead of my laptop charger. I need to mark the cables somehow, o…

I bought a bunch of 45W+ USB-C chargers and put one in every room of the house. They're only about $20 each on Amazon. All of the family computers and phones can be charged on any of them. It's nice not to think about it.

Re: undefined

#34

Post by saagarjha » Sun, Oct 03, 2021, 2:25 AM UTC

PSA, if you're doing investigations like this: there's two important parts to "proving" your hypothesis. One is you look at the experimental results and then ensure that matches your hypothesis, which is what this answer does here. The second is that you remove confounding variables and actually figure out what the real cause is. For an analysis like this, I find it quite disappointing that nobody actually went back to find the actual code that was the culprit and ensure that it was indeed what was causing the issue. Until then, you can show all the fancy graphs you want implying correlation, but you're really no better than "deleting Chrome makes my computer faster"–possibly correct, but you really haven't proved what you set out to.

Re: undefined

#35

Post by kitsunesoba » Sun, Oct 03, 2021, 2:25 AM UTC

walrus01 wrote: post #19

I also recently discovered that something like 12% of CPU on my 2020 macbook air is used whenever I'm in a room with anything greater than perfect silence, by 'coreaudiod', because it apparently is constantly listening to the microphone built in near the keyboard and spending cpu cycles trying to filter out background noise. Apparently running the signal desktop app in the background is enough to keep coreaudiod cons…

The coreaudiod issue might have to do with apps that use the mic "priming" it by keeping it active even when there's no call. Some chat apps do this to avoid the 1-2s delay associated with turning on the mic, so audio starts streaming as soon as the call starts.

Re: undefined

#36

Post by solarkraft » Sun, Oct 03, 2021, 2:28 AM UTC

This answer keeps clogging up my results researching my high kernel_task issues.

This answer being over enthusiastically given even after thermal throttling could reasonably be ruled out is one thing - but it seems like there’s also no real way to troubleshoot this type of issue (what sub-process is going wild?). Best I could find was a command to list the RAM usage of kernel modules.

Re: undefined

#37

Post by saagarjha » Sun, Oct 03, 2021, 2:28 AM UTC

walrus01 wrote: post #19

I also recently discovered that something like 12% of CPU on my 2020 macbook air is used whenever I'm in a room with anything greater than perfect silence, by 'coreaudiod', because it apparently is constantly listening to the microphone built in near the keyboard and spending cpu cycles trying to filter out background noise. Apparently running the signal desktop app in the background is enough to keep coreaudiod cons…

> I also recently discovered that something like 12% of CPU on my 2020 macbook air is used whenever I'm in a room with anything greater than perfect silence, by 'coreaudiod', because it apparently is constantly listening to the microphone built in near the keyboard and spending cpu cycles trying to filter out background noise.

Have you tried sampling the process to ensure that it is actually doing this? If it is, it's likely that Signal is leaving open a recording session and it shouldn't, and given that the app is open source, it should be fairly easy to find where it is doing this. I am sure they would be very interested in such a report, considering the privacy implications of leaving the audio system up when it shouldn't be.

Re: undefined

#38

Post by solarkraft » Sun, Oct 03, 2021, 2:31 AM UTC

scottlamb wrote: post #29

I suspect there's another related reason for high kernel_task % CPU: your battery is very low and your charger is underpowered. It seems to throttle your CPU usage to avoid exhausting the battery while charging, and this throttling shows up in the same way as thermal throttling. I discovered this by accident when I plugged my MBP into my phone charger instead of my laptop charger. I need to mark the cables somehow, o…

Thanks for adding this to my mental list.

The kernel has all of these weird functions/modes and I don’t see any list of which are possible and which one is active.

Re: undefined

#39

Post by interestica » Sun, Oct 03, 2021, 2:35 AM UTC

walrus01 wrote: post #19

I also recently discovered that something like 12% of CPU on my 2020 macbook air is used whenever I'm in a room with anything greater than perfect silence, by 'coreaudiod', because it apparently is constantly listening to the microphone built in near the keyboard and spending cpu cycles trying to filter out background noise. Apparently running the signal desktop app in the background is enough to keep coreaudiod cons…

The 2020 M1 version or previous Intel iteration (or perhaps there's no difference)?

Re: undefined

#40

Post by banjo_turtle » Sun, Oct 03, 2021, 2:36 AM UTC

sprite wrote: post #3

My 2019 16-inch MacBook Pro (2.3ghz i9/64gb ram) becomes unusable with 2 of the LG 5k monitors plugged in (one on each side). Worst MacBook I have ever owned.

I concur, I have a similar setup. Most expensive computer I've ever owned. It just throttles at the drop of a hat. I worked through it all with the apple techs, jumped through hoops for many hours and provided detailed diagnostics. They dismissed me with "throttling is normal, there's no problem".

In the real world, my 2017 MBP performed better just because it didn't throttle. Ridiculous.

As for the supposed 'Low power mode' in Monterey - I'm not convinced that isn't just turning on throttling manually.