3 ms·

This has been a known (albeit apparently not well known) issue/solution since the first USB-C macs were introduced. The cause is the placement of internals and

This has been a known (albeit apparently not well known) issue/solution since the first USB-C macs were introduced. The cause is the placement of internals and sensors and their proximity to the ports on each side. You can “prove” (demonstrate) it without opening the case by taking a stock MBP, putting it under medium load that’s enough to generate heat without other factors, then compare fan speed and performance with power on each side. They’re noticeably different without changing any other factor. And since the case itself is part of the thermal management design, you can feel relatively comparable heat generated by charging on both sides at the edge of the case above the keyboard. Obviously that’s not scientific, but it’s a pretty good clue there’s something else happening. Luckily, this is a known phenomenon and people have already done the legwork you’re asking for.

saagarjha 5y ago
See, that's exactly what I'm talking about. People have looked at the sensor placement and conjectured that kernel_task CPU usage is related to it doing some sort of throttling. It's a good hypothesis! But nobody has finished the job by going through a trace and actually checking that the kernel is doing that. That's the missing portion I'm asking for.
- Aeolun 5y ago
  I’m fairly certain I’ve seen an article somewhere with someone going through the trace and showing that kernel task is just spinning it’s wheels, generating artificial business to dump load.
- eyelidlessness 5y ago
  > conjectured that kernel_task CPU usage is related to it doing some sort of throttling Well it’s definitely doing some different thermal management depending on the port side you use. And performance definitely differs between those cases. Whether it’s technically throttling is kind of immaterial to whether this solution is meaningfully useful, in my opinion.
- bagels 5y ago
  Sure, the root cause hasn't been completely proven. I don't think the burden of that falls on mac users, but upon apple. Besides that, the investigation has produced some practical and useful advice.

saagarjha 5y ago

See, that's exactly what I'm talking about. People have looked at the sensor placement and conjectured that kernel_task CPU usage is related to it doing some sort of throttling. It's a good hypothesis! But nobody has finished the job by going through a trace and actually checking that the kernel is doing that. That's the missing portion I'm asking for.

Aeolun 5y ago
I’m fairly certain I’ve seen an article somewhere with someone going through the trace and showing that kernel task is just spinning it’s wheels, generating artificial business to dump load.
eyelidlessness 5y ago
> conjectured that kernel_task CPU usage is related to it doing some sort of throttling Well it’s definitely doing some different thermal management depending on the port side you use. And performance definitely differs between those cases. Whether it’s technically throttling is kind of immaterial to whether this solution is meaningfully useful, in my opinion.
bagels 5y ago
Sure, the root cause hasn't been completely proven. I don't think the burden of that falls on mac users, but upon apple. Besides that, the investigation has produced some practical and useful advice.