Cool. Thanks for the info.

I decided to pop the case open and peek inside and it doesn't look like PWM control is an option. If I had to guess, the miner uses thermostatic control to prevent overheating. The fan header being used is a 3-pin so I think the only way to gain control with the same kind of granularity as PWM would be to build a voltage regulator for it. This would be more work than one miner is worth.

I was looking as both sides but like you said, the long term average is fine. I was just curious about what the behavior is indicative of. I'll try messing around with diff setting.