What type of control is used to manage the different clock configurations (nvpmodel)?


The TX2 has many clock configurations: MAXN, MAXP, MAXQ, the Jetson_clocks. I am wondering what type of algorithm/control is used to manage frequency?

Is it reactive or proactive?
Reactive: when a trip point is reached, the frequency will diminish.
Proactive: it tries to predict a future temperature or workload and according to that diminish or augment frequency.

Thank you

hello Aizzaac,

you may refer to developer guide,
please check Supported Modes and Power Efficiency for NVPModel clock configurations.
please refer to Maximizing Jetson TX2 Performance for the details of Jetson_clcocks.

in short,
nvpmodel has configuration to set the maximum clock frequency of each power modes. you may have variable clock rates; Jetson_clocks force the clock frequency to current support maximum frequency, you won’t see variable clock rates.
there’s tegrastats utility you may used for checking memory usage and processor usage.

okey, about the variable clock rates what determines it has to change or not? what type of control there is? is it a PID control? like the one that comes within the linux kernel? or something else?

hello Aizzaac,

there’s tegrastats utility you may used for monitor memory usage and processor usage.