How to reduce the CPU usage of CVX Mosek without affecting problem solving ?

How to reduce the CPU usage of CVX Mosek without affecting problem solving ?
When I use parfor for parallel computing, the Mosek solver consumes a lot of CPU and the CPU temperature is too high.
Will setting parameter ‘MSK_IPAR_NUM_THREADS’ affect problem solving?

By default MOSEK will automatically select the number of threads.
However, the maximum number of threads allowed can be changed by setting the parameter ‘MSK_IPAR_NUM_THREADS’ and related parameters. This should never exceed the number of cores.
(link: 9 Technical guidelines — MOSEK Rmosek package 10.2.6)

Since you are solving multiple problems in parallel, then I would force Mosek to solve each problem using 1 thread. Hence, I would use

MSK_IPAR_NUM_THREADS = 1

If you are extremely performance-sensitive, then MATLAB+cvx may not be the best tool for solving many problems in parallel.

Thank you! I have tried it. This method reduces CPU utilization to a certain extent.
The CPU utilization is too high, and the CPU temperature will also be very high. I am worried that it may damage the CPU.

Each Mosek needs at least 1 thread. Maybe the parfor use too many threads.

I have never used parfor though.

Based on your suggestion, I also learned about the method of setting the number of MOSEK threads from the link ( 9 Technical guidelines — MOSEK Rmosek package 10.2.6).
Due to the need for a large number of for-loops in my code, I used parfor for acceleration. In order to prevent CPU utilization and high CPU temperature, I only used a small number of CPU computing cores.

Btw the natural Mosek reference given you use MATLAB is

https://docs.mosek.com/latest/toolbox/index.html

When I use MSK_IPAR_NUM_THREADS = 1, the following prompt ‘NOTE: custom settings have been set for this solver.’ appears in the command line window. Can this prompt be canceled?

Remove a few fprintf’s in lib/cvx_run_solver.m might be the only way CVX/lib/cvx_run_solver.m at a0a775908bc1eec1625d56498d0224132b0d8b7f · cvxr/CVX · GitHub