FastTreeRegression algorithm is not working at full speed

OS & Version: Windows 10 and Linux
ML.NET Version: ML.NET v2.0.1 & AutoML v0.20.1
.NET Version: .NET 7.0

Hi,
The FastTreeRegression algorithm does not use available computing power.
On my VM, with 24 cores, it uses about 10% of the available CPU (distributed among different cores).
On my local PC (4 cores) it uses about 50%, which is also too low.

Is that why this is happening? The program does not use HDD/SSD, so the writing and reading speed is not a limitation. Does the RAM speed block the full CPU speed?

LightGbmRegression works well (80-90% CPU usage on both setups).