By default _ride uses first available GPU which satisfies compute
compatibility >= 6.0
There is command line option 'gpuid=' for _ride script which
allows to specify a particular GPU
So, to run ride on multiple GPUs you can use it like below:
icm64 _ride tamplate1.mol database.molt result1.sdf gpuid=0
icm64 _ride tamplate2.mol database.molt result2.sdf gpuid=1
GPU starts from 0 and should correspond to the output of nvidia-smi.