HPC4 - Tips for selecting CPU/GPU models

Select CPU types for a job

Select Intel CPU types in srun

$ srun --partition=intel --account=your_PI_account_name --time=HH:MM:SS --pty bash

Select AMD CPU types in srun

$ srun --partition=amd --account=your_PI_account_name --time=HH:MM:SS --pty bash

Select Intel CPU types in sbatch

#SBATCH --partition=intel                   # Select intel CPU architect for the task

Select AMD CPU types in sbatch

#SBATCH --partition=amd                    # Select amd CPU architect for the task

 

Select GPU types for a job

Select A30 GPU types in srun

$ srun --partition=gpu-a30 --gres=gpu:1 --account=your_PI_account_name --time=HH:MM:SS --pty bash

Select L20 GPU types in srun

$ srun --partition=gpu-l20 --gres=gpu:1 --account=your_PI_account_name --time=HH:MM:SS --pty bash

Select A30 GPU types in sbatch

#SBATCH --partition=gpu-a30              # Types a30 and Number of GPUs for the task
#SBATCH --gres=gpu:1   

Select L20 GPU types in sbatch

#SBATCH --partition=gpu-l20              # Types l20 and Number of GPUs for the task
#SBATCH --gres=gpu:1               

 

Quality of Service (QoS) for a job

Select either one partition using srun for interactive mode

$ srun --partition=amd --account=your_PI_account_name --qos=debug --pty bash

Select either one partition using sbatch for batch mode

#SBATCH --qos=debug     

 

Learn More

Check out Slurm Partitions, QoS and Resource Limits for available resources could be used.