CPU Jobs for Data Analysis and Processing¶
The cpu partition provides high-memory nodes for data analysis, preprocessing, and workflow orchestration. These nodes do not have GPUs — use them for work that benefits from many CPU cores and large memory rather than GPU acceleration.
CPU Jobs on AICR
AICR is not meant for CPU workloads. A small CPU-only partition is provided as a convenience for small data analysis tasks related to your GPU jobs. Heavy CPU workloads should be done on your home institution's cluster.
When to Use the CPU Partition¶
- Data preprocessing: cleaning, transforming, or merging large datasets before GPU training
- Post-processing: aggregating results, generating reports, statistical analysis
- Workflow orchestration: coordinating multi-step pipelines that launch GPU jobs
For GPU-accelerated work (training, inference, anything using CUDA), use the rtx-* or b200-* partitions instead. See GPU Jobs for more information.
CPU Partition Specs¶
| Property | Value |
|---|---|
| Nodes | AMD EPYC 9745 (5 nodes) |
| Cores per node | 128 |
| Memory per node | 1 TB |
| GPUs | None |
| Max wall time | 24 hours |
| Default time | 15 minutes |
Example: Basic CPU Job¶
The following is an example job script for a CPU job with 16 cores and 32GB of RAM.
#!/bin/bash
#SBATCH --job-name=data_analysis
#SBATCH --partition=cpu
#SBATCH --nodes=1
#SBATCH --cpus-per-task=16
#SBATCH --mem=32G
#SBATCH --time=04:00:00
#SBATCH --output=%x-%j.out
module load miniforge3
python analyze_data.py
See Also¶
- Slurm Basics — general job submission
- GPU Jobs - running GPU jobs