Agate GPUs Overview
In its initial deployment, Agate has a total of 264 A100 GPUs. In addition to general GPU calculations, the A100 GPUs provide accelerated performance for Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and Molecular dynamics workloads. In addition to the A100 batch resources, Agate features 80 A40 GPU nodes for interactive GPU applications through command line or Jupyter notebooks.
AI, Machine Learning, and Deep Learning
- Deep Learning workloads
- General purpose GPU calculations
Interactive A40 GPU platform
- Jupyter Notebook
- Machine learning (Tensorflow, Pytorch)
- Molecular dynamics simulations (GROMACS, NAMD)
- Natural language processing (Tensorflow, Pytorch)
The A100 GPUs are an ideal resource for machine learning jobs (PyTorch, Tensorflow) as well as molecular dynamics simulations (GROMACS, NAMD). As with any parallel application, it is best to start with a single GPU and then see if you get substantial performance improvements when you add a second GPU.