The new 128-node Cabeus GPU system is now available for production use. Each Cabeus GPU node has four NVIDIA A100 GPUs, for a total of 512 virtual nodes (vnodes). You can use the new nodes by accessing a Cabeus front end (CFE) system, and specifying the following model type and queue name in your PBS job request:
For more information about using Cabeus, see the following Knowledge Base articles:
We encourage your feedback on providing performance results and any questions or issues you have when using these nodes. Please contact the Control Room at (800) 331-8737, (650) 604-4444, or support@nas.nasa.gov.