Offered by
National Center for Supercomputing Applications (NCSA)Access
How to gain access:
Requesting Allocations:
– Mainly used by campus researchers for AI model training
– Special request form (https://forms.illinois.edu/sec/6587313)
– Free to allocated users
Available to:
-
Faculty and Staff
-
Grad Students
-
Undergrads
Description
The purpose of HAL is to enable deep learning at scale.
Technical Specifications
– Hal
-
CPU:16IBMPower9nodes(2x20x16=640CPUCores)
-
GPU:4V100oneachnode;totalof64GPUs
-
Memory: 4TB aggregate, 256 GB/node
- Interconnect: HDRInfiniBand (200Gb/s)
- Storage options: 336TB NVMeflash
– Hal-DGX
• NvidiaDGXA100with8A100-80GPUs
Cost
Check the resource site.