Early adoption provides researchers with several benefits, including:
The chance to get a jump start on using the new system with limited competition for resources.
Prioritized support from ARCC staff during EAP.
Exclusive reserved access to any approved resources over the course of the EAP.
EAP users will incur no charges throughout the course of the program, and will be able to continue using MedicineBow incurring standard ARCC charges following the end of the program.
Researchers approved for the early access program will get access to their reserved share of 152 GPUs , 4224 CPUs, and 3PB of research data storage over the course of the program.
The new MedicineBow HPC environment includes the following hardware:
Allocations | #Nodes | Cores / | RAM | GPU | GPUs / Node | Processor | TensorCores/GPU & CUDACores/GPU |
CPU | 25 | 96 CPU Cores per Node
| 1024GB/Node | N/A | N/A | 2x 48-Core/96-Thread 4th Gen AMD EPYC 9454
| N/A |
A30 GPU | 8 | 24GB/GPU | Nvidia A30 | 8 | 224 TC/GPU | ||
L40S GPU | 5 | 48GB/GPU | Nvidia L40S | 8 | 568 TC/GPU | ||
H100 GPU | 6 | 80GB/GPU | Nvidia SXM5 H100 | 8 | 528 TC/GPU |
The total computational power available to the Medicine Bow cluster per year is as follows:
1,331,520 GPU hours
11.86 petaflops @ fp32
2.72 petaflops @ fp64
37,002,240 CPU hours
8.24 petaflops @ fp32
2.36 petaflops @ fp64
3 Petabytes available for research data storage
ARCC will accept applications through Friday May 17th, 2024. Applications are open to all scientific disciplines. To apply, please fill out an application through our portal, here.
The application form will require an abstract associated with your intended work, the technical requirements associated with any computational technologies (if needed), estimated compute hours and storage needs.