RC/HPC available resources
RC Compute Resources
- Multi-node parallel / MPI clusters
- Serial / Shared memory clusters / GPU resources
- Main-campus Kubernetes cluster
Main-campus HPC (SciClone)
The main-campus HPC resources (collectively known as "sciclone") consist of a series a different sub-clusters, each of which has a set of hardware resources tied to it. There are two types of HPC resources in general, those used for multi-node parallel calculations and those best suited for GPU/CPU serial, shared-memory applications within one or, at most, a few nodes. All clusters have a maximum walltime of 72 hrs for jobs, excluding kuro which has a maximum of 48 hrs.
Multi-node parallel / MPI clusters
| name |
processor |
cores/ node |
total # cores |
mem (GB) | network | deployed | notes |
|---|---|---|---|---|---|---|---|
kuro |
|||||||
| kuro.sciclone.wm.edu | 2x AMD EPYC 9124 | 32 | 32 | 384 | 10GbE/HDR IB | 2024 | The kuro cluster is currently reserved from 3pm Friday-8pm Sundays |
| ku01-ku47 | 2x AMD EPYC 9334 | 64 | 3008 | 384 | 1GbE/HDR IB | 2024 | |
femto |
|||||||
| femto.sciclone.wm.edu | 2x Intel Xeon Gold 6130 | 32 | 32 | 96 | 1GbE/EDR IB | 2019 | |
| fm01-fm30 | 2x Intel Xeon Gold 6130 | 32 | 960 | 96 | 1 GbE/EDR IB | 2019 | |
bora |
|||||||
| bora.sciclone.wm.edu | 2x Intel Xeon E5-2640 | 20 | 20 | 64 | 10 GbE/FDR IB | 2017 | |
| bo01-bo55 | 2x Intel Xeon E5-2640 | 20 | 1100 | 128 | 10 GbE/FDR IB | 2017 | |
Serial / Shared memory clusters / GPU resources
| name | GPUs |
processor |
cores/ node |
total # cores | mem (GB) | network | deployed | notes |
|---|---|---|---|---|---|---|---|---|
hima |
||||||||
| bora.sciclone.wm.edu | --- | 2x Intel Xeon E5-2640 | 20 | 20 | 64 | 10GbE/FDR IB | 2017 | The hima cluster uses bora for compiling and job control |
| hi01-hi07 |
2x NVIDIA P100 (16GB) 1x NVIDIA V100 (16GB) |
2x Intel Xeon E5-2683 | 32 | 224 | 256 | 1GbE/FDR IB | 2017 | |
astral |
||||||||
| astral.sciclone.wm.edu | --- | 2x Intel Xeon Gold 6336Y | 48 | 48 | 256 | 1GbE/EDR IB | 2019 | |
| as01 | 8x NVIDIA A30 (24GB) | 2x Intel Xeon Platinum 8362 | 64 | 64 | 512 | 1 GbE/EDR IB | 2019 | |
gust |
||||||||
| gust.sciclone.wm.edu | --- | 2x AMD EPYC 7302 | 32 | 32 | 32 | 10 GbE/EDR IB | 2020 | |
| gt01-gt02 | --- | 2x AMD EPYC 7702 | 128 | 256 | 512 | 1 GbE/EDR IB | 2020 | |
gulf |
||||||||
| gulf.sciclone.wm.edu | --- | AMD EPYC 7313P | 32 | 32 | 128 | 10 GbE/HDR IB | 2024 | |
| gu01-gu02 | --- | AMD EPYC 7313P | 16 | 32 | 512 | 1 GbE/HDR IB | 2024 | |
| gu03-gu06 |
2x NVIDIA A40 (48GB) per node |
2x AMD EPYC 7313P | 32 | 128 | 128 | 10 GbE/HDR IB | 2024 | |
Main-campus Kubernetes cluster
Coming...
VIMS-campus cluster (Chesapeake)
The James cluster has a maximum walltime of 72 hrs for jobs. Potomac and Pamunkey allow 120 hrs.
Multi-node parallel / MPI cluster
| name |
processor |
cores/ node |
total # cores | mem (GB) | network | deployed | notes |
|---|---|---|---|---|---|---|---|
james |
|||||||
| james.hpc.vims.edu | Intel Xeon Silver 4112 | 8 | 8 | 32 | 10GbE/EDR IB | 2018 | Restrictions |
| jm01-jm27 | Intel Xeon Silver 4114 | 20 | 540 | 64 | 1GbE/EDR IB | 2018 | |
Serial / Shared memory clusters
| name |
processor |
cores/ node |
total # cores | mem GB) | network | deployed | notes |
|---|---|---|---|---|---|---|---|
potomac |
|||||||
| chesapeake.hpc.vims.edu | AMD Opteron 4238 | 8 | 8 | 32 | 10GbE/EDR IB | 2014 | |
| pt01-pt30 | AMD Opteron 4334 | 20 | 540 | 64 | 1GbE/EDR IB | 2014 | |
| pamunkey | |||||||
| james.hpc.vims.edu | Intel Xeon Silver 4112 | 8 | 8 | 32 | 10GbE/EDR IB | 2018 |
The pamunkey cluster uses james for compiling and job control. The pamunkey nodes do not have infiniband |
| pm01-pm02 | AMD Opteron 6380 | 64 | 128 | 256 | 10GbE | 2016 | |
Please send an email to [[w|hpc-help]] for advice if you are unsure of the most appropriate type of node for your particular application.
