CPU only partition |
- Number of nodes
184
- core / node (HPE Cray EX425)
2 x 64 core AMD EPYC 7763 (Milan)
- RAM / node
256 GB
- interconnect
200 Gb/s Slingshot
- Total 23.552 db Core
Performance: 0,9+ petaflops (Rmax)
|
GPU (accelerated)
partition |
- Number of nodes
58
- core / node (HPE Cray EX235n)
1 x 64 core AMD EPYC 7763 (Milan)
and 4 x NVIDIA A100 GPU
- RAM / node
256 GB
- interconnect
2x 200 Gb/s Slingshot
- Total 232 db GPU
Performance: 4,6+ petaflops (Rmax)
|
Artificial intelligence
(AI) partition |
- Number of nodes
4
- core / node
2 x 64 core AMD EPYC 7763 (Milan)
- GPU / node
8x NVIDIA A100
- RAM / node
512 GB
- interconnect
2x 200 Gb/s Slingshot
- Total 32 db GPU
Performance: 0,6+ petaflops (Rmax)
|
Big Data
(Data Analytics)
partition |
- Number of nodes
1 db (SMP/NUMA)
- core / node
16 x 18 core Intel Xeon Gold 6254
- RAM / node
12 TB
- Disk
2 TB SSD
- interconnect
2x 200 Gb/s Slingshot
- Total 288 db Core
Performance: 30 teraflops (Rmax)
|
Storage |
- scratch (Cray ClusterStor E1000)
400 TB NVMe Gen 4 flash
- project (Cray ClusterStor E1000)
2,3 PB
- archive (HPE DMF + Spectra Tfinity tape library)
10 PB
|
Network |
- Slingshot switches
36 (32 Colorado and 4 Columbia)
- Compute node edge links
300
- Topology
Dragonfly
- Link speed
200Gb/s
|
