RCIF at WashU

List of Computational Nodes

| Name | CPU | Cores per socket | Sockets | CPU Cores | Real Memory | GPU (gres) | GPU Count | | ————- | ——————— | —————- | ——- | ——— | ———– | ————— | ——— | | node[01-14] | Intel Xeon Gold 6226R | 16 | 2 | 32 | 770000 | | | | node[15-32] | Intel Xeon Gold 6226R | 16 | 2 | 32 | 385000 | | | | gpu01 | Intel Xeon Gold 6226R | 16 | 2 | 32 | 385000 | tesla_a100 | 4 | | gpu02 | Intel Xeon Gold 6226R | 16 | 2 | 32 | 770000 | tesla_v100S | 4 | | gpu03 | Intel Xeon Gold 6226R | 16 | 2 | 32 | 770000 | tesla_v100S | 2 | | gpu04 | Intel Xeon Gold 6226R | 16 | 2 | 32 | 385000 | tesla_v100S | 2 | | gpu05 | Intel Xeon Gold 6226R | 16 | 2 | 32 | 385000 | tesla_v100S | 2 | | gpu06 | Intel Xeon Gold 6226 | 12 | 2 | 24 | 385000 | tesla_v100 | 4 | | gpu07 | Intel Xeon Gold 6226 | 12 | 2 | 24 | 385000 | tesla_v100 | 3 | | gpu08 | Intel Xeon Gold 6226 | 12 | 2 | 24 | 385000 | tesla_t4 | 2 | | gpu09 | Intel Xeon Gold 6226 | 12 | 2 | 24 | 386000 | tesla_v100 | 4 | | highmem01 | Intel Xeon Gold 6240L | 18 | 2 | 36 | 2984000 | | | | highmem02 | Intel Xeon Gold 6226 | 12 | 4 | 48 | 2984000 | | | | gpa[401-412] | Intel Xeon Gold 6338 | 64 | 2 | 128 | 1024000 | tesla_a100 | 4 | | gpua[801-805] | Intel Xeon Gold 6338 | 64 | 2 | 128 | 1024000 | nvidia_a100_80 | 4 | | gpuh[801-802] | AMD EPYC 9334 | 64 | 2 | 128 | 1024000 | nvidia_h100_sxm | 4 | | gpugh01 | Grace Superchip | 72 | 1 | 72 | 1200000 | nvidia_gh200 | 1 | (note: “tesla_a100” is used to denote A100-SXM-40 and A100-PCIe-40 cards)

See also the spreadsheet

Login Nodes

login[1-2]

Summary: 2 x ACTserv x1210 (32-cores, 192GB, 480GB SSD)

top

CPU Nodes

node[01-14]

Summary: 14 x Dual Socket Xeon SP (32-cores, 768GB)

node[15-32]

Summary: 18 x Dual Socket Xeon SP (32-cores, 384GB)

highmem01

Summary: ACTserv x1210 (36-cores, 3TB)

highmem02

Summary: ACT SYS-2049U-TR4 (24-cores, 3TB)

top

GPU Nodes

GPU acceleration adds 432TF of SFLOP performance and 798GB of global memory from 27 GPUs.

gpu01

Summary: ACTserv x2280c (32-cores, 384GB)

gpu02

Summary: ACTserv x2280c (32-cores, 768GB)

gpu03

Summary: ACTserv x2280c (32-cores, 768GB)

gpu[04-05]

Summary: 2 x ACTserv x2280c (32-cores, 384GB)

gpu06

Summary: ACTserv x4170c (24-cores, 384GB)

gpu07

Summary: ACTserv x4170c (24-cores, 384GB)

gpu08

Summary: ACTserv x4170c (24-cores, 384GB)

gpu09

Summary: ACTserv x4170c (24-cores, 384GB)

top

gpua[401-412]

top

gpua[801-805]

top

gpuh[801-802]

top

gpugh01

top

Gateway Nodes

Summary: 4 x ACTserv x1210 (32-cores, 192GB)

top

Storage

For architectural concepts and usage of storage, see here.

High-throughput file system (scratch)

BeeGFS storage with 1.3PB (i.e., 1,327,032 GiB) of total usable storage.

3 x BeeGFS HDD storage block

beegfs-oss01.cluster [ID: 1]

beegfs-oss02.cluster [ID: 2]

beegfs-oss03.cluster [ID: 3]

beegfs-oss04.cluster [ID: 4]

beegfs-oss05.cluster [ID: 5]

beegfs-oss06.cluster [ID: 6]

beegfs-mds01.cluster [ID: 1]

beegfs-mds02.cluster [ID: 2]

Large-volume file system (persistent)

Ceph storage with 8.1PB of total usable storage.

33 x Ceph HDD storage block

3 x Ceph meta data server

croit-mds01

croit-mds02

croit-mds03

top

Infiniband

1 x HDR Network

top

Legacy

CHPC2 has 528 CPU cores in 23x CPU-based nodes and 11x GPU-accelerated nodes. GPU acceleration adds 122.5TF of SFLOP performance and 160GB of global memory from 31 GPUs.

top