HOME > Introduction > Cluster Computing System

Cluster Computing System : HEP Apollo 2000 and DL380

History
Shared-memory System
Cluster Computing System
Applications

HPE Apollo 2000 has been introduced as a Cluster Computing System in Supercomputer System.

Apollo2000
Apollo2000 (CPU nodes) has the following features:

CPU nodeApollo2000
CPUIntel Xeon Gold 6348 (Ice Lake)
Clock frequency2.6GHz
L3 Cache42MB
Number of cores per a node56
CPU peak performance per a node4.65TFLOPS
Memory per a nodeDDR4-3200 256GBDDR4-3200 512GBDDR4-3200 1,024GB
Storage per a node480GB SATA RI SSD
OSRHEL 8
InterconnectBetween nodes:
InfiniBand HDR200 (200Gbps)
Number of nodes444020
CPU peak performance205TFLOPS186TFLOPS93.2TFLOPS
Batch queueSMALL, APC

Apollo2000 (CPU nodes) has the following features:

GPU nodeDL380G11
CPUIntel Xeon Platinum 8462Y+ (Sapphire Rapids)
Clock frequency2.8GHz
L3 Cache60MB
Number of cores per a node64
CPU peak performance per a node5.73TFLOPS
Memory per a nodeDDR5-4800 1,024GB
Storage per a node480GB SATA RI SSD x2 (RAID1)
GPU per a nodeNVIDIA H100 80GB x2
OSRed Hat Enterprise Linux 8
InterconnectBetween nodes:
InfiniBand HDR200 (200Gbps)
Number of nodes9
CPU peak performance51.6TFLOPS
Batch queueAPG
GPUNVIDIA H100 (PCIe)
Compute Capability9.0
Micro architectureNIVIDA Hopper
TF32 Tensor cores(TFlops)378
756(with sparsity）
FP64 Tensor cores(TFlops)51
FP32 (TFlops)51
FP64 (TFlops)26
Tensor cores456
FP32 cores14,592
Interconnect interfacePCI Express Gen5
Interconnect bandwidth128GB/s
GPU Memory80GB HBM2e
GPU Memory bandwidth2,000 GB/s

 

Link

HPE Apollo 2000
HPE ProLiant DL380