Shaheen II [KAUST Supercomputer]


HPC Vendor: Cray XC40 compute cabinets
CPU: 6,174 dual sockets compute nodes (16-core Intel Haswell @2.3GHz) = 197,568 cores
Mem: 128GB of DDR4 memory per node = 790TB of aggregate memory
Network: Cray Aries Interconnect with Dragonfly topology
Burst Buffer: Solid-state Drives fast data cache of 1.5 PB capacity @1.5 TB/s bandwidth
Performance:  7.2 Pflop/s theoretical peak
                     5.53 Pflop/s sustained LINPACK


Ibex [KAUST Computer Cluster]


Intel Skylake 40-core, 384GB Memory Nodes [x106]
Intel Cascade Lake 40-core, 384GB Memory Nodes [x106]
AMD Rome 128-core, 512GB Memory Nodes [x108]
Large Memory Nodes, 3TB Memory [x18]

GPU Models

A100, 80GB HBM GPUs [248]
V100, 32GB GPUs [x272]
P6000, 22GB GPUs [x4]
P100, 16GB GPUs [x20]
RTX 2080ti, 12GB GPUs [x32]
GTX 1080ti, 12GB GPUs [x64]


Shaheen III [FUTURE KAUST Supercomputer]

HPC Vendor: Hewlett Packard Enterprise

CPU Nodes

CPU: 4,608 dual sockets compute nodes (96-core AMD Genoa) = 884,736 cores
Mem: 384GB of memory per node = 1.77PB of aggregate memory

GPU Nodes

700 Grace Hopper Superchip (4x64 ARM cores directly attached to NVIDIA H100 GPUs)
Mem: 384GB of memory per node = 1.77PB of aggregate memory

Expected Performance: 125 Pflop/s theoretical peak