Overview of HBW2 subsystem

The HBW2 subsystem is a complementary system to the HBW2 system.
This system is available to users of RIKEN projects.

Configuration of HBW2 Subsystem

• Wako Large Memory Server (WLMS)
• 2 node (192 core)
• CPU: Intel Xeon Gold 6418H 2.1 GHz, 96 core/node, 3 TiB/node
• GPU Server (GPUS)
• 4 node (448 core)
• CPU: Intel Xeon Platinum 8480+ 2.0 GHz, 112 core/node, 512 GiB/node
• GPU: NVIDIA H100(SXM5), 4 GPUs/node
• FPGA Server (FPGAS)
• 2 node (160 core)
• CPU: Intel Xeon Platinum 8380 2.3 GHz, 80 core/node, 512 GiB/node
• FPGA：IA-840F(Intel Agilex AGF027)、2 FPGAs/node
• Kobe Large Memory Server (KLMS) / Yokohama Large Memory Server (YLMS)
• 1 node (96 core ) / 1 node (96 core)
• CPU: Intel Xeon Gold 6418H 2.1 GHz, 96 core/node, 8 TiB/node
• Attached storage (100TB)
• Located in Kobe West and Yokohama area. Accounts and file systems are independent.

Usage Guide for HBW2 Subsystem

• Terms of use: RIKEN internal project member of HBW2
• Add/Usage application: Please apply for subsystems from the “Other Applications” menu on the HBW2 portal.
• The operation policy may be changed depending on the situation.

Wako Large Memory Server (WLMS)

• Allocation of 1% of the total annual computation time to a project without a usage application
• Up to 20% can be added to a project with an additional application

GPU Server (GPUS)

• Allocation of 1% of the total annual computation time to a project without a usage application
• Up to 20% can be added to a project with an additional application

FPGA Server (FPGAS)

• Available with a usage application
• Permission is granted on a project basis and can be used in subsequent years without application
• Use by occupying a node

Kobe Large Memory Server (KLMS) / Yokohama Large Memory Server (YLMS)

• Available with a usage application
• Permission is granted on an account basis and can be used in subsequent years without application
• Use by logging in to the node