Welcome¶
Info
Production operation started on October 14, 2026. Users continuing from Early Access Phase 2 do not need to submit a new account application.
Supercomputer RIKYU is a system installed at the RIKEN Center for Computational Science to accelerate scientific discovery through AI. The system consists of 400 compute nodes equipped with NVIDIA GB200 NVL4, for a total of 1,600 NVIDIA Blackwell GPUs. Its FP64 performance is 64.160 PFLOPS, and its FP8 performance is 15.539 EFLOPS.
This user guide explains how to use Supercomputer RIKYU.
Account Application¶
To apply for an account on this system, use the RIKYU Account Application System (RAAS) below.
Note
Users continuing from Early Access Phase 2 do not need to submit a new account application. New users must submit an account application. Production operation is available to ARiSE users, members of accepted SPReAD1000 projects, and RIKEN members. Participation in a project is required to use this system. To issue an account, the user must register their personal information in RAAS, and the project representative must register the project, add project members who have registered their personal information, and complete the required project procedures.
Usage Fees¶
The usage fee for this system is 300 yen per GPU hour. For example, using 4 GPUs for 5 hours costs 4 GPUs × 5 hours × 300 yen = 6,000 yen. Consumption tax is charged separately. Usage fees are billed in arrears. Usage from October to December 2026 is scheduled to be billed in January 2027, and the billing schedule from January 2027 onward will be announced separately. Details of the fees will be provided separately to project representatives.
To check your current usage, refer to the RIKYU Portal.
Acknowledgments¶
When you publish research results obtained using this system, such as in papers, and an acknowledgment is needed, write it based on the following example. Include "(in part)" if this system was used for only part of the research results.