System B: ohtaka
With a total theoretical computing performance of approximately 6.9 PFLOPS, System B (ohtaka) is one of the most powerful supercomputer systems in Japan, ranking 87th on the Supercomputer Top500 list as of November 2020. To meet a variety of needs, it is outfitted with a standard configuration CPU node consisting of 1680 nodes and a Fat node comprising 8 nodes equipped with large-scale memory. The CPU node features AMD EPYC 7702 (Rome) processors with 128 cores per node, which allows even a small number of nodes to perform large-scale parallel computations. Users can perform calculations on as many as 144 CPU nodes (~ 590 TFLOPS) when running a job involving a single calculation, with a total of 18,432 cores available for performing very large-scale parallel computations (as of October 2020).
Starting in July 2026, the number of CPU nodes is scheduled to be reduced from 1,680 to 1,536.
For details on the configuration of System B, ohtaka, click here.
- System Configuration
- Guidelines, documents, etc.
*Researchers intending to use this system need to apply in advance.
Please click here for details.