It is expected to deliver outstanding performance, especially in the field of artificial intelligence computing.

NVIDIA's accelerated computing platform is being used to develop an artificial intelligence supercomputer at Tokyo Institute of Technology.

TSUBAME3.0, a new system currently under development at Tokyo Institute of Technology, is expected to deliver more than twice the performance of its predecessor, TSUBAME2.5. Utilizing NVIDIA Pascal architecture-based Tesla P100 GPUs, which offer nearly three times the efficiency, the new system is expected to reach 12.2 petaflops of double-precision computing performance, raising expectations that it will enter the top 10 of the TOP500 supercomputer rankings, which will be announced in November.

▲Tokyo Institute of Technology's TSUBAME utilizing the NVIDIA accelerated computing platform

TSUBAME3.0 delivers AI computing performance exceeding 47 petaflops, promising exceptional performance, particularly in AI computing. When run concurrently with TSUBAME2.5, it is expected to achieve 64.3 petaflops.

TSUBAME3.0, which will be operational starting this summer, will be used for educational and high-tech research purposes within Tokyo Institute of Technology and will also be open to external private researchers, and is expected to serve as an information infrastructure center for major universities in Japan.