Phase III of the Hyperion High Performance Computing cluster has been installed, tested, configured, and placed into production. The 86 new nodes all feature dual socket Intel Xeon Platinum 32-core 2.6GHz processors for a total of 64 cores per node and 960 gigabytes of locally attached NVMe scratch storage. Eight of these systems will be big memory nodes with 2 terabytes of main memory, and 78 will be general computational nodes with 256 gigabytes of main memory.
The Phase III upgrade brings Hyperion to a total of 356 nodes and will increase Hyperion’s theoretical peak performance in quadrillion floating-point operations per second (petaFLOPS or PFLOPS) by 50%, from 1.23 PFLOPS to 1.85 PFLOPS. It will also increase the number of computing cores available by 14%, for a total of 17544 cores.
Research faculty can become Hyperion cluster contributors to gain priority scheduling on their owned nodes for the lifetime of the equipment. Pricing is $10,000 each for standard computational node and $21,000 each for big memory node. To purchase Hyperion Phase III nodes please enter a request through ServiceNow.