The purpose of HOLL-I is to provide access to extreme scale, full wafer processing for AI/ML and other research.

Technical Specifications

–  CS-2 Accelerator (All on a Single wafer)

• Custom Cores: 850,000

• Memory: 40GB (at 20PB/s bandwidth)

• 1200 Gb/s of I/O network bandwidth

–  Interconnect: 100 Gb Ethernet

–  Storage options: 200TB local and shared file system (Taiga)

–  Support cluster running vendor software stack for interfacing with the accelerator system

–  Currently support large subsets of Tensorflow and Pytorch ML frameworks

Free trial available to see if HOLL-I works for you.