HPC Productivity Perspective

High Productivity Computing Systems (HPCS)

The High Productivity Computing Systems Program was initiated by DARPA to address the very real and expanding HPC capability gap documented by a number of DoD studies. The HPCS initiative incorporates strong collaboration and sponsorship from a number of key government agencies such as NSA, NRO, DOE, NNSA, NASA and DDR&E.

The High Productivity Computing Systems program will bridge the gap between the late-80’s based technology of today’s High Performance Computers and the promise of quantum computing for the Department of Defense. DARPA’s challenge is to develop a broad spectrum of innovative technologies and architectures integrated into a balanced total system solution by the end of this decade. There is a subtle but important change in emphasis from past programs in high-end computing. The new emphasis is now on productivity or value.

Using raw theoretical peak computing performance as the single evaluation criteria is not sufficient. Total end user application, computing life cycle costs and mission responsiveness are critical to future large tera- and peta-scale computing installations and end users. An analyst’s idea-to-solution or time-to-solution is more important than raw computing capacity. It is time we move beyond Moore’s Law the doubling of microprocessor performance every 18 months and double the productivity value instead.