Air Force missions, such as persistent surveillance of large areas, require massive data analytics on supercomputers to deliver the critical capability of finding the proverbial ―needle in the haystack and thereby help humans avoid sensory overload. At another extreme, covert special operations forces have limited communications, limited time and limited battery capacity but need functionality from a portable computational capability that only a few years ago would have taken a supercomputer. Even more daunting, autonomous operation of bird-sized micro air vehicles demand high capacity computer operations be carried out in physical spaces equivalent to golf ball sized brains. This challenge is becomes even more difficult when vehicles are shrunk to bug-sized around 2020. The combination of massive data analytics on supercomputers and embedded high performance computing enables new mission capabilities for the Air Force.
As captured in Table 4.1, the first technical challenge that directly addresses all these mission needs is achieving energy efficiency at the system level and finding the technical means for another 700X improvement over the next 15 years. Energy efficiency needs to be a first order, if not the primary, design criterion driving system engineering tradeoffs. Technology advances such as three dimensional stacking can be game changers, but not if the stack overheats from power hungry chips.
No comments:
Post a Comment