Is government intervention necessary for broader HPC adoption?

The largest computing systems have historically required government support, as the market for these machines is too small to interest most vendors. Even today, the US government purchases the biggest machines itself and pushes select corporations through DARPA’s High Productivity Computer Systems (HPCS) initiative. But such intervention does not guarantee economic sustainability.

Indeed, commodization and ease-of-use has given greater access to technology than any government program. It seems more appropriate to supplement a low-cost baseline platform with co-processors and special software. For example, MATLAB and blade-based “personal supercomputers” can be augmented with Star-P and ClearSpeed. So government intervention here seems rather unnecessary.

As a side note, it appears that much of the worry on competitiveness and innovation stems from xenophobia. The Americans were spurred into action more by the success of the Japanese than by a genuine desire to do science. RIKEN’s MDGRAPE-3 may well be the world’s first petaflop computer. Congratulations are in order.