The Government Revitalization Unit, a panel established by the government to eliminate wasteful spending, recommended on Friday to freeze spending on the project. The Institute of Physical and Chemical Research (Riken), which is leading the ¥115 billion (US$1.3 billion) project, has already spent over ¥50 billion and had requested around ¥27 billion for the next financial year.
Its goal was to create a supercomputer by 2011 with a performance of 10 petaflops (a petaflop is 1,000 trillion floating operation points per second). The current fastest supercomputer is a Cray XT5 supercomputer installed at the U.S. Department of Energy’s Oak Ridge Leadership Computing Facility. It has a performance of 1.75 petaflops, released on Friday.
Members of the panel questioned whether Japan really needed to have the world's fastest supercomputer and its relevance to the everyday lives of citizens.
The high-profile withdrawal from the project earlier this year by two participants appears to have contributed to the decision. Riken had originally given a contract for the computer to a consortium of NEC, Hitachi and Fujitsu, which are Japan's three top supercomputer makers, however the former two companies pulled out of the project leaving it in the hands of Fujitsu.
The move also meant the system architecture had to be changed. It was intended to mix vector and scalar components, with NEC and Hitachi responsible for the vector part and Fujitsu working on the scalar part, but the change made the system scalar only. A scalar processor is the most common type of CPU and handles instructions one at a time while a vector processor, more commonly used for scientific work, handles multiple instructions simultaneously.
The decision to freeze spending on the project took about an hour to make and was one of a number of budget items reviewed by the Government Revitalization Unit on Friday. The panel is an initiative of the recently-elected Democratic Party of Japan-led government and is seeking to cut around ¥3 trillion in unnecessary spending to reduce the size of Japan's record ¥95 trillion national budget for next year.
For the supercomputer project the end is not quite decided. A final decision on whether to freeze spending remains to be made although the outlook appears bleak.
Ironically the decision was made the same day Kenichi Miura, a fellow at Fujitsu Labs and director of grid research at Japan's National Institute of Informatics, was awarded the IEEE's 2009 Seymour Cray Computer Engineering Award. The award was given for Miura's contributions to the development of vector supercomputer hardware and software.