DARPA in co-operation with Intel and the San Diego Supercomputer Center are setting their sights very high indeed. DARPA plans on having an exaflop capable computer by 2018.
What is truly amazing is just how fast this system will be. From the physorg blog:
What is truly amazing is just how fast this system will be. From the physorg blog:
- An exaflop is a thousand times faster than a petaflop, itself a thousand times faster than a teraflop. Teraflop computers —the first was developed 10 years ago at Sandia — currently are the state of the art. They do trillions of calculations a second. Exaflop computers would perform a million trillion calculations per second.
That's right, DARPA's computer would be 1 million times faster than present day super computers!
The project is the brain child of Institute for Advanced Architectures, formed cooperatively at Sandia and Oak Ridge national laboratories which are researching some of the drawbacks to the ultrafast machines. The one major problem that the super computers face is, the faster they operate, the farther the data has to travel, effectively slowing the machine down. This slow down comes from the huge amount of processors used. A single processor may call for data that may be stored in a processor tens of thousands processors away.
Physorg article Wiki on FLoating point OPerations per Second (flops)
The project is the brain child of Institute for Advanced Architectures, formed cooperatively at Sandia and Oak Ridge national laboratories which are researching some of the drawbacks to the ultrafast machines. The one major problem that the super computers face is, the faster they operate, the farther the data has to travel, effectively slowing the machine down. This slow down comes from the huge amount of processors used. A single processor may call for data that may be stored in a processor tens of thousands processors away.
Physorg article Wiki on FLoating point OPerations per Second (flops)
No comments:
Post a Comment