Massively Multicore

It seems that the logical extension of multicore architectures are systems with truly massive numbers of cores. Intel, for example, is working on an 80-core chip. The potential of these miniaturized wonders of parallelism is hard to even fathom. Scientific computing comes to mind—massive simulations of complex systems can always get more massive and complex, and consume more processor resources. 3D gaming can get more photorealistic. And as-yet-unimagined AI breakthroughs may become possible with that kind of ubiquitous horsepower. But we won't know about any of that until we have a mature and practical method for moving data between these general-purpose processors. Enter the Network-on-Chip (NoC). Luca Benini and Giovanni De Micheli continue with part two of their series on this cutting-edge concept.

Posted by Kevin Carlson at 02:36 PM  Permalink