One thing that I found particularly interesting is the way that it partitions groups of processors per application. It reminds me of the way that IBM's Scalable

One thing that I found particularly interesting is the way that it partitions groups of processors per application. It reminds me of the way that IBM's Scalable Power 2 system worked. This Tilera chip looks remarkably similar, just a LOT smaller, and less FP oriented; although I suspect that in computational applications, the contemporary GPU solutions, including Larrabee, will probably blow it away, it looks like something that has a lot of potential in I/O bound applications like web servers and DBMS's, as others have pointed out. If they can get enough developer interest, they could have a winner on their hands. I could see this sort of thing being very popular in web server farms, as well as in data centers.