Blank-Powerful-Client Architecture

Imagine a client is based not on a CPU like Intel Core, and not even on a GPU like the one built in a video card by NVIDIA or AMD/ATI. Imagine a client's main brain were some CPLD or FPGA, that could be configured dynamically.

 In other words it would be sort of GPU, that is devided not into 250 "hard-wired" processor units, but rather into 100 000 elementary parts of that 250 units. And you can download the cofiguration into the CPLD/FPGA to force it to be 250 graphic processors, or  force it to be 1000 of just 16 bit general purpose processors, or whatever configuration the current running task is demanding.

Such CPLD/FPGA  hardware is universal, so they could be mass produces (much more than GPUs) and thus be cheap. And also, since the CPLD/FPGA  technology evolves, the number of elements will be getting virtually unlimited. So, why should we "hardwire" CPU architecture? Just download the architecture "On-Demand" to target the current task.

The concept resembles somewhat Visual WebGui "Empty Client Architecture". The client also can be very lite on software initially. The point is that the client is very poweful in terms of hardware and the hardware is very cheap and universal. Microsoft, in my opinion, should rethink the concept of Operating System to address the hardware revolution.

The lowest level of Op System in the future would be not operating a "hard-wired" CPU, but rather gathering the Processor Matrix from a huge set of available elementary blocks and then to build from some repository the most appropriate higher level Software System. The repository could be local, of course, and the client could be autonomous, but, in the future, the communication to some "Cofiguration Knowledge Center" would become more and more vital.

No Comments