In order to scale to the point claimed in the video on their web site, you would need extraordinarily fast crossbar elements to keep all the processors humming. I see none of that design in the web site graphic. The OS has to be very efficient at parsing tasks to the various processors, serving them up, and funneling the completed calculations to appropriate storage or caching. That's why most designs use some other processors (like the Cell and AMD combo, that run a version of RedHat and a controlling systems software)to run like a traffic cop. I don't think that this chip could run like that, at least, not in its present form. In the example of the teraflop computer, above, the system is not doing something (besides a benchmark) that would be useful in the supercomputer world. Like some overclockers, because you can theoretically run a calculation fast does not a system make. The ARM chips for instance, are created on 130 nm CMOS process, and the sheets say that it is good at 720p scaling (when you really want 1080p minimum scaling). And I saw 4K displays at CES, though they aren't ready for prime time. That doesn't mean that this isn't a good design for video processing. Or that furture programming, and running multiple chips together, would not be really great in hand held and other smale video/media devices. But a supercomputer, to do weather forecasting, protein folding, and nuclear weapons stockpile decay? Really.
To address a second point: human stem cells pass along a programmed and path continuum, becoming increasingly differentiated. That process is thought to be goverend by checmical gradients, which change what genes are able to be expressed. I see nothing similar here, except for the overreaching analogy of making something generic something specific. But that ignores the myriad complexities of responding to the environment, and making semi-permanent changes to the cell make up to allow it be differentiated. And this says nothing about a process of dedifferentiation. Maybe I did too much molecular biology! Or maybe I have been burned by too many great ideas at CES/Comdex, that never achieved critical mass for one reason or another. Good comments, though, and I will also keep a close watch on this, the potential is there...
Facebook
Twitter