Hi, im wondering at the ever increasing power requirments for processing especially graphics related.
It seems the move is towards paralleling the procesing such as dual/quad core with cpu and parallel pixel fill in gfx engines.
wich seems reasonable.
but the way they do it is to have the all the ram in one place and all the processing in another, wich leaves a remarkable bottleneck inbetween.
the incredible high speed of the interface must be power hungry.
transputers springs to mind but these needed external RAM. would it not be possible to integrate the RAM and processing on the same chip ?
some mcu im using have 32k ram. its so much much more efficient to use internal ram than to use via an external bus.
it would need a lot more RAM for gfx processing, although im using less than this to bitmap drive a 1/4 vga lcd. although say 100 16 bit micros with 64k internal video ram each might be quite clever.
if you could put them all onto just a few dies youd have a fair amount of redundancy.
some fpgas have 1M gates ive noticed, if this could be translated into ram area this would be interesting.
PC graphics cards have 128mb or more, much of this I assume is for textures, this might be difficult to distribute.
Colin =^.^=