Moore is dead since 2004, and the I think the parallel processing doesn't help too much with the sensor fusion.My talk about comparing chips does make a point. Ever heard of Moore's law? The Russians have better processors available to them than either the i960MX processors used in the F-22 or the PowerPC processors initially used in the F-35. AFAIK Russian military-industrial projects in aviation and space use some Elbrus derivative of the SPARC architecture. Probably the MCST-R1000. That is a quad-core 64-bit CPU with vector instructions fabbed at 90nm. Which is outdated but less so than the PowerPC processors used in the F-35. Let alone the i960MX.
I think that is part of the reason why the F35 is in deep trouble and carry so non affordable development price tag for the US.
A pathetic 1 megapixel consumer grade sensor can generate 1.2 gigabit of raw data stream.
1000 AESA Rx processing element with 100 MHz bandwidth each can generate 100 gigabyte /sec without any trouble.
Now, to synthesise these data require random access, and that doesn't improved a lot in the past decades memory wise.
Actually, that is the reason why the increased memory bandwidth doesn't has too much impact on the CPU performance.
I think it is safe to say the "sensor fusion" has more similar processing requirement to the binary space partitioning than to the matrix multiplication.