Will magnetologic allow for future end-all uber chips?

[zidane1strife]

http://www.edpsciences.org/articles/epl/abs/2004/12/epl8213/epl8213.html

Nowadays logic gates of computers are based on transistors and increasing the computational power means increasing their integration density. Among others, the utilization of magnetism is a promising alternative. We propose a magnetologic gate consisting of a single magnetoresistive element with three independent inputs as a versatile universal gate for computing. Each of these gates can represent one out of at least ten logic functions, among them the four basic and two reversible ones. The functionality can be pre-programmed at run-time and the output is non-volatile. The increased functionality and flexibility as well as the drastically reduced number of computing elements may induce a paradigm shift in microprocessor layout and operation.

I only found an abstract on the web, but I was reading about this in the latest sciam... and d@mn if they're not hot, if they're indeed all they're cracked up to be.

Supposedly these babies allow for h/w that can be reprogrammed at the gate lvl billions of times per second, and you've not to waste clocks/energy synching reads(or something like that) cause it's not volatile. Only problem problem is they're a bit far off, but they appear to be quite impressive.

Can you imagine h/w that can run any application as good as dedicated h/w by reconfiguring on the fly? Mindboggling, and when it comes to the gphx arena, I'd think, these would even surpass unified designs done with more traditional means in terms of performance.

What think?

 

[The_Wolf_Who_Cried_Boy]

Can you imagine h/w that can run any application as good as dedicated h/w by reconfiguring on the fly?
What think?

My initial thought is how would you validate each reconfiguring of the logic? It's good for a chip to be running at optimum efficiency but it's still more important that it runs correctly.