"To analyze this using silicon computers," Tabib-Azar says, "you need a bunch of on-off switches that have to turn on or off in a particular sequence to give you the output, whether you go to dinner or not. But just a single one of these [MEMS logic gate] devices can be designed to perform this computation for you."

That's actually the difference between ASICs and general purpose processors. Nothing says you can't make a complex logic block out of a silicon switch - it's just that if you do design such a part, you can't change its function once it's built. It's made for that one specific task.

Again a temperature in Fahrenheit..!?

It's an interesting concept. But they really need to test this for more than two hours. Long term exposure could lead to embrittlement of MEMS which would also lead to failure.

Though, depending on the timeframe of the task at hand, that can be an acceptable risk.

Nothing says you can't make a complex logic block out of a silicon switch - it's just that if you do design such a part, you can't change its function once it's built.

Why not? If you already have switches then you can also use them to control the connection of the MEMS.

If you think back: That's what mechanical looms did (which were the forerunners of today's computers). They used punchcards to reconfigure so that different patterns could be woven on one loom.

Why not? If you already have switches then you can also use them to control the connection of the MEMS.

Because the circuit itself is not programmable. The ASIC, once built, does nothing else than what it is designed to do, which is why it can substitute the programmable bits with simpler and more efficient shortcuts. It's an Application Specific Integrated Circuit.

What you are thinking about is an FPGA, which is kinda in the middle between a programmable processor and an ASIC. Overall, they may be even more complex than the processor, as in, you can implement a processor inside an FPGA.

What the article talks about is similiar to an ASIC, because it specifically mentions the sort of efficiency where you make a single gate perform a complex, specific function. Trying to re-purpose that same gate then is extremely difficult, just like trying to make the loom knit instead of weave.

back in the 1890,s the computers in telephone exchanges ran in a similar manner.