interesting article FPGA routing field programmable nanowire interconnect (FPNI)

Yes, it is (interesting).

The one big problem that I see is that 20% of the connections don't work, and the defects are random.

So, the chip has to also be able to provide self-healing, or self-routing that is self-testing, or something like that in order to function. Timing closure is yet another issue with random wires and faults.

HP was also pretty harsh: "multi-core is an admission of failure" is pretty nasty to be saying about Intel's strategy. Not sure who reviewed that press release.

There is a long way to go until there is a product that makes use of this technology, but it means that as an IC designer, I am likely to stay employed.

Austin

The author said it might be available as soon as *2020*. In terms of electronics, that is several lifetimes away. It might as well be science fiction. Most of us will be contemplating retirement before that "optimistic" date arrives, and somehow I doubt that the design landscape will even resemble what it is now.

I did see earlier notes on beam cross-bars, which sounded real, but the

HP stuff is 100% vaporware, as they have only simulated it.

It will be impressive if, when they finally build it (~end 2007), it works as simulated :)

No comments on OTP or Reprogrammable, but either way, defect coverage sounds like expensive testing time.

Why is HP researching this stuff ?

-jg