Hacker News new | ask | show | jobs
by jacquesm 5719 days ago
It's a problem because in chip therms that is tiny, and you still need to interface to the outside world, using regular wiring you're much slower than 300GHz, and using optics you have a pretty serious interfacing issue.

Of course even a 10-fold practical increase would be an amazing thing, but I just can't see the 300GHz ever becoming a reality given the constraints that physics imposes on this.
2 comments
sasvari 5719 days ago
but I just can't see the 300GHz ever becoming a reality given the constraints that physics imposes on this.

There are already proven cutoff frequencies for InP-based bipolar transistor up to 755 GHz with demonstrated amplifiers with fmax = 324 GHz.

http://www.ece.ucsb.edu/Faculty/Rodwell/publications/2008_10...

link
jacquesm 5719 days ago
Sure, but that does not translate in to computing speed.

I would not be surprised if that turned out to be 1/10th or less of the device theoretical maximum switching speed.

300GHz is a frequency at which just about every conductor is a coil of sorts. Stripline all the way :)

Or did you miss the 'in chip terms'?

link
sasvari 5719 days ago
Or did you miss the 'in chip terms'?

Sorry, you are right. The more complex your chip is going to be, the more trouble you have with RF issues.

Nevertheless, what I meant is, there will be components working in the low THz range, as they are already demonstration objects. Most probably you won't see this components in consumer products or in computing, due to cost, scaling and implementation issues. But for some applications I think this will be applicable.

link
thefool 5719 days ago
What?

First, what is stopping wires from transmitting these frequencies.

Second, 300GHz switching speed does not necessarily mean that you need to be feeding info at that rate.

link
jacquesm 5719 days ago
Every wire is not 'just' a wire it is also a capacitor and a coil as well as a resistor.

The parasitic LCR filter created by even the shortest stretch of circuitry is going to have a cut-off frequency well below the frequencies quoted here.

The 'C' component is further increased by the capacitance of the input on the other end, and charging that capacitor is slowed down further by any leakage there (another R).

link