However, to achieve true high clock rates, it is my undestanding that it is only possible through optical clock distribution.

https://youtu.be/D3HEltp3OYY?t=788

https://rainerklute.wordpress.com/2015/02/27/video-transcript-poet-technologies-at-city-investors-circle/

13:03 – But gallium arsenide can. So our structure, our basic structure of the optical thyristor can lase, can be an optical driver and can also be an optical receiver, receive light and bring it into the chip.

• We can do edge emitting input and output,
• we can do vertical cavity lasers and receivers,
• and we can also distribute optical lights inside the chip.

13:34 – One thing today that’s plateaued the performance of processors inside computers you can buy today, for the last eight years maybe, if not 10 … Every time you buy a computer the processor’s speed plateaus at about 3 Gigahertz (GHz), 3.2 GHz. If you pay $400 more, you might get 3.6 GHz for a specific processor or MacBook Pro or something.

14:04 – What we’re talking about with distributing clocks internally with optical signaling, is taking that barrier and blow it passed 5 GHz, 10 GHz. The problem with distributing the clock electrically, through wires internally to a processor, is it requires kilometers of wiring, believe it or not, and about 30 percent of the power is used with driving that clocking network really hard, at a 3 GHz heartbeat to get to all of the leafs that need that clock simultaneously. At 3 GHz heartbeats a signal needs to get to everywhere inside the chip.

14:52 – Doing this optically saves about 70 percent of that power, which consumes today about 30 percent of the processor power. So if your processor is a 100 Watt, you are spending 30 Watt just distributing the clock, and doing it optically will save you about 70 percent of that power. So you are going to save 70 percent of 30 Watt, which is like 21 Watt right there that you save, just because you [are distributing optically], and you are going to go at 10 GHz and not 3 GHZ.