Computing speed is defined by two factors: Firstly, how fast data can be transferred, and secondly how fast can data be processed. So, everything depends on the speed of electricity. But photons have a higher bandwidth than electrons. That’s the reason why optical fibre is a faster medium of data transmission. But what about computation?
The current computation scenario
Currently, we are using optics in computation, but not harnessing it fully. Optical Readers are used in form of CDs, DVDs, Blu-Ray Discs. Data transmission over a wired network is being done through fibre optic cables. But behind the scenes: we are converting electrons to photons and vice versa at each node. This slows down the overall computation speed. We’ve greatly reduced latency by switching from vacuum tubes to transistors. Our next step here should be An All-Optical-Computer.
The Laser Transistor
Transistors are fundamental building blocks of a computer system. They are essentially micro-switches that turn on and off and also transfer the output to the next transistor and hold the signal. Currently, computer chips are made of silicon transistors. These silicon chips work on electronic signals thus struggle with the problems of photon-to-electron conversion.
In 2004, University of Illinois researchers Milton Feng and Nick Holonyak Jr. developed the concept of the laser transistor. This replaces one of the two electrical outputs of normal transistors with a light signal in the form of a laser, providing a higher data rate. This laser transistor can be used as an optical processor, thus removing the need to convert photons to electrons for processing data.
“Similar to transistor integrated circuits, we hope the transistor laser will be [used for] electro-optical integrated circuits for optical computing,”
~ Feng told OpenMind.
On the fly computation
Parallelism is a concept in which multiple tasks are executed by a computer simultaneously. Multiple wavelengths of light can be employed for a parallelistic approach. Thus, we can easily build a scalable system with low power and lower error rates. Moreover, optical RAMs will have the ability to process data on the motion. In other words, the CPU won’t need to schedule tasks and transfer data throughout your motherboard to process data. But can deal with pre-processed data!
Current challenges
Optical computers seem to be a far-fetched device. Firstly, while playing with wavelengths of light, we might run into problems. The wavelength of input waves might not be equal to the output waves due to loss. Thus, inappropriate wavelength as output will incur problems in the next transistor.
Also, efficiency and accuracy need a supercooled environmental condition. Those can be achieved in a laboratory and not within your current CPU cabinet (or tower). Also:
In Japan, the NTT company is building a huge optical computer that encloses five kilometres of fibre in a box the size of a room and will be applied to complicated power or communications networks enhancement tasks.
Hence, that won’t be portable at all!
So, it’s hard to create an optical computer. But, research and development are still on. Companies like Optalysys are designing an optical co-processor that can work on heterogeneous system architecture. So, maybe not a 100% optical computer, but a hybrid optoelectronic system might elevate the current computation standards. That might also take us a bit closer to perfection.
References:
- Cover photo by Alex Gorin on Unsplash