Friday, June 19, 2015
Stanford University’s Inverse Design Algorithm for AI designed Fiber Optic Motherboard
This article is somewhat personal to me, as I'm currently using a Dell Lattitude D520 Laptop that gets pretty hot. I know I have to eventually pop her open and clean the fan as explained in my Geezam blog article “Tips and Tricks on how to make your computer run faster” but I'm somewhat of a procrastinator.
So when I read about a research team from Stanford University that has figured out a way to miniaturize Fiber Optic Cables to be used inside of a Laptop as reported in the article
“Breakthrough brings optical Data transport closer to replacing wires”, published May 28 2015, Physorg, I was beside myself with excitement.
Their research, led by Stanford graduate student Alexander Piggott, was published in Nature Photonics Journal. Again, a lot of PhD’s riding on this wagon, so here's yet another long list:
1. Stanford electrical engineer Dr. Jelena Vuckovic
2. Graduate student and Google employee Jesse Lu
3. Graduate student Jan Petykiewicz
4. Postdoctoral Stanford student Thomas Babinec
5. Postdoctoral Stanford student Konstantinos Lagoudakis
What their research yielded was a method of creating Fiber Optic Data circuits using silica glass to replace the copper and gold wires used in the making of motherboards in laptops and computers. Because they are made from silica glass and not wires, they are not prone to the heating effect normally seen in your traditional computer.
Best of all, silica glass naturally refracts IR (infrared) light, the same frequency of light that is used by Fiber Optic Data Transmission systems. Thus, if you could replace the Bus on a computer motherboard, you could eliminate the need to do optical-to-electrical conversion, at least until you reach the Processor.
So how did the Stanford research team shrink the tried and trusted Fiber Optics system that is the basis of modern telecommunications to fit into a computer? And can this potentially be the
Optical computer I've long dreamed in my blog article entitled “IBM develops 25Gbps Photonic Optical Processor at the 90nm level - IBM's Red Dawn (2012) for Optical Processors”?
Stanford University Researcher develop Fiber Optic Motherboard Bus – Cool Running’s for Computers
The research conducted by the team was actually the continuation of previous work done by Professor Jelena Vuckovic with Professor David Miller years prior.
Acting as a guide to Stanford graduate student Alexander Piggott, her previous research foray into computers revealed that as much as 80% of the power wastage in terms of heat loss was not from the Processor but from the motherboards, quote: “Several years ago, my colleague David Miller carefully analyzed power consumption in computers, and the results were striking. Up to 80 percent of the microprocessor power is consumed by sending Data over the wires so called interconnects”.
“Wow” was my first reaction reading article, as I'd always assumed it was the Processor drinking up all that power.
That idea has been reinforced while at MICO College University while doing my Diploma in Professional Studies, as I often have to pop over into the Computer Lab at the e-Learning or at the MICO University Library. Whenever I use a computer, I notice that the computers make an awful lot of noise, almost like an airplane about to take off.
I've always assumed it was the Processor cranking up and needing all that cooling to keep it functioning, as I've had similar problem with my Gaming rig back home in Milk River, Clarendon, which is also noisy despite my best attempts at optimization. However, my computer has a steady fan noise, not an ever increasing hum akin to an airplane taking off as I’d observer with computers at MICO College University, as the fan in my personal PC is rather large.
Up until now I thought the solution lied in improving the cooling technology used in computers using liquid cooling techniques, such as UAH's (University of Alabama in Huntsville)
Fluorinert FC-72 Passive Cooling System as described in my blog article entitled “UAH Graduate Students use 3M's Fluorinert FC-72 in Passive Cooling System – Gaming Rigs and Data Centers Noiseless Cooling Systemss upgrade”.
More exotic solutions such as using CNT’s (Carbon Nanotubes) and room temperature superconductors made from Group 4 Elements as described in my blog article entitled “@UTAustin at Austin develops Silicene Transistors - How to grow Silicene and Group 4 Super-conducting Processors and Batteries on Silver Spoon”.
But they’d still require cooling. A computer Motherboard with the Data Bus based on Fiber Optics, on the other hand, is inherently cool running as it’s not electricity heating up copper and gold connectors and traces but IR (Infrared Light) in silica glass fibers!
Stanford University’s Inverse Design Algorithm – Microscopic Fiber Optic components for Optical Computer
Their idea was to replace those copper and gold connectors and traces used in the Motherboard's Bus to interconnect the various components that the Processor communicates with silica glass based Fiber Optic cables. Most likely, these electrical circuit components that process Data i.e. Processors, Memory, Hard-drive, would sit in or connect via a special socket that does the optical-to-electrical conversion.
This would allow the TTL (Transistor-Transistor Logic) and MosFet based processors and components to work, with capacitors, resistors and inductors still needing basic electrical connections.
By replacing the Motherboard Data Bus with a Fiber Optic Network, it makes the computer run cooler and connect to Fiber Optic Internet the need for optical-to-electrical conversion. A Fiber Optic Motherboard Bus also allows you to multiplex more Data into fewer Data pathways, making such a computer more efficient and powerful to quote Stanford graduate student Alexander Piggott: “Optical transport uses far less energy than sending electrons through wires. For chipscale links, light can carry more than 20 times as much Data”.
To create these Fiber Optic circuits to replace the Motherboard Data Bus, they created a computer program which they dubbed the Inverse Design Algorithm. It's basically like a PCB (Printed Circuit Board) software program, similar to those you can find online.
But their program is more advanced, as it takes into consideration the properties of silica glass as it relates to its ability to refract IR light to design circuit analogs to switches, multiplexers and amplifiers, but specifically geared at manipulating IR light.
Interestingly too, the circuit it designs are microscopic in scale, with circuits so small that twenty (20) of them would make up the width of a human hair.
The PCB diagrams made by the Inverse Design Algorithm are then etched out using their laboratory fabrication facility, albeit not to a high level of precision to quote Stanford graduate student Alexander Piggott: “Our manufacturing processes are not nearly as precise as those at commercial fabrication plants. The fact that we could build devices this robust on our equipment tells us that this technology will be easy to massproduce at stateoftheart facilities”.
Because these structures, made of silica are to be connected to silica glass traces, most likely they'll be encased in epoxy or some other type of cladding like traditional Fiber Optic cables used in Telecoms so as to reduce refractive losses.
Optical Computer with CNT Processors – Fanless Computers designed by Artificial Intelligence
The designs that the Inverse Design Algorithm comes up with are beautifully and visually stunning, but are geared towards creating circuits equivalent in function to their analog counterparts by manipulating the refraction of IR light, to quote Dr. Jelena Vuckovic: “Our structures look like Swiss cheese but they work better than anything we've seen before”.
So it won't be too long before AI (Artificial Intelligent) programs are cooking our food too as predicted in my blog article entitled “US$15,000 Moley Robotics Cooking Robot – Cooking Robot seeking Taste for Human Food to take over in 2017”.
Hopefully, this all means that the CNT (Carbon Nanbotube) Processor that Standard University has been working on as far back a September 2013 as described in my blog article entitled “University of Stanford Designs a Proof-of-concept Processor using Carbon Nanotubes - Practical option to expand Moore’s Law along with Optical, Quantum and Neural Net Processors” now finally has some optical Pathways to efficiently process Data and interface directly with the Fiber Optic Communications standards used in Telecommunications today.
The potential of this breakthrough Inverse Design Algorithm is huge, as it can be used to design all types of Optical communications systems based on Fiber Optics. Not only does it eliminates the need to do optical-to-electrical conversion to connect to Fiber Optic based Internet, at least until you reach the Processor, but it also makes the computer run cooler, more efficient and potentially faster!
Inverse Design Algorithm ties in automation into the design of circuits, which made the manufacturing of telecommunications gear by Telecom Equiptment suppliers achieve economy-of-scale much faster by giving the work to machines instead of humans. If this is how Optical fanless computers will come to pass, then I’m all for it!