Op-Ed: Big Blue turns Big Green, and backs it up with light-based processors

IBM announced earlier this year its “Big Green” initiative, a billion dollar a year project designed to produces less energy-hungry computers. Now they’ve come up with something 100 times faster, with hundreds of thousands of processors on a chip.

Infoworld.com provides the background to “Big Green”: “Large companies are facing a large crisis around energy," says Mike Daniels, senior vice president of IBM Global Technology Services. "This issue is surfacing in a number of different ways, whether it's the capital required to build new data centers because people are out of capacity, or where people are out of power or trying to manage that power. We think this is an issue where we can provide leadership for the industry and for our clients."

IBM claims that the savings for customers are great in going green -- by using IBM technologies such as their blade servers instead of other vendor's technologies, a customer with a 25,000-square-foot data center should be able to save as much as 42 percent on energy consumption.” No doubt about IBM, they don’t think small scale. The Infoworld article also mentions in passing that IBM is redesigning 8 million square feet of data centers worldwide. The new processors, however, are the real deal. They work by converting electrical signals into light pulses. Not too difficult a concept, but the ramifications are huge. It means getting rid of the miles of copper wire, and other conductors, and that’s a huge jump ahead. Meaning your 2Ghz processor would be 200Ghz. That speed is faster than any electronic media on Earth needs to be. The ideas of picture quality, download speeds and other esoterica would vanish. Even displays would be better, and healthier. Current computers work slightly faster than the human eye, which means they blink, but we can’t see them, except on camera, because they blink so fast. Now, they won’t even blink. Photo-based processors have been a holy grail of the industry for some time. But the demands have been rising while the research progresses, and the industry has been seeing some very large problems coming. The sheer volume of data is the problem, and the internet needs all the ability to handle those loads it can get. There are some big capacity issues, as well as power consumption and chip capabilities. Given the decade long drone about the limitations of silicon, temperature issues, and the basic need for more processor power, the electro-optic processors look like they can deliver. This is where all of the major limitations with silicon chips can be addressed. A method which can simultaneously deal with all the hardware problems as well as the power usage is exactly what’s needed. The possibility, because of the much smaller size of the electro-optic processors, is that a supercomputer will now fit on a laptop. IBM’s lead scientist, Will Green, has obviously got his eye on the major issues. Quote from Electronic Design.com: “"Just like fiber optic networks have enabled the rapid expansion of the Internet by enabling users to exchange huge amounts of data from anywhere in the world, IBM's technology is bringing similar capabilities to the computer chip," Green said.” A bit of perspective here. In the last 20 years the human race has gone from paper moving at manual speeds to Gigahertz at light speed. NASA has had to use people’s home computers to process data, because all its mainframes combined can’t handle the data load. Most of the sciences are generating similar amounts of data. Business and government are entirely processor-dependent. Power usage is no joke, either. The United States, with its computers in standby mode, uses as much electricity as the whole of Italy. It's already extremely uneconomic. Between computers and computer dependent equipment, it's positively extravagant. This is an emerald green idea. It will also perhaps finally explain what's meant by energy efficiency as a pretty dramatic cost benefit to the cave dwellers still thinking in kilowatt hours. That idea hasn't really got through since the 70s. With this concept comes some good definition of how much better energy can be used. This idea can also completely rework the internet, and vastly increase its capacity. Processors are about as complex as computers get. What computers and related hardware can do depends on what they can do. The systems which support them have to be able to handle their information, in whatever form it happens. So the net will work a lot better, and much faster, if it adopts the same processors. This is heading to instantaneous transmission, without the vast amounts of infrastructure. A cell phone could do it. The less the demand on system capacity, the less “handling” of vast amounts of data, the better. Any network is weakened if it has to do an assembly line approach to handling data and gets too much data at once. It might even provide better protection against saturation attacks, because it will have the capacity to do a lot more defensive work, rerouting nuisances to oblivion. Put another way, the present version of the net will be like riding a mule. It’s good news for ISPs, too, because this sort of technology is stackable, as far as I can see. You could have the equivalent of a thousand supercomputers in your laundry. Anything goes wrong, you just stick in another chip or so, doing in seconds what could take weeks of downtime and reduced capacity. Costs would go through the floor. What is now very tough IT could be work experience. You could have a thousand different operating systems on one computer. (Might actually need them, too, the way science is going, let alone media.) IBM aren’t expecting to get this going for a decade, but nobody was expecting much from the last decade, either. The market will lap it up. It's a basically simple idea, and it can cover a lot of manufacturing issues at source. Things like sound cards, TV cards, video cards, and a range of mystic odds and ends can all fit straight onto processors, if I'm reading this right. If the outputs use the same methods, back into electrical impulses, all those cards do is copy data. You wouldn't need separate hardware. Fans, temperature controls, would be things of the past, and CMOS batteries would last for decades with much lower power consumption. These things could wind up a lot more cost effective, and actually cheaper. Use solid state, and you no longer need a disk operated system, either. One extremely interesting thing occurs to me: The ability to process micro electrical impulses like that suggests a lot of applications for neurology and the human/computer interfaces which have been getting so much space lately on science news. If these things can do that with micro volts, the rest of it can happen. Artificial retinas, chip-enhanced vision, artificial synapses, there’s a lot of medicine and biology that could come along for the ride, too. Why not cure diseases with a chip in the genes? Diagnosis and case management with a microchip sending routine pathology reports to your doctor? If you can run enough software, it's doable. Dunno about "brave" new world. I think it's as scared as it should be of all the new things. But every once in a while, something that really works comes along, and everybody goes with it.