Then:
The computers that we have come to rely so heavily on in our current day and age have come a long way from their ancestors. Back then, it was not out of the ordinary for one of the first computers to take up roughly half of the room they were in, if not more. Now, computers are small enough to fit in our laps and, in the case of our cellular phones, our hands. The phrase “bigger means better” can apply to a wide variety of objects and concepts, but computers are not on that list. If someone were to take a program or video game from today’s computer, run it on the aforementioned room-sized computer, and by some miracle the computer figured out how to run it, it is very likely that it would seize up like a pair of rusted gears. One may ask why that is, to which I respond with two words: processing power. Over the years, we have invented increasingly more efficient processors for our computers, which operate more quickly with each new variant while taking up less space. There was even an observation made, dubbed Moore’s Law, claiming that processing power would double every couple years. Predictably, this process ended up running into a problem: there is a physical limit to how many transistors we can fit onto a silicon chip.
 |
| ENIAC, the world's first computer. |
The solution to this problem is actually rather easy: instead of trying to fit billions of microscopic transistors onto a lone chip in an effort to make it more powerful, we can instead string together several individual processors that will add up to an even higher amount of processing power (“Where will future computer”, 2004, Sep 29). A mental image would probably help clarify what I mean. Imagine a sack of flour with its contents slowly leaking out of a small hole in its burlap. The sack of flour as a whole is the computer, the flour spilling through the hole is the binary code that is being processed by the microprocessor, and the hole itself is the microprocessor. No matter how efficient the microprocessor becomes at processing binary code, it can only do it so quickly. If one creates more holes in the sack, or make use of multiple microprocessors, the task of spilling the flour, or interpreting the binary code, will speed up as the work load is spread out and ultimately shortened by the additional holes. Back in 2004, scientists doubted whether or not parallel processing would ever become a viable strategy (“Where will future computer”, 2004, Sep 29), but the existence of dual core processors today, including the Intel Core i5-5200U Processor in my laptop, proves that such a feat was possible.
Now:
Even with the addition of parallel processing to our arsenal of processing power, it still was not enough for us. What we needed was a more substantial boost to processing power that did not rely on parallel processing. While parallel processing works, it does have some drawbacks when used by itself: the act of splitting up the processing and inefficient usage of space. In the case of the former, dividing the processing load between several points could very well become complicated as the number of cores increases, though this is admittedly a minor drawback at most. For the latter, it would be much more simple and efficient to use a single processor than using several individual processors arranged side by side; with some of the computers we have today, every bit of space counts. It thankfully did not take too long for us to discover another breakthrough in processing power to replace parallel processing: stacking.
 |
| An example of how stacking works with through-silicon vias. |
For the longest, we have stuck with using two-dimensional processors. Consisting of a silicon chip, long lengths of wiring running all along the chip’s surface, and countless transistors connected by said wiring (Savvas, 2007), they were missing out on the blessing of three-dimensional space. Think of a desk occupied by paper sheets organized in a haphazard pile. One could line the sheets of paper up side by side, eventually filling up the desk’s space with a single layer of paper, or one could just take the sheets of paper and arrange them into a neat stack. The latter method applies to processor chips; not only is space saved by layering the chips on top of one another, but information passing from one chip to the other has less distance to travel, ultimately speeding up the process. Of course, it is not as simple as building one chip on top of the other one, but this is where the main component of the breakthrough shines: through-silicon vias. These vias are minuscule holes etched through the silicon of the chips that are then filled with conductive metal. Information passes from chip to chip using the through-silicon vias, allowing for chips to be stacked on top of each other for massive benefits. Specifically, the numbers show that this technique cuts down on the travel distance of binary code by up to 1,000 times and makes it possible for up to 100 more pathways for data to travel along to be utilized all in comparison to the traditional two-dimensional processor chip. Some of the first applications of through-silicon via technology included power amplifiers for wireless services, high performance servers, and supercomputers (Savvas, 2007).
Later:
Several revolutionary advances have been made in the art of processing over the years; one of the first methods of boosting processing power, parallel processing, made use of multiple chips to divide tasks up into smaller chunks, and that method was later built upon with the use of through-silicon vias to provide data with both less distance and more paths through which to travel. Regardless, we still find ourselves craving for more and more speed. Banks in particular have shown themselves to be utterly insatiable in their demands for ever quicker processing. To give a rough idea of this hunger’s extent, Royal Bank of Scotland employee Barry Childe once increased the computing power of one of the bank’s systems by 40 times only to be asked by his boss, “Great, when can I get 400 times?” (Barnes, 2006). It has gotten to the point where wait times as minuscule as a single nanosecond, a span of time that is virtually impossible for a human being to detect, could potentially cost a bank a deal (Barnes, 2006). How exactly are we supposed to make processors that can interpret vast amounts of data in under the smallest fraction of a second?
 |
| Another electronic-photonic processor chip, this time from University of Colorado. |
It may sound crazy, but scientists have been theorizing that we need to forgo wiring almost entirely in favor of light. Using bursts of light costs negligible energy, produces almost no heat, and has the potential of becoming blindingly fast in comparison to navigating electric currents through solid metal wires. These bursts are planned to be emitted with the use of “photonic crystals,” which in simpler terms could be compared to sending someone a message in Morse code using a flashlight. The article also gave a handy comparison to go along with the flashlight example: with how superior the speed of light processing is expected to be, our current method of processing with wires will be like manually writing out and sending a letter. Will we ever see the usage of light processing in our lifetime? At the time of this article’s publication, this technology was in the laboratory stage of development and was getting substantial results. For example, University of Texas’s electrical engineering professor, Ray Chen, managed to successfully build a processor chip that made use of the aforementioned photonic crystals and pulses of light (Barnes, 2006). While it was not elaborated upon how quick or accurate the chip was, I can personally see it coming to fruition eventually, though I worry that the technology might stay anchored in industrial use instead of ever being exposed to the consumer.
Home
Next ->
No comments:
Post a Comment