In the last entry in this blog, when talking about the CELL processor referring to it as a supercomputer on a chip. This comparison is not free, and does not refer to how quickly the microprocessor will (get it "faster" as most megahertz), but the chip architecture itself, as the CELL processor has a characteristic of supercomputers to be a vector processor.
is when I began to remember old stories of ancient heroes and forgotten and I thought of a hardware designer whose name has always been synonymous with raw power. I refer, of course, Seymour Cray.
tell the story of Seymour would take to many pills and even an entire book (in fact, already have it, although not exclusively dedicated to him), but I am going to focus on a specific project that will help up again as the designer of the world's fastest computer, the Cray-1.
To speak of Cray-1 would first have to talk a little background. In the 60's, IBM was the Almighty and all of the computer world, and everyone else, however great they were, they only danced to Big Blue played. However, one company decided to dispute the leadership at IBM in the field of scientific computers, ie computers specializing in the treatment of complex mathematical operations on a relatively limited set of data.
Thus, the CDC 1604 appeared first and then the jewel in the crown, the CDC 6600, considered the first supercomputer in history and which was at least an order of magnitude faster to any other computer at the time. Famous is the comment of Thomas J. Watson Jr., at that time president of IBM, saying that it was possible for a small company of 34 people could beat them when they were thousands of people. And the answer is also famous chief engineer who designed the CDC 6600, saying that he had answered his own question. The name of this engineer is Seymour Cray. As counter the CDC 6600, IBM announced a new model of family System/360 mainframes, namely the Model 92, which promised to be at least as fast as the CDC 6600, but with all that meant to be an IBM computer. This caused many customers will think before buying a CDC 6600 and wait to see what drew him to compete with IBM. The weather started to happen but that case Model 92 never hit the market, so it sued IBM Data Control for a clear case of what today is known as vaporware .
Against all odds, CDC beat IBM in court, but perhaps this success they went a little to the head and the company began to change course to stop being a company that manufactures and sells computers to be passed one that manufactures and sells solutions, ie Thereafter design, and sell new computers fabtricar was only part of the business that was complemented with printers, terminals, input and output systems, software, etc.
Unfortunately for CDC, this meant that the company should split its always limited resources among more and more projects, and to top it in 1969 not only was funding the development of the CDC 8600, which would be a multiprocessor computer (remember we're talking about 1969, there were no such things outside the realm of theory), but also the design of a long-time collaborator and assistant to Cray, the STAR-100. So in the year mentioned, 1969, CDC management met with Seymour Cray for this would cut costs by 10% (ie, to fire 10% of its workforce). Instead, what he did was he reduced the salary to collect the minimum required by law ($ 1.25 per hour) and save the project.
But his sacrifice was in vain. The CDC 8600 was a machine so complex that the problems seemed to pile up, and perhaps one of the largest was chilling. Cray needed more funds to start the project again, as was done with the CDC 6600 and gave such good results. However, the direction of CDC saw this move as too risky not authorized, which meant that Seymour Cray decided to leave the Control Data Corporation and found his own company, Cray Research end up calling . Once they had sufficient funds (some of which were obtained with lectures and exhibitions Tupperware-style home is Seymour founded the company and began to work again in the construction of a new supercomputer.
In Instead of following the line of development of multiprocessor CDC 8600, Seymour decided to stick with the traditional idea of \u200b\u200ba single CPU for its first project in his new venture. So instead of having multiple processors working independently each other, decided to construct a vector processor, which means, simplifying a little, you have a set or vector (and hence its name) math coprocessors, which could be loaded with different data and all run the same instruction at a time. While this technology is not much in the world of mainframes "office" where you work with lots of data that are given a very limited set of operations for math-intensive applications and scientific applications (or image processing and video-or three-dimensional video games), this technology allowed greatly accelerate performance.
However, the idea of \u200b\u200bvector processor was not really his, but was inspired by her competitor in CDC. Indeed, the CDC STAR-100 was designed as a vector computer. Notwithstanding the differences between the two computers were palpable. Since we do not pretend to give lessons in architecture of processors (not think I have enough knowledge to do it), say that although the idea is copied by Seymour, that develops its own style using their own techniques and conclusions so that the final machine had substantial differences at the structural and performance.
Finally, in 1975, went on sale the Cray-1. In his own original had a 64 bits at 80 MHz and was capable of directing the equivalent of 8 megabytes of RAM. However, at a glance what most stood out was the design of it, he was in a horseshoe shape giving it an extremely advanced science fiction. The cooling system, true workhorse of supercomputers, was based on the freon gas, something completely new at that time. With all this, the first Cray-1 model weighed the negligible amount of 5.5 tons. Initially, Cray Research had to sell a dozen supercomputers (think very specific market that have these machines), but requests for several years of acquisition of Cray-1 was piled reaching more than 80 computers sold, at a price of $ 8.8 million in the second half of the 70.
This success catapulted to fame in the Seymour Cray, who was already working on the Cray-2 and did not have much time for "social" or promotional Cray-1. But yes I would tell a story to understand how much respect was Seymour Cray. Despite
little as it was when appearing in public, Cray attended as a speaker at a lecture in 1976 to the developers of the National Center for Atmospheric Research in Colorado, USA. Finally, when it came question time, the room fell to complete silence. Seymour waited several minutes waiting for someone to do a question, but nobody said anything. Finally, when he left, the head of the computer division of the place asked the audience pretty angry with how difficult it had been getting Cray was up there how is that no one had raised his hand to ask. After a tense moment, one participant replied: "I would talk to God?". Regardless of the future evolution of events, successes and failures of Seymour Cray, this story reflects very well the genius of an engineer whose dream was always to build the fastest computer in the world, the Ferrari of computer ...
0 comments:
Post a Comment