You need only look in your pocket to find evidence of how far computers have come. From the monolithic vacuum tube mainframes of yesteryear to the dizzyingly complex smartphones we now take for granted, the evolution of our computational devices is as fascinating as it is rapid. But the technology wasn’t dropped into our laps overnight, it’s taken a great deal of iteration and the odd breakthrough to reach this point. As it is with most things, sometimes a look back is important to keep things in perspective.
We previously discussed one of the first commercially available computers, the UNIVAC-1, which was a computer that was so massive, that it might be what springs to mind when you think of the term “supercomputer.” The truth is, the reason for UNIVAC’s status as a space hog has nothing to do with its power and complexity, but from the fact that semi-conductors weren’t available at the time, so the systems required huge banks of vacuum tubes to direct their electronic traffic.
Supercomputers are big machines, but all the floor space is packed with power. Used for demanding general computations like molecular modelling, quantum mechanics, and weather forecasting, a supercomputer is built to be substantially faster than even the fastest of commercially available servers or desktops. Due to their usage in scientific fields, supercomputer effectiveness is typically measured in floating point operations per second(FLOPS). The higher the flop, the faster the supercomputer.
Modern supercomputers, while impressive, aren’t too imaginative: huge banks of computer shards running hundreds of thousands of processors in tandem to achieve speeds of several quadrillion FLOPS. Back in the 1960’s and ‘70’s, on the other hand, designers of early supercomputers had to get creative when managing issues like efficient usage of space and dispersal of heat. No one was better at the game than the godfather of Supercomputers, Seymour Cray.
THE CRAY EMPIRE
One cannot delve into the history of supercomputing without making significant reference Seymour Cray. The dude’s a legend who, in a perfect world, would be mentioned with the same reverence reserved for the likes of Steve Jobs and Bill Gates. He was a World War 2 radio operator, seeing action in both the Pacific and European theatres of operation. When he got home, he went to school and picked up degrees in both electrical engineering and applied mathematics. From there, he joined the Engineering Research Associates (ERA) where he remained, even after it was bought out by Remington Rand and folded into their UNIVAC division.
In the 1960’s, he joined a number of other ex-ERA employees in forming the Control Data Corporation (CDC). This new company formed around the time that the transistor hit the market. The CDC initially started into the market building average computers with average aims at crunching business and commercial data, but it wasn’t long before Cray set his eyes on designing the fastest computer in the world. He set his goals at lofty heights – he wasn’t content with just building the next iteration of the computer, he wanted to build something that would outpace absolutely everything by a wide margin.
This took the form of the CDC 6600, the world’s first commercial supercomputer. It was a pricy investment that largely kept it out of the private market, but for those with big needs for number crunching, there was nothing that came even close. It was tremendously disruptive. Even massive corporations like IBM struggled to produce anything that even came close.
Cray continued to design systems for CDC until the company could no longer keep up with his vision. After a fairly amicable split, Cray founded Cray Research, and with the help of some Wall Street seed money, went to work creating the Cray-1.
THE SUPERCOMPUTER GRAND-DADDY
Released in 1975, the Cray-1 ran at 80MHz and could crunch around 160-250 Mflops of data. To put that into perspective, the original iPhone pushed about 300Mflops in 2007. That may not sound like much against our modern, spaceman technology, but back in the day, this provided the Cray-1 with the title of World’s Most Powerful Supercomputer from 1976-1982, when the Cray X-MP came along with its 713Mflops of power.
That’s nothing, though. In 1985, the Cray-2 was released with a top performance of 1.9Gflops. To put that into frame, in 2010, the iPhone 4 was somewhere around 1.6GFlops. Its reign as top dog didn’t last as long as it’s the Cray-1 (only lasting from 1985-1987), but it’s still a highly regarded machine in the history of supercomputers. It helps that the Cray-2 simply looks cool, with its bubbling Fluorinert cooling system that cascades over a transparent waterfall unit. Sleek gear and rows of cabinets stacked with server shards may still get a technician’s blood pumping, but there’s a certain romance to the clunky-yet-futuristic designs of yesteryear.
THE END OF AN ERA
As the 1990’s rolled around, supercomputer design began to shift more and more towards massively parallel computers which functioned by roping a growing multitude of processors together to achieve incredible speeds at lower cost. Cray was struggling to keep up, with Seymour Cray personally resisting this market shift. He began experimenting with novel designs that often didn’t quite pan out, all under the philosophy that a single very fast processor would always outpace a cluster of slower ones. Tragically, we’ll never find out if he was correct in that theory, as he died in 1996 following a traffic accident.
Cray Research continues to make supercomputers that are fast enough to crack the top 500(November 2016’s list has the Cray XK7 in the number 3 spot). The market is in a much different place than it was in Seymour Cray’s day, but the need for speed remains the same and shows no sign of vanishing. These days, the supercomputers of the 70’s and 80’s can now fit nicely into our pockets, but the newest machines are still staggeringly powerful, pushing up to 93 quadrillion FLOPS and consisting of millions of parallel CPU’s. Gosh. I certainly wish one of those would pass through Just Fix It’s help desk. Since we don’t have a quantum physicist on the team, I’m not sure what we’d use it for, but I bet it can run spreadsheets like a dream.