Home
Overview
History
Market
Future
References
Semantics
History
The first supercomputer was Control Data Corporation 6600, CDC6600 for short, which was comprised of a single CPU and was released around 1964. The cost of it was roughly 8 million dollars and in today’s time rounds to about 60 million dollars. The size was small and it operated at 3 million floating points per second.
The second supercomputer to be created was the Cray 1, which was released in 1976. The primary task was to model nuclear weapons. The computing power of the Cray1 was fueled by parts that were its integrated circuits and it increased to 64 bit, a higher word size. It allows Cray1 to operate at a 136 megaflops which was a huge difference compared to the CDC6600 which only had 3 megaflops. It had a unique shape giving it the ability to operate at higher speeds. It was able to sell on the market around five to eight million dollars, rounding around 25 million dollars in today’s currency.
The third supercomputer was The Cray X-MP, which was released in 1982. The new supercomputer has the capability to hold four CPUs but was still housed in the same type of chassis as the Cray1. The CPUs were clocked at a higher speed, an increase from 80 to 105MHz and had increased bandwidth memory by more than double allowing for performance of 200 megaflops per CPU. Speeds of 800 megaflops for the Cray X-MP ran for 15 million dollars, the equivalent of 32 million dollars. It was able to support 16 million 64-bit words of memory; although it sounds like a lot, it is only 128MB of data, which is about 100 pictures worth of data.
The fourth supercomputer was the Cray 2 and was released 1985. The chassis was the same horse shoe shape but with a little size change. Logic boards were tightly packed close together to improve performance and that made the Freon cooling system useless, so they submerged the whole supercomputer in Fluorient. The CPU was increased to eight and was the first supercomputer to run mainstream software. Performance was measuring 1.9 gigaflops giving it the title of the world’s fastest computer until the year 1990. These supercomputers were the first to be used by more than just the US government agencies, found in universities and corporations.
After 20 years of dominance by America the next supercomputer was from Japan known as Hitachi SR2201. Holding a whopping 2048 processor and performance report score 600 gigaflops. Cray went bankrupt around this time because they tried to make Cray 4 but the research lived on. Intel beings its involvement with supercomputers when it created and released the i860, a 32-bit and 64-bit RISC chip, which lead to the creation of the Intel Paragon which also lead the making of ASCI Red around 1996. It was the first supercomputer to be built with off the shelf CPUs, rocking 6000 200MHz CPUs and costing 46 million dollars. It made it the first supercomputer to break the one teraflop barrier and when they upgraded the CPUs with 9,298 Pentium II Xeons reaching 3.1 teraflops, making it the fastest supercomputer for the next four years. Japan retook the crown in 2002 with their NEC Earth Simulator and the price of only 900 million dollars pulling 35 teraflops of power. In the year 2004 IBM released the Blue Gene/L which had 16,000 compute nodes and performing 70 teraflops and after upgrades throughout the years in 2007 it peaked at 600 teraflops and costing a large amount of money that is unknown. The seat was taken over by the IBM Roadrunner running through the 1 petaflop barrier. China came in the picture around 2010 with its Tianhe-1A, performing at 2.5 petaflops. Around this time china shows it built their own homegrown supercomputer running at 1 petaflop called Sunway. Japan is running the show with its 10 petaflop supercomputer K, it is the most expensive supercomputer priced at $1.25 billion. We started with a supercomputer that could perform at 3 megaflops and after 48 years we have increased our performance by 3.3 billion times leaving us at 10 petaflops.