History of Supercomputing
Modern supercomputers are giving scientists the tools they need to improve our health care, tackle climate change, and reach farther into the recesses of deep space.
As the name implies, a supercomputer is a computer that’s capable of performing extremely large and complex calculations at blazing speeds.
A supercomputer can perform highly advanced calculations that typical computers cannot.
Let’s speculate on the future of supercomputing by tracing its history:
Just how super are supercomputers?
Supercomputer speeds are measured in FLOPS: floating point operations per second.
Basically, these are calculations with numbers so massive, they have to be expressed in exponents: 70 x 15010.
- MegaFLOPS: 1 million FLOPS
- GigaFLOPS: 1 billion FLOPS
- TeraFLOPS: 1 trillion FLOPS
- PetraFLOPS: 1 quadrillion FLOPS
- ExaFLOPS: 1 quintillion FLOPS
The early days of supercomputing
In the 1950s, several emerging tech companies compete to build the fastest computer.
IBM is the industry leader with its super-fast IBM 7030 Stretch.
- A group of computer engineers formed Control Data Corporation (CDC) in Minneapolis, Minnesota – one of them was a brilliant computer scientist named Seymour Cray
- At CDC, Cray’s team completes the CDC 6600 computer
- The CDC 6600 becomes the world’s fastest computer, 3x faster than the IBM 7030 Stretch
- The computer is so fast that it’s dubbed a supercomputer
- The CDC 6600
- 400,000 transistors
- Over 100 miles of wiring
- Innovative cooling system
- Can perform calculations at speeds up to 3 megaFLOPS - slow by modern standards, but revolutionary for its time
The Cray Revolution
Cray’s computer designs at CDC were impressive, but expensive. Frustrated by corporate oversight, Cray left CDC to form his own company, Cray Research.
Cray Research dominates the supercomputer market during the ‘70s and ‘80s.
- The Cray-1
- Released in 1976
- One of the first supercomputers to use integrated circuits
- Performs calculations as fast as 160 megaFLOPS
- Notable for its “C” shape, which improved the efficiency of the hardware
- The Cray X-MP
- Released in 1982
- Added 4 processors and greater memory bandwidth
- Performs calculations as fast as 800 megaFLOPS
- The Cray-2
- Released in 1985
- First supercomputer to use liquid cooling
- Performs calculations as fast as 1.9 gigaFLOPS
The end of the Cray Era
Other tech companies sought to make supercomputers more affordable by developing massively parallel processing (MPP). MPP works by harnessing thousands of side-by-side processors that work in conjunction with a single memory unit.
Cray resisted the idea, but it was more successful than his vector-based supercomputers – Cray Research ultimately went bankrupt in 1995.
New age of supercomputing
- 2 NASA contractors, Don Becker and Thomas Sterling, built the Beowulf supercomputer from off-the-shelf computer parts
- Beowulf innovated the cluster model: a supercomputer that’s made from a cluster of computer units working together
- Intel used the cluster model to develop the first 1 teraflop supercomputer: the ASCI Red
Most modern supercomputers use the cluster model, though now they’re designed to be more compact.
Modern supercomputers use both CPUs and GPUS working in tandem to perform calculations.
In fact, some NVIDIA GPUs in HP OMEN gaming computers can perform some supercomputer-caliber calculations.
- IBM completes the fastest supercomputer in the world: the Summit.
- The Summit can perform calculations as fast as 200 petaFLOPS
- Some types of calculations can be performed as quickly as 3.3 exaFLOPS
Also in 2019, Cray was acquired by Hewlett Packard Enterprise.
Hewlett Packard Enterprise and Cray will work to build supercomputers that can routinely reach exascale speeds.
Future of supercomputing
Modern supercomputers have barely just breached the exascale, so expect supercomputer manufacturers to develop new technologies that will try and push supercomputing up to zettascale speeds (that’s a whopping sextillion FLOPS or more!)
The next wave of exascale supercomputers can help us make advancements in:
- Space exploration
Research in these fields requires calculations that have a massive number of data points, so only high-tier supercomputers can make those computations. As supercomputing technology improves, so will our world.
 HewlettPackardEnterprise.com; A super-fast history of supercomputers: From the CDC 6600 to the Sunway TaihuLight
 TheVerge.com; Hewlett Packard Enterprise acquires supercomputer maker Cray in the race to ‘exascale’ performance
 Engineering.com; INFOGRAPHIC: What’s the Future of Supercomputing?
 KnowledgeBase.IndianaUniversity.edu; Understand measures of supercomputer performance and storage system capacity