Supercomputers

Read the following words and word combinations and use them for understanding and translation of the text:

successive - последующий, преемственный

at a clock speed - с тактовой частотой

to take advantage of - воспользоваться

entire - целый, полный

the ultimate - в конечном итоге

formerly - раньше

to soak up - впитывать, поглощать

protein - белок

cost-effective - рентабельный

distributed - распределенный

to feature - показывать, изображать

synergistic processing element - ядро специального назначения

ad hoc - специальный, на данный случай

to parcel out - выделять, делить на части

extraterrestrial - внеземной

proximity - близость

mesh - сетка, ячейка

to rank - классифицировать

high-performance computing - высокопроизводительные вычисления

benchmark - отметка, стандарт, эталонный тест

to retain - сохранять. удерживать

The term supercomputer is not really an absolute term describing a unique type of computer. Rather it has been used through successive generations of computer design to describe the fastest, most powerful computers available at a given time. However, what makes these machines the fastest is usually their adoption of a new technology or computer architecture that later finds its way into standard computers.

The first supercomputer is generally considered to be the Control Data CDC 6600, designed by Seymour Cray in 1964. The speed of this machine came from its use of the new faster silicon (rather than germanium) transistors and its ability to run at a clock speed of 10 MHz (a speed that would be achieved by personal computers by the middle 1980s).

Cray then left CDC to form Cray Research. He designed the Cray I in 1976, the first of a highly successful series of supercomputers. The Cray I took advantage of a new technology: integrated circuits, and new architecture: vector processing, in which a single instruction can be applied to an entire series of numbers simultaneously. This innovation marked the use of parallel processing as one of the distinguishing features of supercomputers.

The next generation, the Cray X-MP carried parallelism further by incorporating multiple processors (the successor, Cray Y-MP, had 8 processors which together could perform a billion floating point operations per second (1 gigaflop).

Soon other companies (particularly the Japanese manufactures NEC and Fujitsu) entered the market. The number of processors in supercomputers increased to as many as 1, 024, which can exceed 1 trillion floating-point operations per second (1 teraflop)

The ultimate in multiprocessing is the series of Connection Machines built by Thinking Machines Inc. (TMI) and designed by Daniel Hillis. These machines have up to 65, 000 very simple processors that run simultaneously and can form connections dynamically, somewhat like the process in the human brain. These “massively parallel” machines are thus attractive for artificial intelligence research.

As the power of standard computers continuous to grow, applications that formerly required a multimillion-dollar supercomputer can now run on a desktop workstation.

On the other hand there are always applications that will soak up whatever computing power can be brought to bear on them. These include: analysis of new aircraft designs, weather and climate models, the study of nuclear reactions, and the creation of models for the synthesis of proteins.

For many applications it may be more cost-effective to build systems with numerous coordinated processors (a sort of successor to the 1980s Connection Machine). For example, the Beowolf architecture involves “clusters” of ordinary PCs coordinated by software running on UNIX or Linux. The use of free software and commodity PCs can make this approach attractive, though application software still has to be rewritten to run on the distributed processors.

A new resource for parallel supercomputing came from an unlikely place: the new generation of cell processors found in game consoles such as the Sony Playstation 3. This architecture features tight integration of a central “power processor element” with multiple “synergistic processing elements”.

Finally, an ad hoc “supercomputer” can be created almost for free, using software that parcels out calculation tasks to thousands of computers participating via the Internet, as with SETI@Home (searching for extraterrestrial radio signals) and Folding@Home (for protein-folding analysis). In another approach, a large number of dedicated processors are placed in close proximity to each other (e.g. in a computer cluster); this saves considerable time moving data around and makes it possible for the processors to work together (rather than on separate tasks), for example in mesh and hypercube architecture.

The Top500 project ranks and details the 500 most powerful (non-distributed) computer systems in the world. The project was started in 1993 and publishes an updated list of the supercomputers twice a year. The project aims to provide a reliable basis for tracking and detecting trends in high-performance computing and bases rankings on HPL, a portable implementation of the high-performance LINPACK benchmark written in FORTRAN for distributed memory computers.

According to the 42^nd edition (November, 2013) of the Top500 list of the world’s most powerful supercomputers, Teanhe-2, a supercomputer developed by China’s National University of Defense Technology, retained its position as the world’s No.1 system with a performance of 33.86 petaflops/s (quadrillions of calculations per second).

Titan, a Cray XK7 system installed at the Department of Energy’s (DOE) Oak Ridge National Laboratory, remains the No.2 system. It achieved 17.59 Pflops/s on the Linpack benchmark. Titan is one of the most energy-efficient systems on the list.

Notes:

NEC (Nippon Electric Corporation) - японская компания, производитель электронной, компьютерной техники

Fujitsu - крупная японская корпорация, производитель электроники

SETI@Home (Search for Extra-Terrestrial Intelligence at Home – поиск внеземного разума на дому) – научный некоммер­ческий проект добровольных вычислений на платформе BOINC, использующий свободные вычислительные ре­сурсы на компьютерах добровольцев для поиска радиосиг­налов внеземных цивилизаций

LINPACK benchmark - тест производительности вычисли­тельных систем, по результатам которого составляется спи­сок 500 наиболее высокопроизводительных систем мира