In computing, floating point per second ( FLOPS , flops or flop/s ) is a measure of computer performance, useful in the field of scientific calculations that require the calculation of floating point. For such cases the size is more accurate than measuring instructions per second.
A similar term FLOP is often used for floating-point operations, for example as a floating-point counting unit performed by an algorithm or computer hardware.
Video FLOPS
Floating point-arithmetic
Floating-point arithmetic is required for very large or very small real numbers, or calculations requiring large dynamic ranges. The floating-point representation is similar to scientific notation, unless it's all done in base two, not ten. The coding scheme stores the sign, exponent (in basis two for Cray and VAX, base two or ten for IEEE floating point format, and base 16 for IBM Floating Point Architecture) and Significand (number after radix point). While some similar formats are in use, the most common is ANSI/IEEE Std. 754-1985. This standard defines the format for 32-bit numbers called single precision , as well as the 64-bit number called double precision and the longer number is called expanded precision (used for intermediate results). Floating-point representation can support a much wider range of values ​​than fixed-points, with the ability to represent very small numbers and large numbers.
Dynamic range and precision
The inherent exponents in floating-point calculations ensure a much larger dynamic range - the largest and smallest numbers that can be represented - which is very important when processing large data sets or where ranges may be unpredictable. Thus, floating-point processors are ideal for computationally intensive applications.
Computational performance
FLOPS and MIPS are the unit of measure for computer numerical computing performance. Floating-point operations are typically used in areas such as scientific computing research. The MIPS unit measures the performance of integers from a computer. Examples of integer operations include movement of data (A to B) or test values ​​(If A = B, then C). MIPS as an adequate performance benchmark when computers are used in database queries, word processors, spreadsheets, or to run multiple virtual operating systems. Frank H. McMahon, from Lawrence Livermore National Laboratory, invented the term FLOPS and MFLOPS (megaFLOPS) so that he could compare the day's supercomputers to the number of floating-point calculations they do per second. This is much better than using an ordinary MIPS to compare computers because these statistics usually have little effect on the machine's arithmetic abilities.
FLOPS dapat dihitung menggunakan persamaan ini:
Maps FLOPS
FLOPs per siklus untuk berbagai prosesor
Catatan kinerja
Catatan komputer tunggal
In June 1997, Intel ASCI Red was the first computer in the world to reach one teraFLOPS and beyond. Sandia Director Bill Camp said ASCI Red has the best reliability of all the supercomputers ever made, and "is a supercomputer in water levels in longevity, price, and performance".
The NEC SX-9 supercomputer is the world's first vector processor that surpasses 100 gigaFLOPS per single core.
For comparison, handheld calculators perform relatively few FLOPS. The computer response time under 0.1 seconds in the context of calculation is usually considered instantaneous by human operators, so a simple calculator requires only about 10 FLOPS to be considered functional.
In June 2006, a new computer was announced by the Japanese research institute RIKEN, MDGRAPE-3. Computer performance peaks on a FLLPS map, almost twice as fast as Blue Gene/L, but MDGRAPE-3 is not a general purpose computer, which is why it does not appear in the Top500.org list. It has special purpose pipes to simulate molecular dynamics.
In 2007, Intel Corporation unveiled an experimental POLARIS multi-core chip, which reached 1 teraFLOPS at 3.13 GHz. The 80-core chip can increase this result to 2 teraFLOPS at 6.26 GHz, although thermal dissipation at this frequency exceeds 190 watts.
On June 26, 2007, IBM announced the second generation of top supercomputing, dubbed Blue Gene/P and is designed to continue operating at speeds exceeding one FLOPS map. When configured to do so, it can reach speeds of more than three FLLPS maps.
In June 2007, Top500.org reported the world's fastest computer to become IBM Blue Gene/L supercomputer, which measures the peak of 596 teraFLOPS. The Cray XT4 reached second place with 101.7 teraFLOPS.
On October 25, 2007, the NEC Corporation of Japan issued a press release announcing its SX-9 SX series, claiming it as the world's fastest vector supercomputer. The SX-9 features the first CPU capable of producing a peak vector performance of 102.4 gigaFLOPS per single core.
On February 4, 2008, the NSF and the University of Texas at Austin opened a full-scale research run on AMD, a solar supercomputer named Ranger, the world's strongest supercomputer system for open science research, operating at a sustainable speed of 0.5 FLOPS maps.
On May 25, 2008, an American supercomputer built by IBM, named 'Roadrunner', reached the computing milestone of a FLLPSPS map. It leads the TOP500 list of June 2008 and November 2008 from the most powerful supercomputer (excluding computer grid). This computer is located at Los Alamos National Laboratory in New Mexico. The computer name refers to a New Mexico country bird, the larger roadrunner ( Geococcyx californianus ).
In June 2008, AMD released the ATI Radeon HD 4800 series, which was reported as the first GPU to reach one teraFLOPS. On August 12, 2008, AMD released an ATI Radeon HD 4870X2 graphics card with two Radeon R770 GPUs with a total of 2.4 teraFLOPS.
In November 2008, upgrading to the Jaguar Cray XT supercomputer at the Department of Energy's (DOE's) Oak Ridge National Laboratory (ORNL) enhanced the computing power of the system to the top of the 1.64 FLOPS map, making Jaguar the world's first dedicated FLOPS map system for open research. In early 2009 the supercomputer was named after a mythical creature, the Kraken. The Kraken is declared the fastest supercomputer run by the university and the sixth fastest overall in TOP500 2009 list. In 2010 the Kraken is improved and can operate faster and stronger.
In 2009, Cray Jaguar performed at 1.75 FLOPS maps, beating IBM Roadrunner for the number one spot on TOP500's list.
In October 2010, China launched Tianhe-1, a supercomputer that operates at the peak computing level of 2.5 FLOPS maps.
In 2010, the fastest six-core PC processor reached 109 gigaFLOPS (Intel Core i7 980 XE) in double precision calculations. GPU is much more powerful. For example, the Nvidia Tesla C2050 GPU computing processor performs about 515 gigaFLOPS in double precision calculations, and AMD FireStream 9270 peaks at 240 gigaFLOPS. In a single-precision performance, the Nvidia Tesla C2050 computing processor performs about 1.03 teraFLOPS and AMD FireStream 9270 cards at 1.2 teraFLOPS. Both Nvidia and AMD consumer game GPUs can achieve a higher FLOPS. For example, the AMD HemlockXT 5970 achieves 928 gigaFLOPS in double precision calculations with two GPUs on board and Nvidia GTX 480 reaching 672 gigaFLOPS with one GPU in it.
On December 2, 2010, the US Air Force launched a defense supercomputer consisting of 1,760 PlayStation 3 consoles that can run 500 teraFLOPS.
In November 2011, it was announced that Japan had reached 10.51 FLLOPS maps with its K computer. This is still in the development and adjustment of software performance is currently underway. It has 88,128 SPARC64 VIIIfx processors on 864 racks, with a theoretical performance of 11.28 FLOPS maps. Named from the Japanese word "kei", which means 10 quadrillion, corresponds to the target speed of 10 FLOPS maps.
On 15 November 2011, Intel demonstrated a single x86-based processor, codenamed "Knights Corner", retaining more than one teraFLOPS on DGEMM operations. Intel emphasized during the demonstration that this is a continuous teraFLOPS (not a "raw teraFLOPS" used by others to get higher numbers but less meaningful), and that is the first general-purpose processor ever crossed teraFLOPS.
On June 18, 2012, IBM's Sequoia supercomputer system, based at Lawrence Livermore US National Laboratory (LLNL), reached 16 FLOPS maps, set a world record and claimed first place in the latest TOP500 list.
On November 12, 2012, TOP500's list states Titan as the world's fastest supercomputer per LINPACK benchmark, at 17.59 FLOPS maps. It was developed by Cray Inc. at Oak Ridge National Laboratory and incorporates an AMD Opteron processor with graphics processing technology (GPU) "NVIDIA Tesla" Kepler.
On June 10, 2013, China Tianhe-2 was ranked the world's fastest with 33.86 FLOPS maps.
On June 20, 2016, Sunway TaihuLight China was ranked the world's fastest with 93 FLLPS maps on the LINPACK benchmark (from 125 tops of the FLLPS maps). The system, which is almost exclusively based on technology developed in China, is installed at the National Supercomputing Center in Wuxi, and represents the performance of more than the next five most powerful systems on the TOP500 list combined.
Distributed computing records
Distributed computing using the Internet to connect personal computers to achieve more FLOPS:
- As of October 2016, the @ home home network has more than 100 FLOPS maps of total computing power. This is the first computing project of any kind to traverse 1, 2, 3, 4, and 5 original mapFLOPS maps. This performance level is mainly enabled by the cumulative effort of various GPU units and powerful CPUs.
- Beginning in January 2018, the entire BOINC network averages around 20à ¢ â,¬Â, the FLOPS map.
- Since July 2014, SETI @ Home, using the BOINC software platform, averages 681 teraFLOPS.
- In July 2014, Einstein @ Home, a project using the BOINC network, was churning out at 492 tera FLLOS.
- In July 2014, MilkyWay @ Home, using BOINC infrastructure, calculated at 471 teraFLOPS.
- Starting January 2017, GIMPS, looking for Mersenne prime numbers and maintaining 300 teraFLOPS types.
Further developments
In 2008, James Bamford The Shadow Factory reported that the NSA told the Pentagon that it would require an exaflop computer by 2018.
Given the current speed of progress, the supercomputer is projected to reach 1 exaFLOPS (EFLOPS) by 2018. Cray, Inc. announced in December 2009 plans to build an EFLOPS 1 supercomputer before 2020. Erik P. DeBenedictis of Sandia National Laboratories theorizes that zettaFLOPS (ZFLOPS) computers are required to complete full weather modeling from a two-week span. Such a system may be built around 2030.
Cost of computing
Hardware cost
See also
References
Source of the article : Wikipedia
0 Comments