Floating point operations per second

From testwiki
Jump to navigation Jump to search

Template:Short description Template:Other uses Template:Redirect-distinguish Template:Use mdy dates Floating point operations per second (FLOPS, flops or flop/s) is a measure of computer performance in computing, useful in fields of scientific computations that require floating-point calculations.[1]

For such cases, it is a more accurate measure than measuring instructions per second.Template:Cn

Floating-point arithmetic

Multipliers for flops
Name Unit Value
kiloFLOPS kFLOPS 103
megaFLOPS MFLOPS 106
gigaFLOPS GFLOPS 109
teraFLOPS TFLOPS 1012
petaFLOPS PFLOPS 1015
exaFLOPS EFLOPS 1018
zettaFLOPS ZFLOPS 1021
yottaFLOPS YFLOPS 1024
ronnaFLOPS RFLOPS 1027
quettaFLOPS QFLOPS 1030

Floating-point arithmetic is needed for very large or very small real numbers, or computations that require a large dynamic range. Floating-point representation is similar to scientific notation, except computers use base two (with rare exceptions), rather than base ten. The encoding scheme stores the sign, the exponent (in base two for Cray and VAX, base two or ten for IEEE floating point formats, and base 16 for IBM Floating Point Architecture) and the significand (number after the radix point). While several 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 64-bit numbers called double precision and longer numbers called extended precision (used for intermediate results). Floating-point representations can support a much wider range of values than fixed-point, with the ability to represent very small numbers and very large numbers.[2]

Dynamic range and precision

The exponentiation inherent in floating-point computation assures a much larger dynamic range – the largest and smallest numbers that can be represented – which is especially important when processing data sets where some of the data may have extremely large range of numerical values or where the range may be unpredictable. As such, floating-point processors are ideally suited for computationally intensive applications.[3]

Computational performance

FLOPS and MIPS are units of measure for the numerical computing performance of a computer. Floating-point operations are typically used in fields such as scientific computational research, as well as in machine learning. However, before the late 1980s floating-point hardware (it's possible to implement FP arithmetic in software over any integer hardware) was typically an optional feature, and computers that had it were said to be "scientific computers", or to have "scientific computation" capability. Thus the unit MIPS was useful to measure integer performance of any computer, including those without such a capability, and to account for architecture differences, similar MOPS (million operations per second) was used as early as 1970[4] as well. Note that besides integer (or fixed-point) arithmetics, examples of integer operation include data movement (A to B) or value testing (If A = B, then C). That's why MIPS as a performance benchmark is adequate when a computer is used in database queries, word processing, spreadsheets, or to run multiple virtual operating systems.[5][6] In 1974 David Kuck coined the terms flops and megaflops for the description of supercomputer performance of the day by the number of floating-point calculations they performed per second.[7] This was much better than using the prevalent MIPS to compare computers as this statistic usually had little bearing on the arithmetic capability of the machine on scientific tasks.

FLOPS by the largest supercomputer over time

FLOPS on an HPC-system can be calculated using this equation:[8]

FLOPS=racks×nodesrack×socketsnode×coressocket×cyclessecond×FLOPscycle.

This can be simplified to the most common case: a computer that has exactly 1 CPU:

FLOPS=cores×cyclessecond×FLOPscycle.

FLOPS can be recorded in different measures of precision, for example, the TOP500 supercomputer list ranks computers by 64 bit (double-precision floating-point format) operations per second, abbreviated to FP64.[9] Similar measures are available for 32-bit (FP32) and 16-bit (FP16) operations.

Template:Anchor

Floating-point operations per clock cycle for various processors

Template:Table alignment Template:Sort-under

Floating-point operations per clock cycle per core[10]
Microarchitecture Instruction set architecture FP64 FP32 FP16
Intel CPU
Intel 80486 x87 (32-bit) Template:Dunno 0.128[11] Template:Dunno
Template:Plainlist x87 (32-bit) Template:Dunno 0.5[11] Template:Dunno
Template:Plainlist MMX (64-bit) Template:Dunno 1[12] Template:Dunno
Intel P6 Pentium III SSE (64-bit) Template:Dunno 2[12] Template:Dunno
Intel NetBurst Pentium 4 (Willamette, Northwood) SSE2 (64-bit) 2 4 Template:Dunno
Intel P6 Pentium M SSE2 (64-bit) 1 2 Template:Dunno
Template:Plainlist SSE3 (64-bit) 2 4 Template:Dunno
Template:Plainlist Template:Plainlist 4 8 Template:Dunno
Intel Atom (Bonnell, Saltwell, Silvermont and Goldmont) SSE3 (128-bit) 2 4 Template:Dunno
Intel Sandy Bridge (Sandy Bridge, Ivy Bridge) AVX (256-bit) 8 16 0
Template:Ublist AVX2 & FMA (256-bit) 16 32 0
Intel Xeon Phi (Knights Corner) IMCI (512-bit) 16 32 0
Template:Plainlist AVX-512 & FMA (512-bit) 32 64 0
AMD CPU
AMD Bobcat AMD64 (64-bit) 2 4 0
Template:Plainlist AVX (128-bit) 4 8 0
AMD K10 SSE4/4a (128-bit) 4 8 0
AMD Bulldozer[13]
(Piledriver, Steamroller, Excavator) 		Template:Ublist 4 8 0
Template:Ublist AVX2 & FMA
(128-bit, 256-bit decoding)[14] 		8 16 0
Template:Ublist AVX2 & FMA (256-bit) 16 32 0
ARM CPU
ARM Cortex-A7, A9, A15 ARMv7 1 8 0
ARM Cortex-A32, A35 ARMv8 2 8 0
ARM Cortex-A53, A55, A57,[13] A72, A73, A75 ARMv8 4 8 0
ARM Cortex-A76, A77, A78 ARMv8 8 16 0
ARM Cortex-X1 ARMv8 16 32 Template:Dunno
Qualcomm Krait ARMv8 1 8 0
Qualcomm Kryo (1xx - 3xx) ARMv8 2 8 0
Qualcomm Kryo (4xx - 5xx) ARMv8 8 16 0
Samsung Exynos M1 and M2 ARMv8 2 8 0
Samsung Exynos M3 and M4 ARMv8 3 12 0
IBM PowerPC A2 (Blue Gene/Q) Template:Dunno 8 8
(as FP64)		 0
Hitachi SH-4[15][16] SH-4 1 7 0
Nvidia GPU
Nvidia Curie (GeForce 6 series and GeForce 7 series) PTX Template:Dunno 8 Template:Dunno
Nvidia Tesla 2.0 (GeForce GTX 260–295) PTX Template:Dunno 2 Template:Dunno
Nvidia Fermi

(only GeForce GTX 465–480, 560 Ti, 570–590)

PTX Template:1/4
(locked by driver,
1 in hardware)		 2 0
Nvidia Fermi

(only Quadro 600–2000)

PTX Template:Frac 2 0
Nvidia Fermi

(only Quadro 4000–7000, Tesla)

PTX 1 2 0
Nvidia Kepler

(GeForce (except Titan and Titan Black), Quadro (except K6000), Tesla K10)

PTX Template:Frac
(for GK110:
locked by driver,
Template:2/3 in hardware)		 2 0
Nvidia Kepler

(GeForce GTX Titan and Titan Black, Quadro K6000, Tesla (except K10))

PTX Template:2/3 2 0
Template:Ublist PTX Template:Frac 2 Template:Frac
Nvidia Pascal (only Quadro GP100 and Tesla P100) PTX 1 2 4
Nvidia Volta[17] PTX 1 2 (FP32) + 2 (INT32) 16
Nvidia Turing (only GeForce 16XX) PTX Template:Frac 2 (FP32) + 2 (INT32) 4
Nvidia Turing (all except GeForce 16XX) PTX Template:Frac 2 (FP32) + 2 (INT32) 16
Nvidia Ampere[18][19] (only Tesla A100/A30) PTX 2 2 (FP32) + 2 (INT32) 32
Template:Plainlist PTX Template:Frac Template:Nowrap
or
Template:Nowrap		 8
Nvidia Hopper PTX 2 2 (FP32) + 1 (INT32) 32
AMD GPU
AMD TeraScale 1 (Radeon HD 4000 series) TeraScale 1 0.4 2 Template:Dunno
AMD TeraScale 2 (Radeon HD 5000 series) TeraScale 2 1 2 Template:Dunno
AMD TeraScale 3 (Radeon HD 6000 series) TeraScale 3 1 4 Template:Dunno
AMD GCN
(only Radeon Pro W 8100–9100) 		GCN 1 2 Template:Dunno
AMD GCN
(all except Radeon Pro W 8100–9100, Vega 10–20) 		GCN Template:Frac 2 4
AMD GCN Vega 10 GCN Template:Frac 2 4
AMD GCN Vega 20
(only Radeon VII)		 GCN Template:1/2
(locked by driver,
1 in hardware)		 2 4
AMD GCN Vega 20
(only Radeon Instinct MI50 / MI60 and Radeon Pro VII) 		GCN 1 2 4
Template:Plainlist RDNA Template:Frac 2 4
AMD RDNA3 RDNA Template:Frac? 4 8?
AMD CDNA CDNA 1 4
(Tensor)[20]		 16
AMD CDNA 2 CDNA 2 4
(Tensor)		 4
(Tensor)		 16
Intel GPU
Intel Xe-LP (Iris Xe MAX)[21] Xe Template:1/2? 2 4
Intel Xe-HPG (Arc Alchemist)[21] Xe 0 2 16
Intel Xe-HPC (Ponte Vecchio)[22] Xe 2 2 32
Intel Xe2 (Arc Battlemage) Xe2 Template:Frac 2 16
Qualcomm GPU
Qualcomm Adreno 5x0 Adreno 5xx 1 2 4
Qualcomm Adreno 6x0 Adreno 6xx 1 2 4
Graphcore
Graphcore Colossus GC2[23][24] Template:Dunno 0 16 64
Template:Plainlist Template:Dunno 0 32 128
Supercomputer
ENIAC @ 100 kHz in 1945 0.004[25]
(~Template:Val)
48-bit processor @ 208 kHz in CDC 1604 in 1960
60-bit processor @ 10 MHz in CDC 6600 in 1964 0.3
(FP60)
60-bit processor @ 10 MHz in CDC 7600 in 1967 1.0
(FP60)
Cray-1 @ 80 MHz in 1976 2
(700 FLOPS/W)
CDC Cyber 205 @ 50 MHz in 1981

FORTRAN compiler (ANSI 77 with vector extensions)

8 16
Transputer IMS T800-20 @ 20 MHz in 1987 0.08[26]
Parallella E16 @ 1000 MHz in 2012 2[27]
(5.0 GFLOPS/W)[28]
Parallella E64 @ 800 MHz in 2012 2[29]
(50.0 GFLOPS/W)[28]
Microarchitecture Instruction set architecture FP64 FP32 FP16

Performance records

Single computer records

In June 1997, Intel's ASCI Red was the world's first computer to achieve one teraFLOPS and beyond. Sandia director Bill Camp said that ASCI Red had the best reliability of any supercomputer ever built, and "was supercomputing's high-water mark in longevity, price, and performance".[30]

NEC's SX-9 supercomputer was the world's first vector processor to exceed 100 gigaFLOPS per single core.

In June 2006, a new computer was announced by Japanese research institute RIKEN, the MDGRAPE-3. The computer's performance tops out at one petaFLOPS, almost two times faster than the 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 pipelines for simulating molecular dynamics.

By 2007, Intel Corporation unveiled the experimental multi-core POLARIS chip, which achieves 1 teraFLOPS at 3.13 GHz. The 80-core chip can raise this result to 2 teraFLOPS at 6.26 GHz, although the thermal dissipation at this frequency exceeds 190 watts.[31]

In June 2007, Top500.org reported the fastest computer in the world to be the IBM Blue Gene/L supercomputer, measuring a peak of 596 teraFLOPS.[32] The Cray XT4 hit second place with 101.7 teraFLOPS.

On June 26, 2007, IBM announced the second generation of its top supercomputer, dubbed Blue Gene/P and designed to continuously operate at speeds exceeding one petaFLOPS, faster than the Blue Gene/L. When configured to do so, it can reach speeds in excess of three petaFLOPS.[33]

On October 25, 2007, NEC Corporation of Japan issued a press release announcing its SX series model SX-9,[34] claiming it to be the world's fastest vector supercomputer. The SX-9 features the first CPU capable of 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 full scale research runs on an AMD, Sun supercomputer named Ranger,[35] the most powerful supercomputing system in the world for open science research, which operates at sustained speed of 0.5 petaFLOPS.

On May 25, 2008, an American supercomputer built by IBM, named 'Roadrunner', reached the computing milestone of one petaFLOPS. It headed the June 2008 and November 2008 TOP500 list of the most powerful supercomputers (excluding grid computers).[36][37] The computer is located at Los Alamos National Laboratory in New Mexico. The computer's name refers to the New Mexico state bird, the greater roadrunner (Geococcyx californianus).[38]

In June 2008, AMD released ATI Radeon HD 4800 series, which are reported to be the first GPUs to achieve one teraFLOPS. On August 12, 2008, AMD released the ATI Radeon HD 4870X2 graphics card with two Radeon R770 GPUs totaling 2.4 teraFLOPS.

In November 2008, an upgrade to the Cray Jaguar supercomputer at the Department of Energy's (DOE's) Oak Ridge National Laboratory (ORNL) raised the system's computing power to a peak 1.64 petaFLOPS, making Jaguar the world's first petaFLOPS system dedicated to open research. In early 2009 the supercomputer was named after a mythical creature, Kraken. Kraken was declared the world's fastest university-managed supercomputer and sixth fastest overall in the 2009 TOP500 list. In 2010 Kraken was upgraded and can operate faster and is more powerful.

In 2009, the Cray Jaguar performed at 1.75 petaFLOPS, beating the IBM Roadrunner for the number one spot on the TOP500 list.[39]

In October 2010, China unveiled the Tianhe-1, a supercomputer that operates at a peak computing rate of 2.5 petaFLOPS.[40][41]

Template:As of the fastest PC processor reached 109 gigaFLOPS (Intel Core i7 980 XE)[42] in double precision calculations. GPUs are considerably more powerful. For example, Nvidia Tesla C2050 GPU computing processors perform around 515 gigaFLOPS[43] in double precision calculations, and the AMD FireStream 9270 peaks at 240 gigaFLOPS.[44]

In November 2011, it was announced that Japan had achieved 10.51 petaFLOPS with its K computer.[45] It has 88,128 SPARC64 VIIIfx processors in 864 racks, with theoretical performance of 11.28 petaFLOPS. It is named after the Japanese word "kei", which stands for 10 quadrillion,[46] corresponding to the target speed of 10 petaFLOPS.

On November 15, 2011, Intel demonstrated a single x86-based processor, code-named "Knights Corner", sustaining more than a teraFLOPS on a wide range of DGEMM operations. Intel emphasized during the demonstration that this was a sustained teraFLOPS (not "raw teraFLOPS" used by others to get higher but less meaningful numbers), and that it was the first general purpose processor to ever cross a teraFLOPS.[47][48]

On June 18, 2012, IBM's Sequoia supercomputer system, based at the U.S. Lawrence Livermore National Laboratory (LLNL), reached 16 petaFLOPS, setting the world record and claiming first place in the latest TOP500 list.[49]

On November 12, 2012, the TOP500 list certified Titan as the world's fastest supercomputer per the LINPACK benchmark, at 17.59 petaFLOPS.[50][51] It was developed by Cray Inc. at the Oak Ridge National Laboratory and combines AMD Opteron processors with "Kepler" NVIDIA Tesla graphics processing unit (GPU) technologies.[52][53]

On June 10, 2013, China's Tianhe-2 was ranked the world's fastest with 33.86 petaFLOPS.[54]

On June 20, 2016, China's Sunway TaihuLight was ranked the world's fastest with 93 petaFLOPS on the LINPACK benchmark (out of 125 peak petaFLOPS). The system was installed at the National Supercomputing Center in Wuxi, and represented more performance than the next five most powerful systems on the TOP500 list did at the time combined.[55]

In June 2019, Summit, an IBM-built supercomputer now running at the Department of Energy's (DOE) Oak Ridge National Laboratory (ORNL), captured the number one spot with a performance of 148.6 petaFLOPS on High Performance Linpack (HPL), the benchmark used to rank the TOP500 list. Summit has 4,356 nodes, each one equipped with two 22-core Power9 CPUs, and six NVIDIA Tesla V100 GPUs.[56]

In June 2022, the United States' Frontier was the most powerful supercomputer on TOP500, reaching 1102 petaFlops (1.102 exaFlops) on the LINPACK benchmarks. [57]Template:Circular reference

In November 2024, the United States’ El Capitan exascale supercomputer, hosted at the Lawrence Livermore National Laboratory in Livermore, displaced Frontier as the world's fastest supercomputer in the 64th edition of the Top500 (Nov 2024).

Distributed computing records

Distributed computing uses the Internet to link personal computers to achieve more FLOPS:

  • Template:As of, the Folding@home network has over 2.3 exaFLOPS of total computing power.[58][59][60][61] It is the most powerful distributed computer network, being the first ever to break 1 exaFLOPS of total computing power. This level of performance is primarily enabled by the cumulative effort of a vast array of powerful GPU and CPU units.[62]

Cost of computing

Hardware costs

Date Approximate USD per GFLOPS Platform providing the lowest cost per GFLOPS Comments
Unadjusted Template:Inflation-yearTemplate:Inflation-fn
1945 $1.265T $Template:InflationT ENIAC: Template:US$ in 1945 and $Template:Inflation in 2023. Template:US$ / Template:Val. First-generation (vacuum tube-based) electronic digital computer.
1961 $18.672B $Template:InflationB A basic installation of IBM 7030 Stretch had a cost at the time of Template:US$ each. The IBM 7030 Stretch performs one floating-point multiply every Template:Val.[68] Second-generation (discrete transistor-based) computer.
1964 $2.3B $Template:InflationB Base model CDC 6600 price: $6,891,300. The CDC 6600 is considered to be the first commercially-successful supercomputer.
1984 $18,750,000 $Template:Inflation Cray X-MP/48 $15,000,000 / 0.8 GFLOPS. Third-generation (integrated circuit-based) computer.
1997 $30,000 $Template:Inflation Two 16-processor Beowulf clusters with Pentium Pro microprocessors[69]
Template:Sort $1,000 $Template:Inflation Bunyip Beowulf cluster Bunyip was the first sub-Template:Val computing technology. It won the Gordon Bell Prize in 2000.
Template:Sort $640 $Template:Inflation KLAT2 KLAT2 was the first computing technology which scaled to large applications while staying under Template:Val.[70]
Template:Sort $83.86 $Template:Inflation KASY0 KASY0 was the first sub-Template:Val computing technology. KASY0 achieved 471 GFLOPS on 32-bit HPL. At a cost of less than $39,500, that makes it the first supercomputer to break $100/GFLOPS.[71]
Template:Sort $48.31 $Template:Inflation Microwulf As of August 2007, this Template:Val "personal" Beowulf cluster can be built for $1256.[72]
Template:Sort $1.80 $Template:Inflation HPU4Science This $30,000 cluster was built using only commercially available "gamer" grade hardware.[73]
Template:Sort 75.00¢ Template:Inflation¢ Quad AMD Radeon 7970 System A quad AMD Radeon 7970 desktop computer reaching 16 TFLOPS of single-precision, 4 TFLOPS of double-precision computing performance. Total system cost was $3000; built using only commercially available hardware.[74]
Template:Sort 21.68¢ Template:Inflation¢ Sony PlayStation 4 The Sony PlayStation 4 is listed as having a peak performance of Template:Val, at a price of $399[75]
Template:Sort 16.11¢ Template:Inflation¢ AMD Sempron 145 & GeForce GTX 760 system Built using commercially available parts, a system using one AMD Sempron 145 and three Nvidia GeForce GTX 760 reaches a total of Template:Val for a total cost of Template:US$.[76]
Template:Sort 12.41¢ Template:Inflation¢ Pentium G550 & Radeon R9 290 system Built using commercially available parts. Intel Pentium G550 and AMD Radeon R9 290 tops out at Template:Val grand total of Template:US$.[77]
Template:Sort 7.85¢ Template:Inflation¢ Celeron G1830 & Radeon R9 295X2 system Built using commercially available parts. Intel Celeron G1830 and AMD Radeon R9 295X2 tops out at over Template:Val at a grand total of Template:US$.[78][79]
Template:Sort 6.00¢ Template:Inflation¢ AMD Ryzen 7 1700 & AMD Radeon Vega Frontier Edition system Built using commercially available parts. AMD Ryzen 7 1700 CPU combined with AMD Radeon Vega FE cards in CrossFire tops out at over Template:Val at just under Template:USD for the complete system.[80]
October 2017 2.73¢ Template:Inflation¢ Intel Celeron G3930 & AMD RX Vega 64 system Built using commercially available parts. Three AMD RX Vega 64 graphics cards provide just over 75 TFLOPS half precision (38 TFLOPS SP or 2.6 TFLOPS DP when combined with the CPU) at ~$2,050 for the complete system.[81]
November 2020 3.14¢ Template:Inflation¢ AMD Ryzen 3600 & 3× NVIDIA RTX 3080 system AMD Ryzen 3600 @ 484 GFLOPS & $199.99

3× NVIDIA RTX 3080 @ 29,770 GFLOPS each & $699.99

Total system GFLOPS = 89,794 / TFLOPS = 89.794

Total system cost incl. realistic but low cost parts; matched with other example = $2839[82]

Template:US$/GFLOP = $0.0314

November 2020 3.88¢ Template:Inflation¢ PlayStation 5 The Sony PlayStation 5 Digital Edition is listed as having a peak performance of 10.28 TFLOPS (20.56 TFLOPS at half precision) at a retail price of $399.[83]
November 2020 4.11¢ Template:Inflation¢ Xbox Series X Microsoft's Xbox Series X is listed as having a peak performance of 12.15 TFLOPS (24.30 TFLOPS at half precision) at a retail price of $499.[84]
September 2022 1.94¢ Template:Inflation¢ RTX 4090 Nvidia's RTX 4090 is listed as having a peak performance of 82.6 TFLOPS (1.32 PFLOPS at 8-bit precision) at a retail price of $1599.[85]
May 2023 1.25¢ Template:Inflation¢ Radeon RX 7600 AMD's RX 7600 is listed as having a peak performance of 21.5 TFLOPS at a retail price of $269.[86]


See also

Template:Div col

Template:Div col end Template:Clear

References

Template:Reflist

Template:Graphics Processing Unit Template:CPU technologies

Template:Authority control

  1. ↑ Template:Cite web
  2. ↑ Floating Point Retrieved on December 25, 2009.
  3. ↑ Summary: Fixed-point (integer) vs floating-point Template:Webarchive Retrieved on December 25, 2009.
  4. ↑ Template:Cite book
  5. ↑ Fixed versus floating point. Retrieved on December 25, 2009.
  6. ↑ Data manipulation and math calculation. Retrieved on December 25, 2009.
  7. ↑ Template:Cite book
  8. ↑ Template:Cite web
  9. ↑ Template:Cite web
  10. ↑ Template:Cite web
  11. ↑ 11.0 11.1 Template:Cite web
  12. ↑ 12.0 12.1 Template:Cite web
  13. ↑ 13.0 13.1 Cite error: Invalid <ref> tag; no text was provided for refs named tpeak_jos
  14. ↑ Template:Cite web
  15. ↑ Template:Cite journal
  16. ↑ Template:Cite web
  17. ↑ Template:Cite web
  18. ↑ Template:Cite web
  19. ↑ Template:Cite web
  20. ↑ Template:Cite web
  21. ↑ 21.0 21.1 Template:Cite web
  22. ↑ Template:Cite web
  23. ↑ Template:Cite web
  24. ↑ Archived at GhostarchiveTemplate:Cbignore and the Wayback MachineTemplate:Cbignore: Template:Cite webTemplate:Cbignore
  25. ↑ ENIAC @ 100 kHz with 385 Flops Template:Cite web
  26. ↑ Template:Cite web
  27. ↑ Epiphany-III 16-core 65nm Microprocessor (E16G301) // admin (August 19, 2012)
  28. ↑ 28.0 28.1 Template:Cite web
  29. ↑ Epiphany-IV 64-core 28nm Microprocessor (E64G401) // admin (August 19, 2012)
  30. ↑ Template:Cite web
  31. ↑ Template:Cite web
  32. ↑ Template:Cite news
  33. ↑ Template:Cite news
  34. ↑ Template:Cite news
  35. ↑ Template:Cite web
  36. ↑ Template:Cite web
  37. ↑ Template:Cite web
  38. ↑ Template:Cite news
  39. ↑ Template:Cite news
  40. ↑ Template:Cite news
  41. ↑ Template:Cite web
  42. ↑ Template:Cite web
  43. ↑ Template:Cite web
  44. ↑ Template:Cite web
  45. ↑ Template:Cite web
  46. ↑ See Japanese numbers
  47. ↑ Template:Cite web
  48. ↑ Template:Cite web
  49. ↑ Template:Cite news
  50. ↑ Template:Cite news
  51. ↑ Template:Cite web
  52. ↑ Template:Cite web
  53. ↑ Template:Cite web
  54. ↑ Template:Cite magazine
  55. ↑ Template:Cite web
  56. ↑ Template:Cite web
  57. ↑ Template:Cite web
  58. ↑ Template:Cite web
  59. ↑ Template:Cite web
  60. ↑ Template:Cite web
  61. ↑ Template:Cite web
  62. ↑ Template:Cite press release
  63. ↑ Template:Cite web
  64. ↑ Template:Cite web
  65. ↑ Template:Cite web
  66. ↑ Template:Cite web
  67. ↑ Template:Cite web
  68. ↑ Template:Cite web
  69. ↑ Template:Cite web
  70. ↑ Template:Cite web
  71. ↑ Template:Cite web
  72. ↑ Template:Cite web
  73. ↑ Adam Stevenson, Yann Le Du, and Mariem El Afrit. "High-performance computing on gamer PCs." Ars Technica. March 31, 2011.
  74. ↑ Template:Cite web
  75. ↑ "Sony Sparks Price War With PS4 Priced at $399." CNBC. June 11, 2013.
  76. ↑ Template:Cite web
  77. ↑ Template:Cite web
  78. ↑ Template:Cite web
  79. ↑ Template:Cite web
  80. ↑ Template:Cite web
  81. ↑ Template:Cite web
  82. ↑ Template:Cite web
  83. ↑ Template:Cite web
  84. ↑ Template:Cite web
  85. ↑ Template:Cite web
  86. ↑ Template:Cite web
Retrieved from "https://en.wiki.beta.math.wmflabs.org/w/index.php?title=Floating_point_operations_per_second&oldid=2405"