Peak FP32 rate (CPU) over time
Theoretical single-precision peak from unit count, clock and operations per cycle. An upper bound, not an achieved rate.
- Device scope
- cpu
- Method
- Theoretical peak
- Plotted in
- floating-point operations per second
Chart
4 figures from 4 machines, spanning 1992 to 2006. The lowest peak fp32 rate here is 8 MFLOP/s (Commodore Amiga 4000, 1992) and the highest is 218 GFLOP/s (Sony PlayStation 3, 2006). That is a range of about 27 250× within one comparability group; the figures are plotted in FLOP/s. Points use the release date of the machine the figure belongs to. The table below lists absent figures with their recorded reasons.
Data
| Machine | Released | Figure | Confidence | Sources |
|---|---|---|---|---|
| Commodore Amiga 4000 Standard 68040 | October 1992 | 8 MFLOP/s | Theoretical provisional | |
| Silicon Graphics Octane MIPS R10000 | 1997 | 500 MFLOP/s | Derived | |
| Sony PlayStation 2 Emotion Engine | March 2000 | 6.2 GFLOP/s | Vendor-rated | |
| Sony PlayStation 3 Cell Broadband Engine | November 2006 | 218 GFLOP/s | Vendor-rated provisional | |
| Apollo Guidance Computer (Block II) Apollo Guidance Computer Block II processor | 1966 | not applicable The AGC has no floating-point hardware. Its arithmetic unit is a one's-complement fixed-point adder with shifting gates, so there is no rate to state — the question does not apply. | not plotted | — |
| Saturn Launch Vehicle Digital Computer Saturn Launch Vehicle Digital Computer | 1966 | not applicable The LVDC is a fixed-point machine, as its own characteristics table states in the type row. There is no floating-point rate to give. | not plotted | — |
| Apple II MOS Technology 6502 | June 1977 | unknown The manual documents software floating point but does not explicitly state whether the Apple II configuration has floating-point hardware. | not plotted | — |
| Atari 2600 MOS Technology 6507 | September 1977 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Intel 8086 PC Intel 8086 | 1978 | not applicable The 8086 has no integrated floating-point unit; an optional 8087 is a separate processor and is not fitted to this configuration. | not plotted | — |
| Atari 800 MOS Technology 6502 | November 1979 | not applicable The Atari 800's 6502 contains no floating-point hardware. | not plotted | — |
| IBM Personal Computer 5150 Intel 8088 | August 1981 | not applicable The selected 5150 configuration contains no 8087 floating-point coprocessor. | not plotted | — |
| BBC Micro Model B MOS Technology 6502 | December 1981 | unknown The detailed hardware documentation does not identify a floating-point processor in the standard Model B. | not plotted | — |
| Intel 80286 PC Intel 80286 | 1982 | not applicable The 80286 has no integrated floating-point unit; an 80287 would be a separately fitted coprocessor. | not plotted | — |
| ZX Spectrum 48K Zilog Z80A | April 1982 | not applicable The Z80A contains no floating-point hardware. | not plotted | — |
| Commodore 64 MOS Technology 6510 | August 1982 | not applicable No floating-point hardware. The 6510 does decimal and binary integer arithmetic in an eight-bit accumulator; the Commodore 64's floating point is a software library in the BASIC ROM, which is not a property of the processor. | not plotted | — |
| MSX Zilog Z80A | 1983 | not applicable The Z80A contains no floating-point hardware. | not plotted | — |
| Apple Macintosh 128K Motorola 68000 | January 1984 | not applicable The original Macintosh does not fit a 68881 floating-point coprocessor. | not plotted | — |
| Amstrad CPC 464 Zilog Z80A | June 1984 | not applicable The Z80A contains no floating-point hardware. | not plotted | — |
| Intel 80386DX PC Intel 80386DX | 1985 | not applicable The 80386DX has no integrated floating-point unit; an 80387 would be a separate coprocessor. | not plotted | — |
| Atari 520ST Motorola 68000 | June 1985 | not applicable The 520ST does not fit a 68881 floating-point coprocessor. | not plotted | — |
| Nintendo Entertainment System Ricoh 2A03 | October 1985 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Sega Master System Zilog Z80 | June 1986 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Commodore Amiga 500 Motorola MC68000 | 1987 | not applicable The 68000 has no floating-point hardware on the die. Floating point requires a separate 68881 or 68882 coprocessor, which the Amiga 500 does not have, so any floating-point rate would describe software running on an integer ALU. | not plotted | — |
| Acorn Archimedes A310 Acorn ARM2 | June 1987 | not applicable ARM2 has no integrated floating-point hardware; the later FPA is a separate option. | not plotted | — |
| NEC PC Engine Hudson Soft HuC6280 | October 1987 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Sega Mega Drive Motorola 68000 | October 1988 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Intel 80486DX PC Intel 80486DX | 1989 | unknown The 80486DX integrates an x87 unit, but the cited documentation does not state an FP32 peak rate in FLOP/s. | not plotted | — |
| Nintendo Game Boy Sharp LR35902 | April 1989 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| SNK Neo Geo AES Motorola 68000 | April 1990 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Super Nintendo Entertainment System Ricoh 5A22 | August 1991 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Commodore Amiga 1200 Motorola MC68EC020 | October 1992 | not applicable No floating-point hardware on the die. The Amiga 1200 has a socket for a 68881 or 68882 coprocessor, but the manual records it as factory-installed only, so a stock machine has none and any floating-point figure would describe software on an integer ALU. | not plotted | — |
| Intel Pentium PC Intel Pentium | 1993 | unknown The processor includes a floating-point unit, but the cited specification does not publish an FP32 peak in FLOP/s. | not plotted | — |
| Intel Pentium II PC Intel Pentium II | 1997 | unknown The processor has floating-point hardware, but the cited specification does not publish an FP32 peak in FLOP/s. | not plotted | — |
| Game Boy Advance ARM7TDMI | March 2001 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Intel Pentium 4 PC Intel Pentium 4 | 2002 | unknown The processor has floating-point hardware, but the cited specification does not publish an FP32 peak in FLOP/s. | not plotted | — |
| Nintendo DS ARM946E-S | November 2004 | not applicable This console CPU has no integrated floating-point execution unit. | not plotted | — |
| Intel Core 2 Duo E6600 PC Intel Core 2 Duo E6600 | July 2006 | unknown The processor supports scalar and SIMD floating point, but the cited specification does not state an FP32 peak in FLOP/s. | not plotted | — |
| Apple iPhone (1st generation) | June 2007 | unknown The cited analysis states that the ARM11 CPU in the iPhone has a vector floating point unit, and states no rate for it. Apple states nothing at all. | not plotted | — |
| Intel Core i7-2600K PC Intel Core i7-2600K | January 2011 | unknown The cited product table gives cores, clock and power, not a single-precision peak rate. | not plotted | — |
| Intel Core i9-14900K PC Intel Core i9-14900K | October 2023 | unknown Intel does publish a floating-point rate for this processor — 1228.8 GFLOPS, in the export-compliance matrix it maintains for the US Export Administration Regulations — and states no precision for it anywhere in that document. Without a precision the figure cannot answer a single-precision question, and it cannot be assumed to be one: the same 1228.8 is listed for the faster-clocked 14900KS. | not plotted | — |
Why this chart holds one group only
Every point on a chart is implicitly compared with every other point on it. This chart uses one comparability group: one metric, one device scope, one method . There is no chart of all clock frequencies over time because they are not one quantity.
Points use the release date of the machine a figure belongs to, including figures recorded against a part. A release date applies only to a machine that shipped. What this metric measures.