opengl:cpuflops
CPU FLOPS
CPU FLOPS
| CPU | FLOPS (32bit fp) | Clock | core | op/clock | op | |
|---|---|---|---|---|---|---|
| PC Haswell Core i7-5960X | 768.0 GFLOPS | 3.0GHz | 8 | 256 | AVX: 2(mad) x8(simd) x2(unit) x8(core) x3.0(clock) = 768 GFLOPS | |
| PC Haswell Core i7-5820K | 633.6 GFLOPS | 3.3GHz | 6 | 192 | AVX: 2(mad) x8(simd) x2(unit) x6(core) x3.3(clock) = 633.6 GFLOPS | |
| PC Haswell Core i7-4790K | 512.0 GFLOPS | 4.0GHz | 4 | 128 | AVX: 2(mad) x8(simd) x2(unit) x4(core) x4.0(clock) = 512 GFLOSP | |
| PC Haswell Core i7-4770K | 448.0 GFLOPS | 3.5GHz | 4 | 128 | AVX: 2(mad) x8(simd) x2(unit) x4(core) x3.5(clock) = 448 GFLOPS | |
| PC Sandy Bridge Core i7-2770K | 224.0 GFLOPS | 3.5GHz | 4 | 64 | AVX: 8(simd) x2(unit) x4(core) x3.5(clock) = 224 GFLOPS | |
| PS3 | Cell | 217.6 GFLOPS | 3.2GHz | 8 | 68 | SPU: 2(mad) x4(simd) x7(core) x3.2(clock) = 179.2<html><br></html> VMX: 2(mad) x4(simd) x1(core) x3.2(clock) = 25.6<html><br></html>FPU: 2(mad) x2(unit) x1(core) x3.2(clock) = 12.8 = Total 217.6 GFLOPS |
| PC Sandy Bridge Core i7-2720QM | 140.8 GFLOPS | 2.2GHz | 4 | 64 | AVX: 8(simd) x2(unit) x4(core) x2.2(clock) = 140.8 GFLOPS | |
| PC Ivy Bridge Core i7-3615QM | 147.2 GFLOPS | 2.3GHz | 4 | 64 | AVX: 8(simd) x2(unit) x4(core) x2.3(clock) = 147.2 GFLOPS | |
| PC Kaveri A10-7850K | 118.4 GFLOPS | 3.7GHz | 2 | 32 | AVX: 2(mad) x8(simd) x2(module) x3.7(clock) = 118.4 GFLOPS | |
| Xbox 360 | Xenon | 115.2 GFLOPS | 3.2GHz | 3 | 36 | VMX: 2(mad) x4(simd) x3(core) x3.2(clock) = 76.8<html><br></html> FPU: 2(mad) x2(unit) x3(core) x3.2(clock) = 38.4 = Total 115.2 GFLOPS |
| Xbox One | Jaguar | 112.0 GFLOPS | 1.75GHz | 8 | 64 | AVX: 4(simd) x2(unit) x8(core) x1.75(clock) = 112.0 GFLOPS |
| Apple A9 | Twister | 108.0 GFLOPS | 2.25GHz | 2 | 48 | NEON: 2(mad) x4(simd) x3(unit) x2(core) x2.25(clock) = 108.0 GFLOPS |
| Apple A9 | Twister | 88.8 GFLOPS | 1.85GHz | 2 | 48 | NEON: 2(mad) x4(simd) x3(unit) x2(core) x1.85(clock) = 88.8 GFLOPS |
| APQ8084 | Krait 450 | 86.4 GFLOPS | 2.7GHz | 4 | 32 | NEON: 2(mad) x4(simd) x4(core) x2.7(clock) = 86.4 GFLOPS |
| PC Nehalem Core i7-920 | 85.1 GFLOPS | 2.66GHz | 4 | 32 | SSE: 4(simd) x2(unit) x4(core) x2.66(clock) = 85.12 GFLOPS | |
| PC Ivy Bridge Core i5-3210M | 80.0 GFLOPS | 2.5GHz | 2 | 32 | AVX: 8(simd) x2(unit) x2(core) x2.5(clock) = 80.0 GFLOPS | |
| MSM8994 | Cortex-A15<html><br></html>Cortex-A7 | 76.0 GFLOPS | 2.0GHz<html><br></html>1.5GHz | 4+4 | 32+8 | NEON: 2(mad) x2(simd) x2(unit) x4(core) x2.0(clock) = 64<html><br></html> NEON: 2(mad) x1(simd) x4(core) x1.5(clock) = 12 Total 76 |
| PC Sandy Bridge Core i5-2415M | 73.6 GFLOPS | 2.3GHz | 2 | 32 | AVX: 8(simd) x2(unit) x2(core) x2.3(clock) = 73.6 GFLOPS | |
| MSN8974 | Krait 400 | 73.6 GFLOPS | 2.3GHz | 4 | 32 | NEON: 2(mad) x4(simd) x4(core) x2.3(clock) = 73.6 GFLOPS |
| Apple A8X | Typhoon | 72.0 GFLOPS | 1.5GHz | 3 | 48 | NEON: 2(mad) x4(simd) x2(unit) x3(core) x1.5(clock) = 72.0 GFLOPS |
| Exynos 5 Octa | Cortex-A15<html><br></html>Cortex-A7 | 71.2 GFLOPS | 1.9GHz<html><br></html>1.3GHz | 4+4 | 32+8 | NEON: 2(mad) x2(simd) x2(unit) x4(core) x1.9(clock) = 60.8<html><br></html> NEON: 2(mad) x1(simd) x4(core) x1.3(clock) = 10.4 Total 71.2 |
| MSN8974 | Krait 400 | 70.4 GFLOPS | 2.2GHz | 4 | 32 | NEON: 2(mad) x4(simd) x4(core) x2.2(clock) = 70.4 GFLOPS |
| Tegra K1 | Cortex-A15 | 70.4 GFLOPS | 2.2GHz | 4 | 32 | NEON: 2(mad) x2(simd) x2(unit) x4(core) x2.2(clock) = 70.4 GFLOPS |
| PC Athlon 5350 Kabini Jaguar | 64.0 GFLOPS | 2.0GHz | 4 | 32 | AVX: 4(simd) x2(unit) x4(core) x2.0(clock) = 60.8 GFLOPS | |
| Tegra 4 | Cortex-A15 | 60.8 GFLOPS | 1.9GHz | 4 | 32 | NEON: 2(mad) x2(simd) x2(unit) x4(core) x1.9(clock) = 60.8 GFLOPS |
| Tegra 4 | Cortex-A15 | 57.6 GFLOPS | 1.8GHz | 4 | 32 | NEON: 2(mad) x2(simd) x2(unit) x4(core) x1.8(clock) = 57.6 GFLOPS |
| Tegra K1 | Denver | 55.2 GFLOPS | 2.3GHz | 2 | 24 | NEON: 2(mad) x4(simd) +4(add) x2(core) x2.3(clock) = 55.2 GFLOPS |
| PC Atom J1900 (Silvermont) | 48.0-57.6 GFLOPS | 2.0-2.41GHz | 4 | 24 | SSE: (4(simd add) + 2(simd mul)) x 4(core) x 2.0(clock) = 48.0GFLOPS (TB 57.6 at 2.41GHz) | |
| APQ8064 | Krait | 48.0 GFLOPS | 1.5GHz | 4 | 32 | NEON: 2(mad) x4(simd) x4(core) x1.5(clock) = 48.0 GFLOPS |
| Apple A7 | Cyclone | 44.8 GFLOPS | 1.4GHz | 2 | 32 | NEON: 2(mad) x4(simd) x2(unit) x2(core) x1.3(clock) = 44.8 GFLOPS |
| Apple A7 | Cyclone | 41.6 GFLOPS | 1.3GHz | 2 | 32 | NEON: 2(mad) x4(simd) x2(unit) x2(core) x1.3(clock) = 41.6 GFLOPS |
| PC Core2 Duo P8600 | 38.4 GFLOPS | 2.4GHz | 2 | 16 | SSE: 4(simd) x2(unit) x2(core) x2(clock) = 32 GFLOPS | |
| PC Core2 Duo P7350 | 32.0 GFLOPS | 2.0GHz | 2 | 16 | SSE: 4(simd) x2(unit) x2(core) x2(clock) = 32 GFLOPS | |
| MT8135 | Cortex-A15<html><br></html>Cortex-A7 | 28.8 GFLOPS | 1.5GHz<html><br></html>1.2GHz | 2+2 | 16+4 | NEON: 2(mad) x2(simd) x2(unit) x2(core) x1.5(clock) = 24<html><br></html> NEON: 2(mad) x1(simd) x2(core) x1.2(clock) = 4.8 Total 28.8 |
| Tegra 3 | Cortex-A9 | 27.2 GFLOPS | 1.7GHz | 4 | 16 | NEON: 2(mad) x2(simd) x4(core) x1.7(clock) = 27.2 GFLOPS |
| Exynos 5D | Cortex-A15 | 27.2 GFLOPS | 1.7GHz | 2 | 16 | NEON: 2(mad) x2(simd) x2(unit) x2(core) x1.7(clock) = 27.2 GFLOPS |
| Apple A6X | Swift | 22.4 GFLOPS | 1.4GHz | 2 | 16 | NEON: 2(mad) x4(simd) x2(core) x1.4(clock) = 22.4 GFLOPS |
| Tegra 3 | Cortex-A9 | 19.2 GFLOPS | 1.2GHz | 4 | 16 | NEON: 2(mad) x2(simd) x4(core) x1.2(clock) = 19.2 GFLOPS |
| MSN8260 | Scorpion | 19.2 GFLOPS | 1.2GHz | 2 | 16 | NEON: 2(mad) x4(simd) x2(core) x1.2(clock) = 19.2 GFLOPS |
| Atom Z2560 | Saltwell | 19.2 GFLOPS | 1.6GHz | 2 | 12 | SSE: (4(simd add) + 2(simd mul)) x2(core) x1.6(clock) = 19.2 GFLOPS |
| PC Atom Z540 | Bonnell | 11.2 GFLOPS | 1.86GHz | 1 | 6 | SSE: (4(simd add) + 2(simd mul)) x1.86(clock) = 11.16 GFLOPS |
| OMAP 4430 | Cortex-A9 | 9.6 GFLOPS | 1.2GHz | 2 | 8 | NEON: 2(mad) x2(simd) x2(core) x1.2(clock) = 9.6 GFLOPS |
| PC Atom N270 | 9.6 GFLOPS | 1.6GHz | 1 | 6 | SSE: (4(simd add) + 2(simd mul)) x1.6(clock) = 9.6 GFLOPS | |
| MT8125 | Cortex-A7 | 9.6 GFLOPS | 1.2GHz | 4 | 8 | NEON: 2(mad) x1(simd) x4(core) x1.2(clock) = 9.6 GFLOPS |
| Apple A5 | Cortex-A9 | 8.0 GFLOPS | 1.0GHz | 2 | 8 | NEON: 2(mad) x2(simd) x2(core) x1.0(clock) = 8.0 GFLOPS |
| QSD8250 | Scorpion | 8.0 GFLOPS | 1.0GHz | 1 | 8 | NEON: 2(mad) x4(simd) x1.0(clock) = 8.0 GFLOPS |
| Raspberry Pi 2 | Cortex-A7 | 7.2 GFLOPS | 0.9GHz | 4 | 8 | NEON: 2(mad) x1(simd) x4(core) x0.9(clock) = 7.2 GFLOPS |
| PS2 | Emotion Engine | 6.2 GFLOPS | 0.299GHz | 1 | 20? | |
| Apple A4 | Cortex-A8 | 4.0 GFLOPS | 1.0GHz | 1 | 4 | NEON: 2(mad) x2(simd) x1.0(clock) = 4.0 GFLOPS |
| Tegra 2 | Cortex-A9 | 4.0 GFLOPS | 1.0GHz | 2 | 4 | VFP: 2(mad) x 2(core) x 1(clock) = 4.0 GFLOPS |
| PSP | MIPS 4000 | 2.6 GFLOPS | 0.333GHz | 1 | 8 | SIMD: 2(mad) x4(simd) x0.333(clock) = 2.664 GFLOPS |
| iPhone 3GS | Cortex-A8 | 2.4 GFLOPS | 0.6GHz | 1 | 4 | NEON: 2(mad) x2(simd) x0.6(clock) = 2.4 GFLOPS |
| Xbox | Pentium 3 | 1.5 GFLOPS | 0.733GHz | 1 | 2 | SSE: 2(simd) x0.733(clock) = 1.466 GFLOPS |
| Dreamcast | SH4 | 1.4 GFLOPS | 0.2GHz | 1 | 7 | SIMD: 7(dotproduct 4mul + 3add) x 0.2(clock) = 1.4 GFLOPS |
| Raspberry Pi | ARM1176JZF-S | 0.7 GFLOPS | 0.7GHz | 1 | 1 | VFP: 2(mad) x 0.5(cycle) x 0.7(clock) = 0.7 GFLOPS |
| iPhone 3G | ARM1176JZF-S | 0.4 GFLOPS | 0.412GHz | 1 | 1 | VFP: 2(mad) x 0.5(cycle) x 0.412(clock) = 0.412 GFLOPS |
理論上のピーク値です。実際にこの数値が出るわけではありません。表には間違いが含まれている可能性があります。
測定できるものは下記のように実測にて mad(fma) の bit 幅と Unit 数を推測しています。
計測アプリを公開しました。
参考
FPU
下記ページに移動しました
opengl/cpuflops.txt · 最終更新: 2016/04/08 21:25 by oga