GPGPUやDeep learning用にGPUを購入する際の参考にするために，GPUのスペックをまとめました．
RTX40シリーズ
| clock (GHz) | CUDA cores | Tensor Cores | メモリGB | メモリ帯域GB/s | Single (Tflops) | Double (Gflops) | Half (Tflops) | TDP (W) | |
|---|---|---|---|---|---|---|---|---|---|
| RTX 4060 | 1.83 (2.460) | 3072 | 96 | 8 | 272.0 | 15.11 | 236.2 | 15.11 | 115 |
| RTX 4060Ti | 2.31 (2.535) | 4352 | 136 | 8 | 288.0 | 22.06 | 344.8 | 22.06 | 160 |
| RTX 4060Ti | 2.31 (2.535) | 4352 | 136 | 16 | 288.0 | 22.06 | 344.8 | 22.06 | 165 |
| RTX 4070 | 1.92 (2.475) | 5888 | 184 | 12 | 504.2 | 29.15 | 455.4 | 29.15 | 200 |
| RTX 4070 SUPER | 1.98 (2.475) | 7168 | 224 | 12 | 504.2 | 35.48 | 554.4 | 35.48 | 220 |
| RTX 4070Ti | 2.31 (2.61) | 7680 | 240 | 12 | 504.2 | 40.09 | 626.4 | 40.09 | 285 |
| RTX 4070Ti SUPER | 2.34 (2.61) | 8448 | 264 | 16 | 716.8 | 44.10 | 689.0 | 44.10 | 285 |
| RTX 4080 | 2.21 (2.51) | 9728 | 304 | 16 | 716.8 | 48.74 | 761.5 | 48.74 | 320 |
| RTX 4080 SUPER | 2.295 (2.550) | 10240 | 320 | 16 | 716.8 | 52.22 | 816.0 | 52.22 | 320 |
| RTX 4090 | 2.23 (2.52) | 16384 | 512 | 24 | 1,008 | 82.58 | 1,290 | 82.58 | 450 |
RTX30シリーズ
| clock (GHz) | CUDA cores | Tensor Cores | メモリGB | メモリ帯域GB/s | Single (Tflops) | Double (Gflops) | Half (Tflops) | TDP (W) | |
|---|---|---|---|---|---|---|---|---|---|
| RTX 3060 | 1.32 (1.777) | 3584 | 112 | 12 | 360 | 12.74 | 199.0 | 12.74 | 170 |
| RTX 3060Ti | 1.41 (1.665) | 4864 | 152 | 8 | 448 | 16.2 | 263.1 | 16.2 | 200 |
| RTX 3070 | 1.5 (1.725) | 5888 | 184 | 8 | 448 | 20.31 | 317.4 | 20.31 | 220 |
| RTX 3080 | 1.44 (1,71) | 8704 | 272 | 10 | 760.3 | 29.77 | 465.1 | 29.77 | 320 |
| RTX 3080Ti | 1.365 (1,665) | 10240 | 320 | 12 | 912.4 | 34.10 | 532.8 | 34.10 | 350 |
| RTX 3090 | 1.395 (1.695) | 10496 | 328 | 24 | 936.2 | 35.58 | 556.0 | 35.58 | 350 |
RTX20シリーズなど
| clock (MHz) | CUDA cores | Tensor Cores | メモリGB | メモリ帯域GB/s | Single (Gflops) | Double (Gflops) | Half (Gflops) | TDP (W) | |
|---|---|---|---|---|---|---|---|---|---|
| GTX 1650 | 1485 | 896 | - | 4 | 128 | 2661 | 83 | 5322 | 75 |
| RTX 2060 | 1365 | 1920 | 240 | 6 | 336 | 5242 | 164 | 10483 | 160 |
| RTX 2060 Super | 1470 | 2176 | 272 | 8 | 448 | 6123 | 191 | 12246 | 175 |
| RTX 2070 | 1410 | 2304 | 288 | 8 | 448 | 6497 | 203 | 12994 | 175 |
| RTX 2070 Super | 1605 | 2560 | 320 | 8 | 448 | 8218 | 257 | 16435 | 215 |
| RTX 2080 | 1515 | 2944 | 368 | 8 | 448 | 8920 | 279 | 17840 | 215 |
| RTX 2080 Super | 1650 | 3072 | 384 | 8 | 496 | 10138 | 317 | 20275 | 250 |
| RTX 2080Ti | 1350 | 4352 | 544 | 11 | 616 | 11750 | 367 | 23500 | 250 |
| Titan RTX | 1350 | 4608 | 576 | 24 | 672 | 12442 | 389 | 24884 | 280 |
GRX10シリーズなど
| clock (MHz) | CUDA cores | Tensor Cores | メモリGB | メモリ帯域GB/s | Single (Gflops) | Double (Gflops) | Half (Gflops) | TDP (W) | |
|---|---|---|---|---|---|---|---|---|---|
| K20c | 706 | 2496 | - | 5 | 208 | 3524 | 1175 | - | 225 |
| GTX750Ti | 1020 | 640 | - | 2 | 86.4 | 1306 | 40.8 | - | 60 |
| k2200 | 1046 | 640 | - | 4 | 80 | 1300 | 40 | - | 68 |
| GTX1050Ti | 1290 | 768 | - | 4 | 112 | 1981 | 62 | - | 75 |
| GTX1060 6G | 1506 | 1280 | - | 6 | 192 | 3855 | 120 | - | 120 |
| GTX1080 | 1607 | 2560 | - | 8 | 320 | 8228 | 257 | - | 180 |
| GTX1080Ti | 1480 | 3584 | - | 11 | 484 | 10609 | 332 | - | 250 |
| Titan V | 1200 | 5120 | 640 | 12 | 652.8 | 12288 | 6144 | 24576 | 250 |