Table 1.
GPU | # of CUDA Cores | FPP (TFLOPS) | MMB (GB/s) | Approximate Price ($) | Reference |
---|---|---|---|---|---|
NVIDIA GeForce GTX Series | |||||
970 | 1664 | 3.4 | 224 | 157 | [65,66] |
980 TI | 2816 | 5.6 | 337 | 250 | [52,67,68] |
1050 | 640 | 1.7 | 112 | 140 | [69,70] |
1050 TI | 768 | 2.0 | 112 | 157 | [71,72,73] |
1060 | 1280 | 3.9 | 121 | 160 | [60,74,75] |
1070 | 1920 | 5.8 | 256 | 300 | [76,77] |
1070 TI | 2432 | 8.2 | 256 | 256 | [78] |
1080 | 2560 | 8.2 | 320 | 380 | [79,80] |
1080 TI | 3584 | 10.6 | 484 | 748 | [81,82,83], etc. |
1660 TI | 1536 | 5.4 | 288 | 290 | [84] |
TITAN X | 3072 | 6.1 | 337 | 1150 | [45,59] |
NVIDIA GeForce RTX Series | |||||
2080 | 4352 | 10.6 | 448 | 1092 | [85,86,87], etc. |
2080 TI | 4352 | 14.2 | 616 | 1099 | [42,88,89], etc. |
TITAN | 4608 | 16.3 | 672 | 2499 | [47,90] |
NVIDIA Tesla Series | |||||
C2075 | 448 | 1.0 | 144 | 332 | [43] |
K20 | 2496 | 3.5 | 208 | 200 | [91] |
K40 | 2880 | 4.3 | 288 | 435 | [92,93] |
K80 | 4992 | 5.6 | 480 | 200 | [94,95] |
P100 | 3584 | 9.3 | 732 | 5899 | [96,97,98] |
NVIDIA Quadro Series | |||||
P2000 | 1024 | 2.3 | 140 | 569 | [99] |
P5000 | 2560 | 8.9 | 288 | 800 | [53] |
NVIDIA Jetson Series | |||||
NANO | 128 | 0.4 | 26 | 100 | [89] |
TK1 | 192 | 0.5 | 6 | 60 | [100] |
TX2 | 256 | 1.3 | 60 | 400 | [89] |
Others | |||||
NVIDIA TITAN XP | 3840 | 12.2 | 548 | 1467 | [101,102,103] |
Cloud server | — | — | — | — | [54,104] |
CPU only | — | — | — | — | [64,105,106], etc. |
Note: GPU is graphics processing unit; CPU is central processing unit; CUDA is computer unified device architecture; FPP is floating-point performance; TFLOPS is tera floating point operation per second; and MMB is maximum memory bandwidth. “—” indicates missing information.