显卡型号 | NVIDIA CUDA® Cores | Memory | Memory Bandwidth (bit) | 单精度浮点运算 ( FP32 TFlops) |
---|---|---|---|---|
GTX 1060 6G | 1280 | 6 GB GDDR5 | 192 | 4.4 |
GTX 1070 | 1920 | 8 GB GDDR5 | 256 | 6.5 |
GTX 1080 | 2560 | 8 GB GDDR5X | 320 | 8.8 |
GTX 1080 Ti | 3584 | 11 GB GDDR5X | 484 | 11.5 |
TITAN Xp | 3840 | 12 GB GDDR5X | 547.7 | 12 |
TITAN X | 3584 | 12 GB GDDR5X | 480 | 7 |
P100 (阿里云gn5) | 3584 | 16 GB GDDR5X | 549 | 10.6 |
M40 (阿里云gn4) | 3072 | 12 GB GDDR5 | 288 | 7 |
K520 (AWS G2) | 1536 | 4 GB GDDR5 | 160 | 4.6 |
M4000 | 1664 | 8 GB GDDR5 | 192 | 5.3 |
K80(AWS P2) | 2496 | 12 GB GDDR5 | 280 | 8.7 |
V100(AWS P3) | 5120 | 16 GB GDDR5 | 900 | 14 |
M60(AWS G3) | 2048 | 8 GB GDDR5 | 160 | 9.6 |
P4 (华为Pi1) | 2560 | 8 GB GDDR5 | 256 | 5.5 |
P40 | 3840 | 24 GB GDDR5 | 384 | 12 |
深度学习GPU性能调研
最新推荐文章于 2024-05-21 02:04:55 发布