Table 3.
NVIDIA model | Architecture | CUDA cores | Clock rate (MHz) | Memory (GB) | SP throughput (Gflop/s) | ≈ Price (€) (net) | |
---|---|---|---|---|---|---|---|
Tesla K20X[a] | Kepler GK110 | 2688 | 732 | 6 | 3935 | 2800 | ✓ |
Tesla K40[a] | Kepler GK110 | 2880 | 745 | 12 | 4291 | 3100 | ✓ |
GTX 680 | Kepler GK104 | 1536 | 1058 | 2 | 3250 | 300 | ✓ |
GTX 770 | Kepler GK104 | 1536 | 1110 | 2 | 3410 | 320 | ✓ |
GTX 780 | Kepler GK110 | 2304 | 902 | 3 | 4156 | 390 | ✓ |
GTX 780Ti | Kepler GK110 | 2880 | 928 | 3 | 5345 | 520 | ✓ |
GTX TITAN | Kepler GK110 | 2688 | 928 | 6 | 4989 | 750 | ✓ |
GTX TITAN X | Maxwell GM200 | 3072 | 1002 | 12 | 6156 | – | |
Quadro M6000 | Maxwell GM200GL | 3072 | 988 | 12 | 6070 | Figure 5 | |
GTX 970 | Maxwell GM204 | 1664 | 1050 | 4 | 3494 | 250 | – |
GTX 980 | Maxwell GM204 | 2048 | 1126 | 4 | 4612 | 430 | ✓ |
GTX 980+ | Maxwell GM204 | 2048 | 1266 | 4 | 5186 | 450 | ✓ |
GTX 980‡ | Maxwell GM204 | 2048 | 1304 | 4 | 5341 | 450 | ✓ |
The upper part of the table lists HPC‐class Tesla cards, below are the consumer‐class GeForce GTX cards. Checkmarks (✓) indicate which were benchmarked. For the GTX 980 GPUs, cards by different manufacturers differing in clock rate were benchmarked (+ and ‡ symbols).