Thank you for your fast answer.
So, to narrow down my options, do you mean, that theoretically if I consider double precision performance also, I could get much better speeds using the FIRST one from the following two options due to double precision performance, even though the second has higher results in single precision calculations?
1. GTX Titan (GK110) released in 2013, 384 bit, 6 GB, single precision: 4500, double: 1500, 2688 CUDA cores, 837 MHz base clock
2. GTX 1070 (GP104) released in 2016, 256 bit, 8 GB, single precision: 5783, double: only 181, 1920 CUDA cores, 1506 MHz base clock
Do you think that 6 GB will be enough for HCP? Why is it important to check the double precision performance?
|