Re: excavator by NiceHash - multi-algorithm advanced NVIDIA CUDA miner [1.1.4a]

I haven't played around with the settings as much as some of you, but my best speed (and OC) is around 648 S/s. Power lim at 120%, +140 core, +500 mem. Temp gets as high as 85c. I think efficiency was somewhere around 2.x - Core was at around 1980 MHz while mem is 5508 MHz. Power draw was around 275W I think. Fan was at around 80%

I found the best efficiency I could do was just leave it at base clock (1480/5005). Dialed the power lim it 63% and leave the core and mem clock at +0. This leaves me with 540 S/s at 3.4 efficiency, 72c (lower fan speed - 42%) and power draw of 157W.

For GTX 1080 Ti, best setting is: +240 core, -300 mem, 60% TDP. You get 630 sol/s (if you keep temps low) at 150 W.

I'm only getting around 580 Sol/s with those settings, but efficiency gone up a bit - now at 3.8. If I drop to 58% TDP, I get 563 Sol/s but my efficiency goes up to 4.

What are your temps? You will not be able to reach 600+ without properly cooled card. In our experiments, all 1080 Tis go to 620+ with these settings (tested on 6 of them) but it is also true that all are water cooled and have around 40 C working temperature.

67c. The 60% TDP seems to be limiting the clock speed.

And temperature, too. Clock is starting to get lower after around 40-45 C. Cool it down to 45 and you will see the difference. Also what we noticed is that slow PCIe interfaces play some role here too; with PCIe 2.0 x1, there is around 10 sol/s missing compared to PCIe 3.0 x1. If you have PCIe 1.1 x1, then you might be loosing even more due to slow interface.