try this -g 2 -I 13 --shaders 2048 -w256
use cgminer 2.11.4 do not set thread-concurrency let cgminer detect the best value for your cards, maybe try the last beta drivers 13.3, i get more than 720 kh/s with core1040/memory1750 with vcore unervolted and vmemory overvolted , i did order Arctic extreme cooler tried it on 1 card and it is amazing I am getting around 67 c°with these setting comparing to the other cards that are running @ 86 c° I did order more coolers for the other cards, it is worth every cent I did give.
This worked very well for me. I went from:
I 20
TC 24000
g 1
to
I 13
TC 10240
g 2
With the core clock staying at 1025, memory at 1800, and voltage at 1.063 each GPU went from 675 KH to 720 KH.
I am glad that did help you .
What's weird now is that I went to try and implement the same settings on my rigs downstairs and they only get 650 with almost the exact same settings. The only difference is my upstairs computer has:
"gpu-platform" : "1",
While the downstairs ones have:
"gpu-platform" : "0",
If I try to set the downstairs ones to 1 they won't start cgminer. Guess it is time for more testing.
As soon as I closed MSI Afterburner the hashrate shot right up to 720 like the upstairs cards. Picked up a nice 289 KH between 3 rigs and managed to lower the voltage on two cards for a reduction in power consumption as well.
I also worked to convert my rig with 4x7870 LE Tahiti(s) to mine LTC. Results are mixed as I can get the GPUs up to 500 KH where they theoretically should be at, but with 90% HW errors or 390 KH with zero HW errors. The things are a pain in the ass.
So, my Pool allows me to set 'difficulty' from 32, 64 and 128. What should I choose GPU mining with 78xx/79xx? What/how does it matter?
Basically, one accepted share of difficulty 32 is equal to 32 shares on difficulty 1. Set it as low as you can as when you have rejected shares you lose 128 shares versus losing 32. You may lose more 32 difficulty as they get submitted more quickly than those of a higher difficulty. From my not exact science of testing that having a lower difficulty is slightly more profitable on pools with higher reject ratios.