i don’t pay any electricity, so i’m bit more able to push the voltage slightly higher… i found two optimal settings, obviously both with bios modified. Have to say that the bios is just slightly modified, and not pushed to much, as my memory are micron and not samsung, were a bit afraid of bricking the card, so i pushed it just to a double, and not triple values…
the pads have been changed, and thermal paste too
that being said my 2 best options were:
-
(electricity costs 0.00/kwh) core: 1800mhz - voltage: 950mv - mem clock: 1894 - fans 30 to 60% stable at 52% most of the time - thermal junction 79/80c - Tmem 96/98c - 154watts with about 165watts at the wall —> getting 59.4 MH/s stable since 1 month non stop
-
(cost efficient) core 1450mhz - voltage 800mv (you can put 750mv if you’re able to change driver, but in that case you’ll have to low down the clock to 1350/1400mhz) - mem clock 1900 - fans 30 to 60% stable at 52% most of the time - thermal junction ~50c or less - Tmem 98/102c (depending on room temperature you can push them down by putting an extra fan on lateral of the gpu where memory are) - 110watts with 115/120watts at the wall —> getting 57 MH/s stable since 2+ weeks continually
please note also if you’ll be able to low down the voltage to 700/750mv your power usage would be around 80/85watts, obviously the power consumption at the wall will depend on your psu specifications obviously an 80+ white will have much more power consumption at the wall then an 80+ platinum.
what i discovered using the -openclLocalWork parameter is that it is definitely increasing the hashrate, pushing my hashrate to 68.4 mhs wich is impressive, but is not giving any improvement since is getting a lot of invalid shares (20/30% in my case, but many other people reported 40/50% or even more), which makes the end result much less profitable then usual settings