Help please (ideas).
I recently took on a second had panda miner with 6 RX470s and 2 RX570s after changing motherboard because of broken PCIe slots I still can’t get GPU 2 & 3 to fire up and mine.
I have:
Refreshed the VBIOS for both cards with stock.
Run miner up with just a single card on the board.
Tried altering the overclock settings.
I have used Polaris free to 1 click the VBIOS and the others are running quite well (I think).
if you have them with custom bios ( edited the stock with polaris bios editor ), you can use these settings:
1168 core clock , 860 core voltage, 800 memory controller voltage, and 2050-2100 mem clock.
Dont use DPM states.
amdmemtweak --REF 20 or 30
Aggressive undervolting on.
i make with these settings 32.22 MHs
it consumes a lot of power coz i got the bad luck with the silicon lottery ( meaning some cards came badly from the manufacturer despite they are the same maker and model )
i meant 1168 any wahy test what fits best. 1186 might be a bit higher and you can get dead gpu meaning you need to lower it or increase a bit the core voltage. but increasing core clock and voltage results in increasing temps and watts so find the sweet spot. also with fan speed, mine runs on 65% to keep the gpu at 53-54 degrees.
I’ve got some 580s on the water so keen to look at those. In New Zealand power is not that cheap. About 0.195c per kWh. So power matters in balance. It is still just profitable before OC. Now it is reasonable. I was quite please at how much Bios Editor Pro reduced the 470s power and confirmed the drop at the wall too. I like you got a GPU that was dropped on its head as a baby… last 570 is hungry for W’s.
i always use polaris bios editor to modify my card. its the free version. i have tested various memory timings and came to the conclusion that the one click bios mod is the best.
with it i can run higher memory clocks. i used to use 2050 because the timings were bad but now i can go over 2100 but it doesnt work better so no need. I see you cards are elphida memory so maybe you can try with 2100+ memory clock , 2120 2150. you have to make trials to see when the card dies and then reduce a bit to stabilize it.
i did some math equations to figure out how much power does the GPU consume and how much money it equals to and how much power per MHs is that and money and came to the conclusion in my case that 1mh is 7 cents while 10 watts is 4 cents so i get it to work with 103-105 watts but with 32.22 MHs. it was working before with 95 watts and 31.1 MHs.
your RX 470 should be working with very low power consumption as far as i know. so its a matter of optimization i think.
That is very insightful. You should write a thread on how to do the calculation. I’m sure others would enjoy.
Everything you say makes sense. I’ll look into the Elphida ranges.
Very happy with outcomes. First time above 175mh with 6 cards…so pleased with that. I’ll try the math myself. I also have solar but take that as variable with weather and therefore leave it out of equation.
my pleasure.
last card runs a bit hot i think , you can try with 1150 or 1168 core clock and if you reduce the core clock you can try also reducing a bit the core voltage to 850 maybe or 845 if you like to test. if it dies then increase it a bit. its all a matter of finding what works best. now that you know the stable settings you can try to add more. the ones with 31 MH might improve a bit if you set 1168 core also on them. keep in mind that core clock is tied to core voltage so a certain level of core voltage can support up to a certain level of core clock. so if you increase the core clock and the card dies, then you need to increase a bit the core voltage to sustain the higher clock or revert back to previous settings of the clock.
for the temps i usually prefer to be not more than 60-65 but i like more under 60. you can set the fan speed to 50-55% stable run and see what happens