I've tried three different versions of nvOC now and I still just CAN'T GET THE OVERCLOCKS TO WORK!
I've tried individual and whole system and I even see the overclocks apply when onebash starts - but they aren't sticking.
I'm mining eth and in Windows with Afterburner I set all 5 cards (2 1070s and 3 1060s) to core +75 and mem +800 and I get 132mh/s
Using nvOC and trying to set the same overclocks, I get 120mh/s.
It's really frustrating. I just don't understand. Does Claymore set everything to factory clocks when it starts or something?
Anyone else mining ETH and managing to ue overclocking?
Also linux OC offsets are scaled differently than windows; you will need to use higher offsets to get the same results in linux.
I am mining ETH and am overclocked (~30.9 MH/sec on 1070 Founders Ed cards and ~29.5 MH/sec on a frankenstein rig with 3 different kinds of 1070 cards).
I recommend using v0016 if you aren't already and trying the OC settings:
__CORE_OVERCLOCK=-100
MEMORY_OVERCLOCK=1000
if this is stable I would try:
__CORE_OVERCLOCK=-100
MEMORY_OVERCLOCK=1100
then:
__CORE_OVERCLOCK=-100
MEMORY_OVERCLOCK=1200
... until you get a soft crash in Claymore, then I would backtrack to the previous setting
OK... I have a very weird (at least to me) situation. I'm using v0016 and I followed your suggestion of trying memory overclocks of 1000, 1100 & 1200.
I do see an improvement. previously I saw a hashrate of 120mh/s compared to Win8.1 / MSI Afterburner (core -200 memory +800) getting me around 132mh/s.
Now when I tried your suggestion nvOC is giving me around 127mh/s but the weird thing is - I get this hashrate regardless of whether it's based on memory 1000, or 1100 or 1200. The same improvement. This suggests that your memory overclock is just somehow selecting some arbitrary amount whether I tell it 1000, 1100 or 1200. Not some precise amount which increases as I raise the number. I know the actual memory figure isn't raising it by 800 or 1000 or 1100 or whatever, because the (actual) memory is running in the 5000s.
It would be nice to be able to set a precise number and see a precise gain. It would be especially nice to be able to at least equal (or even improve) what I can achieve in Windows, which seems to be about 132mh/s - not 127mh/s which is what nvOC is reporting. I know it's only 5mh/s but don't we all want the highest hash-rate we can achieve?
I guess I also must consider power consumption. If nvOV is delivering 127mh/s but at much lower power than Windows at 132mh/s of course I will use it.
I just wish it was clear to me what these overclocking setting numbers actuually mean.