Having an issue... As I good remember I didnt read about that here before. I have 3 rigs 1x 7xSapphire RX580Nitro+LE / 1x 7xSapphire RX580Nitro+OC / 1x 3xSapphire RX580Nitro+OC. All cards are bios modded.
I use the same settings in all my rigs and DUAL mining ETH+XVG.
mem: 2200mhz / 885mv
core: 1180mhz / 885mv
Today I check all my rigs with HWinfo and on my 1st rig with LE cards, all GPUs has a bunch of memorie errors. Other 2 rigs has zero or a few mem errors. Im play with mem/core/mv settings and try all possible but no change in mem errors. The other thing is that Claymore shows me 300-400 invalid shares (if you see this offen make sure that you dont overclock it to much, blablabla) on #GPU3 in 24 hours (did not do that earlier)
Any idea?
I use the same settings in all my rigs and DUAL mining ETH+XVG.
mem: 2200mhz / 885mv
core: 1180mhz / 885mv
Today I check all my rigs with HWinfo and on my 1st rig with LE cards, all GPUs has a bunch of memorie errors. Other 2 rigs has zero or a few mem errors. Im play with mem/core/mv settings and try all possible but no change in mem errors. The other thing is that Claymore shows me 300-400 invalid shares (if you see this offen make sure that you dont overclock it to much, blablabla) on #GPU3 in 24 hours (did not do that earlier)
Any idea?
EDIT:
Beleive me or not a simple rig cleanup solve my problem with incorect shares on #GPU3 and now I have much less memorie errors in HWinfo
Another thing people seem to miss is that even though a card can hold a specific overclock for graphics work, that doesn't mean that running in almost 100% compute mode it is the best setting. Typically we stress the cards much more than even the best graphics-based benchmarking tools by running the compute intensive miner 24x7 on our GPU cards.
Just using my single-card rig as an example: it's a GTX1060/6GB Gigabyte card that will easily hold +230 core clock offset, but I have backed it down to +220 and later +210 because Claymore is much more stable and tends to get better quantity of shares at that setting. It would run with +225 forever, but it would get those messages about invalid shares, and regardless of what many people think - 99% of the problems we're having trace back to one of two problems: either OC is set too high or there is a problem with a riser card(s). Some of them are better than others, and if you run into a batch with problems it is likely going to affect the whole group, not just one riser. That other 1% can be elusive, but almost every problem can be cured by backing off overclock by 5% or swapping out the riser for the "problem card".
