@crys
can you let us set the difficulty manually? while you try to fix you're automatic difficulty algorith.
Totally agree! The automatic algorithm doesn't work at all. Every Rig get 409.6diff on pool (1.6 in the miner) in 15-30 minutes. While one has 6*1080 and another only 2*1080.
P.S. Right now something weird is shown on the pool all numbers are divided by 10.
agreed ...
but without writing more code into the stratum - this cannot happen just yet ...
but this would be a good addition - and IS on the todo list for us ...
there is just no way to do it right now ...
we will keep trying to tame this aggression - until it is fine with the diff levels ...
#crysx
I think the quickest way to "solve" the problem is to lower maximum diff to 1 for lodiff port. At least it will help fot 2GPU or 3GPU rigs at the moment.
actually ...
great idea ...
this means that the bigger farms will be slamming the connection - BUT - the CWI-Stratum is custom built to handle a LOT of hammering ...
ok ... 1 is a little too low - so 1.24 ( if my calculations are right ) seems to be a fair figure for small and large on the lodiff port ...
get ready for a stratum interruption ... 25 seconds - here we go ...
#crysx
i experimented a lot with ccminer on nova and yiimp with trying to find a sweetspot for my 1070 gpus.
so far this is what i've come up with. hope this helps you dial in the pool.
they like a worker diff to be a +32 number. 32, 64, 96, 128 etc, etc, etc. they seem to mine just a bit better, and submit shares that are just a bit higher with this pattern of diff, vs random numbers like 40.8, or 490.6 etc. they also seem to hit that "home run" just a bit more often. for example, a single 1070 mining at diff of 64 will submit something like five or six low diff shares like 0.6 or 0.4 etc, but then bam - 12 or 26. that "home run" share just seemed to pop up much more often with a +32 number. not a scientific fact, just something i noticed.
so 0.125 is 32, 0.25 is 64, 1 is 256 etc etc. i found that 64 (0.25) was a good long term sweetspot per 1070 (0.375 or 96 did well too). i would love to see how it would perform if it could adjust to the speed/difficulty of the network, so it could be at 0.125 (32) on low diff fast moving blocks, and top out at 0.5 (128) on the slower moving high diff ones, but you can't set a range in ccminer, so i picked the middle of the road value of 0.25 and stuck with it.
i think that 0.125 to 0.5 per gpu (x6 for a 6 card rig) is a very good range for 1070's. anything lower and they thrash too much, and anything higher and they start to choke up. again, not a scientific fact, just an observation, and an observation on a different miner and different pools, so not even sure how valid it would be here.
just trying to help.
...
