Quote from: ltc_bilic on July 07, 2014, 11:21:44 AM

Today I was testing it, but unfortunately there was not a lot of switching done today, but when I was running it yesterday I remember there were quite a lot of switching happening for 1/2 min back and forth,...so I think you should consider implementing a "5 min buffer", where it would keep track of the new price and it won't switch immediately as of now, but after 5 min if the new price is still better then current,...this would avoid the constant switching if you're specs are very similar for different algos like x11 and x13. Yes when you minimize it you cannot bring up the miner window and you can only manually stop and start in order to see what's going on well it's not an elegant solution. IMO the next step would be to include the miner window on Control Form right next to Activity tab, and print out the last few lines of what's happening, so in this case you would avoid the previous issue and have an elegant overview of what's happening with the miner. Some people would like to see the gpu temp fan speed and all the neat stuff that cuda manager has, but I don't think it's necessary because it's meant to be used with newer generation nvidia cards and those suckers are running cool and quiet Roll Eyes

The time buffer is definitely moving up on my priority list as I was getting hit by the same flickering between x11 and x13 yesterday. A couple of days ago I actually had it go between x11, x13, x15, and Nist5 all in the span of a couple of hours.

My plan is to implement a pre-buffer and a post-buffer for the switching. The pre-buffer will prevent switching for an amount of time in case the algorithm switches and switches back over just a minute or two. The post-buffer will act as a minimum amount of time to mine so that once a switch does occur it won't switch again for that amount of time. If the kid goes to bed early enough the next couple of nights I might have time to work on it during my 9:00 PM to 10:30 PM "me time" window.

Regarding showing a section of the output in the control program, I've thought about that. CUDA Manager does exactly that. This wouldn't be too difficult with cudaMiner and ccminer as they use basic output which can be captured and shown elsewhere. The problem comes in with programs like sgminer or the ccminer-splitscreen branch which has a more interactive display. I have no idea how bad this would interact with a straight dump.

Quote from: ltc_bilic on July 07, 2014, 11:21:44 AM

More important feature for "todo list" would be to the lack of backup pools in ccminer/cudaminer so to detect if there's a problem with hashing and miner starts running on a backup pool. When situation goes to normal tries to switch back to the most profitable nicehash algo.

Nothing I can do there. I could be wrong but cudaMiner may support backup pools already. I know ccminer doesn't but from what I've been reading on the appropriate thread there are a couple of people interested in implementing it.

I've thought about trying to do something from the NiceHash control program to handle switching but there are so many issues it just hasn't been worth looking too deep into so far. Fortunately, the down time for NiceHash has been infrequent enough that it hasn't been a major issue for me.