This is strange. Have you tried to run clinfo.exe to see information about your OpenCL setup if its complete? That point of miner close is directly where it should try to detect the OpenCL drivers available.


I 0 sol/s appear (usually) when writing to a non available memory address. There can be two reasons for this: a) either invalid program code / the program really tries to write to a non existing addr or b) some bits flipped so some calculation went inaccurate. The latter often happens with too much OC.
Plz try to run again and see if there is a systematics here / always after some time or always on same GPU. That may indicate slightly too high clocks. E.g. we had a 580 in testing with +10% mem running stable for hours, but at +15% ending up in 0 sol/s after less then 5 minutes and that reproducable.


Old miner speed was better on 1070? Can someone confirm? On my 1080 I went from 16.4k to over 20.5k from 0.34 to 0.4. Wonder (on same code) if 1070 or 1080 scale differently here (I do not have such cards, but if that confirms I can provide the lagacy kernels for the new miner, so newer features can be used with older kernel code).

By the way: right now building up an complete new Windows build chain, switching from MingW64 cross build to Microsoft Visual Studio compilers and runtime libs. I hope this reduces the problems we face in Windows