The prototype rig
2 x RX470 8GB Mining Edition GPUs (on Samsung mem) on latest Blockchain AMD drivers, flashed BIOS, etc.
The Claymore .bat file
timeout /t 15 setx GPU_FORCE_64BIT_PTR 0 setx GPU_MAX_HEAP_SIZE 100 setx GPU_USE_SYNC_OBJECTS 1 setx GPU_MAX_ALLOC_PERCENT 100 setx GPU_SINGLE_ALLOC_PERCENT 100 ethdcrminer64.exe -epool us1.ethermine.org:4444 -ewal 0xADDRESS.Rig001 -epsw x -mode 1 -tstop 85 -tt 65 -fanmin 40 -fanmax 100 -cclock 1150 -mclock 2150 -cvddc 850 -mvddc 875 -dcri 5 pause
28.3 Mh/s per GPU. Decently happy with this - haven't pushed the cards very far. Drawing very little power. Pretty efficient.
The system runs very stable. It can go 5-12 hours without hiccups, low temperatures, the whole thing.
Every once in a while, however, there is a
Socket was closed remotely (by pool) error in the console.
After this, there is no recovering the hashrate on one of my GPUs until I do a full driver re-install.
To be clear: restarting Claymore or Windows 10 itself will never get back to my 28.3 Mh/s regular rate.
Here's the Claymore output pinpointing exactly when this happens - and it always happens every day:
Now, I don't care that the socket is closed every now and then.. the recovery seems to be immediate.
...But why on earth do I suffer an irreversible drop in hashrate - particularly in one of my GPUs - after this?
DDU + Driver re-install + patch (with restarts in between) will 100% restore my hashrate.
Very strange, and hard to pass off as being hardware related.
Claymore 10.0. Everything I'm running is pretty much the latest as you'd expect as of Oct 11th 2017.
In ethermine.org you can see the corresponding drop:
I've seen several references online to mysterious drops in hashrate connected to either 1) lost connections, or 2) closing and re-opening Claymore, but I haven't seen anyone track down or identify this issue.