There is something strange going on with the 16.12.2 AMD driver (I skipped 16.12.1) and these mining softwares (Claymore ZEC and ETH):
- after a fresh OS boot I rarely (if ever) get all cards (usually 6 per machine) up to full speed, some get stuck at their lowest GPU clock speeds
- restarting the same miner with the same settings usually reproduces the same pattern (the same GPUs remain stuck as before)
- starting a different miner (ETH after stopping ZEC or vice versa) usually draws a different pattern of stuck GPUs (there is overlap but it's not the same)
- starting two processes of the same miner in parallel with 3-3 different cards instead of all 6 usually gets all cards up to full speed
- starting a single process after getting all cards "unstuck" once with parallel processes usually results in all cards running at full speed (even after switching miners)
There is no detectable/reported driver crash (it's no like those card instantly crash and get reset to "safe" speeds, they never get up to full speed).
Sometimes even smaller machines have similar problems, like 1 card getting stuck from 3 (but it's more often for 6 cards and the 3-3 parallel trick is usually enough for 6 cards)