For the Sapphire R9 290, I only need to specify threads 32765, gpu memory 1350 and I get 880H/s. Changing core clock makes no difference over time (and you SICK/DEAD if you go too high).
It doesn't like memory 1500 much ether, some cards lock up after a while. 1425 gives 890KH/s but 1350 has been rock solid (1% less KH/s and a good night's sleep vs 1% more and checking all the time...hmmmm

).
I've not had a problem with sick or dead, I've had them clocked at 1000 core 1500 memory for weeks and they never gave an issue like that.
How many shares per min you getting each card ?
Usually I get about 0.8 share/min per card.
Could be a number of things, I'm leaning towards the mobo though (some are on powered risers but that doesn't seem to matter), an Asus P8Z77 WS with a Lucid chip on it (I missed that bit when I bought it, otherwise I would have avoided it). Also, I can't seem to get the rejects below 4-5%, even at I13 (which gives me really poor performance) or I18. Ditto for another rig with 7970/280X (same mobo).