My way of testing for stability has been like this with 290X cards:
Unigine Valley until it reaches max temps, then leave it running for 30min. Takes a while to get to max temp if cards are watercooled.
If it passes, then I just fire up BF4 and play couple of runs or until it crashes.
Valley has been great for checking initial stability and upping the clocks, because it really pushes the gpu on shader / ram / power areas.
BF4 is better for checking if the clocks are really game stable.
OCCT / Furmark are bad choices for stability testing because they make unrealistic load on the gpu's. That's why they are defined as power viruses.
My completely game stable clocks (1200/1600) gives me a Black Screen on my 290X CF and a couple of seconds later, my PSU shuts down the moment I fire OCCT / Furmark.
My 1200w PSU can't keep up when power virus pays a visit
BF4 usage is at around 1100w, measured from socket. Valley eats 1050-1080w.
Setup: 3930K, 2x290X CF, 1200w Strider Gold PSU
I actually did test Furmark vs Unigine Valley when these cards were released and it was interesting journey so to speak:
Unigine Valley settings: Ultra / 1600p / 8XAA
Power usage measured from socket.
Valley : 345w
(Powertune 0% & 50%)
Core: 1200Mhz @ 1.337v (real: 1.25v), Powertune: 50%
Valley : 415w
Furmark : 590w !!!
Those test were done with 2600K (stock clocks) and single 290X (watercooled).