SOLVED:
Having no page file was causing the issues, which have now stopped. Lesson learned, even with SSD, use a page file. For reference it required me to go to 8192mb, which is conveniently twice that of the R9 290s VRAM.
==========================================================
Crossfire is giving me headaches this time around. Together they crash/black screen often in games and tests. Also, they occasionally black screen at the desktop. When they do crash games, it is after it's loaded a map or assets. BF4 for example, let's me work through a map for about a a minute and then gives me a Direct X texture allocation error and occasionally the VRAM error. It then proceeds to crash Windows Explorer when I try to open task manager and says system RAM allocation is out of memory.
Skyrim on the other hand loads up fine and then when I pan the camera it slows to a crawl and all textures go black. If I disable crossfire, it works.
Bioshock Infinite black screen/not responding after about a minute in.
Firefall won't launch.
On the plus side, they have yet to BSOD or turn off the computer/power down.
Alone, each card has passed multiple tests.
I've tried :
-Different combinations of the BIOS switch on each card
-Both cards physically alone in both PCI E slots, and also with them installed simultaneously but Crossfire disabled via software.
-Undervolt/overvolt, underclock/overclock shared or separate profiles
-Several complete driver reinstalls, fresh from 14.3 to 14.7
-Checked all the fans, cooling solutions and monitored no excessive or new high temperatures
-Installed CCC to ensure Catalyst AI was on standard instead of advanced setting
-Disabled any overlays AB, RP etc.
-Run on single monitor
I thought maybe it was PSU related, but wouldn't I be facing constant and instantaneous crashing rather than a delayed fail? Also would I not see more BSODs and complete computer shut-offs if it was PSU? Additionally, Heaven benchmark runs and other synthetics do not crash....
I read somewhere that 290s had a problem with supplying enough voltage when used in PCI E 2.0 configurations. It just so happens my board runs both these cards at 2.0 x8
It seems unlikely, though just in case anyone got a PCI E 3.0 board in England I could borrow to test?