I recently picked up one of these graphics cards (XFX Radeon HD7990), they dump a boatload of heat into box so there are certain things to consider. Please Register or Log in to view the hidden image! Know your limits Apparently they can operate into temperatures around 85°C safely, the maximum is 95°C and thats where most manufacturers would suggest it's severely overheating and probably has a problem with it's cooling (either from being in a small poorly ventilated box or from a problem with the onboard cooling array) If you get sudden reboots, its' likely that so much heats been dumped into the box that it's raised your CPU temp to the point that it causes a safe shutdown. The card is PCI-E 3 capable, DirectX11.2 (with recent drivers) and is extremely long. (squeezed it into a Aerocool Deepsilence mini-ATX case but only by removing the 2.5" bay and then seating the card before fixing the bay back in position, then it was fiddly fixing the extra Molexs) The cards claimed to use up to 540w's on it's own, so make sure you spec any PSU requirements. Before installing new drivers, delete all of the old ones People using Nividia card's don't have this problem no a days as their installation can actually "clean" first before installing. AMD hasn't quite got that far yet, so you have to uninstall previous drivers, reboot and then remove drivers associated with the GPU cores manually from the System before installing a new driver build. Remember to apply Manual Settings for Catalyst to both GPU's The current AMD Catalyst Control Center allows manual overclocking where it's possible to also "underclock". So reducing core speeds of the GPU and RAM can reduce temperature, what should be noted though is that when you use the manual settings, it's set for one GPU, so remember to mirror the settings to the other core or you'll end up with one GPU burning a lot hotter than the other (which in turn causes the temp to rise in the cooler one due to how the stock cooling system works.) It's a good idea to MAX the fans when you have a workload as the drivers preset curve just doesn't seem to like to throttle up. IF it flickers when you scroll down a webpage: Enable GPU Scaling I did have a screen concern when I hit high temps. I thought I wrecked the board, however it turned out in CCC I'd been using an Overscan method to remove a black border from around the screen output (basically it stretches the output to fit) What I hadn't set however was "Enable GPU scaling", this meant that the rescaling of the screen was being pushed through the CPU and it's Bus causing a bit of a bottleneck. It wouldn't be noticeable normally, however it just meant that when surfing a webpage and scrolling down you'd get a screen tare at the bottom. Enabling the GPU to do the rescaling fixed it. Biosis.... The card has two BIOS slots, one can be flashed apparently and there are some modded BIOS's out there to reduce the voltage used and clock the card in all manner of ways. It's not something for the faint hearted. If all else fails, downscale Currently those involved in Bitcoin/Litecoins can be after these particular cards for Litemining since they crunch quite well. So if it dumps too much heat, downgrade to a single GPU card and sell off the huge monster of a card to a person that wants to hash at it.