A bit of research would have told you that the hot V-REG with furmark is associated with the 4850/4870 cards not the 5000 cards, so the problem is not associated with all ATI cards.
IT does still put an unnatural load on the v-reg's on a 5series, just doesn't crash it as the 4 series just wasn't designed for such an unnatural power load to be placed on a card. I wouldn't run it on any card, it puts a higher load on Nvidia cards than any game does aswell and even a ridiculously well optimised Cuda app would be hard pushed to match the load furmark places on it.
INcredibly simple and predictible stuff like Furmark, like super pi, is basically as good as it gets in terms of efficiency of code/shader usage and as bad as it can get in terms of power load.
The simple fact is, even more so with AMD, but Nvidia to a large degree aswell, Furmark doesn't replicate real life conditions, being stable or temp stable in Furmark is completely and utterly pointless. Lets say you're testing for stability at 900Mhz, Furmark fails so you downclock, if you tested in games you might never find a game or cuda app, open cl ap, that is unstable at 950Mhz. LIkewise the temps you see in Furmark, especially on the v-reg's, might persaude you to downclock or go no further when in real life situations, gaming, opencl/cuda apps, you could be running much lower temps with huge room for overclocking.
Furmark is completely and 100% useless and serves no purpose, stability in Furmark doesn't prove or disprove your stable speeds in games so simply don't use it.
Not only that but it is inconsistant and has problems fully testing many crossfire and sli setups, it does seem fairly consistant for single card setups being put under absolute maximum loads, but again, theres simply no point to that.
