[...]
Nornally I use ubuntu to mine. But since I didn't want to screw up all the hard work I have put in making the ubuntu work rather seamlessly I prefered making the test under Windows 7 wich is my secondary boot partition. It might be possible to do it under an *nix environment but I don't want to lose anymore time pausing my rig to install/unistall etc etc, specially because I will have to forcibly shut down the rig due to a melted connector in the PSU...No idea how that happned!!!... so if anyone is willing to do it in ubuntu/linux/xubuntu, that would be very welcomed and valuable to the community.
Nornally I use ubuntu to mine. But since I didn't want to screw up all the hard work I have put in making the ubuntu work rather seamlessly I prefered making the test under Windows 7 wich is my secondary boot partition. It might be possible to do it under an *nix environment but I don't want to lose anymore time pausing my rig to install/unistall etc etc, specially because I will have to forcibly shut down the rig due to a melted connector in the PSU...No idea how that happned!!!... so if anyone is willing to do it in ubuntu/linux/xubuntu, that would be very welcomed and valuable to the community.
I too use Ubuntu for mining, too lazy to port all this GPU stuff to FreeBSD
The PSU problem -- was it melted at the GPU side? Those connectors are very bad indeed. The original Molex connectors are sort of reasonable, but the clones are... just absurd. There are several different specs for the connectors, especially the leads inside, some for very low currents --- and sadly most PSU/cards come with the lowest spec connectors!
I started doing my own cables and PSU (based on the IBM BladeCenter 2000W PSU, my power distribution board is now done) -- and it is very obvious why they do it -- the real good connectors (9A per pin) are noticeably more expensive than the fakes.
Will be useful, if you can do some charts for performance/power consumption. Also, some "real world" measurement of hashing performance.
Anecdote: Just now waiting for my car to go out of inspection and, asked the technicians to do tests to identify why it would show higher fuel consumption in the city -- their answer was: "you were looking at the board computer measurement, right? those computers usually lie, something might have happened and it no longer shows real numbers. you should do an full tank test to measure your actual fuel consumption".
They are indeed much more efficient. 55W for 7 Mhash.. plus they are close to half the size of the 280x and not to mention the non-existant heat. As for the PSU, indeed, a 860W or even a 600W PSU would suffise for 6 GPUs. It wil be a blessing for the electricity bill!!
No it was melted at the PSU side. Something extremely strange though as I hadn't noticed it before and 2 of the PCI-E (empty) connectors were partially melted. Not even where the GPUs were connected.....
... go figure! Me thinking someone/somehow/at some point disconnected the melted connector and plugged it on another port (twice)... If it was me then I could only have done that while sleep walking, which I don't do. 
Anyways whatever the reason I find this really abusive since I bought the top of the line PSUs (2 x Seasonic 860W 80 Plus Platinum rated). They are supposed to be the caviar of the PSUs. Four 280x for two 860W PSU is way more than enough so obviously the problem did not come from me overloading the PSU's.
If I had the knowelegde I would do like you and build my own PSU.
I can try to do some chart but it wont be anything scientific, more of a homemade benchmark. But you can already see without going into too much detail that those 750 Ti are killers!
As for fuel consumption inside the city, that is totally normal. Driving in a city means slowing down, accelerating, braking etc etc much more often than on long distances or highway, which leads to higher fuel consumption. Basically on my 650CC bike I can consume twice more in the city than if I go for long uninterrupted trips outside the city boundaries.
)