no .bin in the folder = cgminer was not able to compile the kernel cause amdocl crashed - your initial crash, happens to me on 13.4
amdocl.dll and OpenCL.dll and some .exe in System32 left after cleanup for my system means that Windows installed drivers from auto-update -> disable installing drivers from internet and cleanup cause those files will be used every time not the driver original libraries.
on my system I get best results for 7950 using 12.6 driver
and ofc no Catalyst install - driver only, AMD External events service - disabled

setx GPU_MAX_ALLOC_PERCENT 100 - allows to use all GPU memory
setx GPU_USE_SYNC_OBJECTS 1 - no idea
setx DISPLAY ":0" - allows w2k8 not to use dummys

check if were set - regedit, HKCU/Enviroment

I am using tc 44000
Intensity 20, if you are getting errors on higher intensity - tc is too low.
I overclock with AfterBurner, make profile and auto set it for both 2d and 3d.
On w2k8 AfterBurner auto-starts with low prio, so I've disabled auto-start in AfterBurner and created scheduled task manually with different name so AfterBurner won't delete it.

Is voltage regulation available on *nix?

Just because you did not have issues - does not mean powered risers are FUD ; )
Yesterday my friend burnt couple of non-powered 16x-1x risers and before that mb and the same conf with powered risers have no issues.

Also using powered risers makes perfect sense technically speaking unless you have proof that all the GPUs are using no more than 10W from PCI-E slot and that all the motherboards are capable providing 10W on each PCI-E slot(or any other W combination which won't burn mb power regulator).

p.s.
I'm also running non-powered risers on all my rigs with no issues.

edit 2013/07/28: After running ~2months 24/7 my rig burned ; ) PSU ATX motherboard connector to be exact.

yeah.. the easiest way to measure temperature is using your finger ; )))
unless you short something and you won't have some static on you and temperature is below 50c it will be okay

imagine 5 alcoholics(GPU) drinking beer from a small beer keg(motherboard voltage regulator) connected with tiny in diameter straw(the stuff which burnt) to the huge beer canister(PSU)

so.. since alcoholics drink too fast, the tiny straw will burst cause of pressure; )))

now.. would not it be better to give each alcoholic a straw and plug each straw directly into huge canister(GPU+risers)?

I'm sorry.. but imo mining on hackintosh.. seems like the dumbest thing ever(no offense).

also, cgminer works for me on mac mini.

though, you can try to perform a finger test : D
plug 1 7870, let it run for a while and measure temperature on the ATX power connector and motherboard power elements. If it is ok - add one more 7870 and so on..

mm.
See, GPUs are using not only the power supplied by PCI-E power cables(http://www.hardcore-games.tk/images/psu/pcie6.jpeg) but also the power from PCI-e slot itself - http://en.wikipedia.org/wiki/PCI_Express#Pinout, see there are +12V power lines?

So what happened to your rig -
motherboard takes +12V from PSU using the ATX power plug(the one which burned)
motherboard then does something with that power(or not) using voltage regulator on the motherboard and gives it to GPUs on the PCI-e slot.
And your GPUs took too much power from PCI-e slots - so the weakest link burnt, usually the weakest link is the ATX power connector itself.

Now to anwser your question(replace 7870 with 7850) we need to know exactly how much power 7850/7870 are using from PCIe slot itself and for how much max power your ATX connector is rated and assuming that not only the connector burnt but also motherboard voltage regulator also - how much max power it can give.

That's a lot of questions ; )) That's why ppl usually use powered risers - consuming +12V directly from PSU and not using it from motherboard.

DIY PCB - http://www.users.senet.com.au/~astrohog/EasyDIYpcb/
U probably meant something else, right? ; )

totally, make sure there are no shorts and don't turn the rig on with the same config(5x7870).

He meant this - http://www.google.com/search?q=M2A581&rls=com.microsoft:en-za:IE-SearchBox&oe=&um=1&ie=UTF-8&hl=en&tbm=isch&source=og&sa=N&tab=wi&ei=TsS4UZfLD4jK4ASOrIG4Bg&biw=1440&bih=704&sei=UMS4UaF2yffhBJ7DgeAL

see, there is a molex power above the first PCI-e slot.

Clean - take all the melted plastic away and make sure there is a contact between those pins that have burned.

It burned because there were too much Amps going, powered risers would have prevented this.
It was not PSU problem.

clean it and plug again, it's probably alive ; )

thread concurrency is too high.
Also on 13.4 cgminer can't create .bin for some reason on my rigs, 13.1 works, check if there are *.bin files in your cgminer folder to confirm.

Win 2k8 R2 ;p

But seriously, OS does not matter. The straightness of hands does matter
I've mined on CentOS, Windows Server 2008 R2, Windows 7 x64/x86, MacOS 10.6.8, Windows Xp x64/x86, Windows Server 2003 x64/x86

p.s.
MacOS 10.6.8 - poor OpenCL
CentOS - no voltage control(?)
Windows Server 2003 - problems with OpenCL installation
Windows Server 2008 - OpenCL must be installed manually + setx DISPLAY ":0" for headless - takes a lot of HDD space ~30GB

I'd go for USB *nix(BAMT f.e.) if I'd be certain that I could change voltage either in OS or in BIOS.

Ok, I've successfully tested 6GPUs setup.

Gigabyte GA-990FXA-UD3 rev. 3.0
2x6950 + 4x7950 connected with x16_x1 PCIe non-powered risers
Windows 2k8 R2

Catalyst 12.6 - driver only, AMD service disabled(Device Manager-> right click on device -> update driver -> point to unpacked Catalyst folder)
After driver installation I've run OpenCL.msi, located in Catalyst_12-6/Packages/Apps/OpenCL64/ <- this probably is not needed for Windows 7, but on server OpenCL never installs with the driver for me.

cgminer 3.1.1 / 3.2.0

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
setx DISPLAY ":0" <- without this windows server won't start any GPU.


edit 2013/07/28: After running ~2months 24/7 my rig burned ; ) PSU ATX motherboard connector to be exact.

AMD publishes PSU requirements like this because if they said PSU must be able to deliver ~20A on +12V rail nobody would understand what they mean ; )
Also because PSU manufacturers often lie about their PSU specs.

I'm running 3x7950 + 2x6950 mining LTC with Enermax 1250W works fine. One of 7950 was even using ~320W until I've noticed it and lowered voltage.

PSU = Power Supply Unit ; )

Throw your PSU away immediatly, it's very close to being defective! ))
Even at idle it is very low voltage.
11.43 is barely within the specs +/-5%, that's between 12.6 and 11.4
I suspect that capacitors inside the PSU are bulged or will be soon. Their ESR is already through the roof probably.

As for PSU power, I have PC with Core i5(dunno) + GTX 560(150W off. spec) + HD 7950 220W(measured by me) + Thermaltake 600W PSU.
I've tested it - 7950 is hashing LTC at ~600 khs, GTX 560 - Furmark and Core i5 - prime95, +the lowest value for 12V in both GPU-Z and SpeedFan is 11.95V
I've even tried swapping GTX 560 with 6950 (220W) and 12V was showing ~11.90V
I've confirmed readings with multimeter and run tests for several hours.

So. You need good working PSU, not powerfull. I can recommend Seasonic or Enermax, they are expensive though.

works now

+1