Changing URL (btcguild.com:8332) to a fixed IP in their cluster fixed the problem. It seems they have a cluster now - each time I ping the server btcguild.com it would give a different ip address in network 78.46/16

Perhaps CGMiner stored the IP address resolved at the first place, expecting response from this IP address, but performed a reconnection and got a different IP address before retrieving initial work, therefore get confused? Just my guess.

Today at my computer CGMiner can't connect to any pool other than a local server, but I am having no problem with phoenix so it is not a networking problem. Is there anyone having the same problem?

No. I can't find anyone to sell me 50~150Ohm resistor in the city I live.
I didn't try Linuxcoin. After all, Linuxcoin is still Linux.

Now, problem solved - just as I predicted, the problem was my Ubuntu was 32-bit! They should really add some documentation on this, most likely to be a design problem in kernel, not ATI's fault.

Now with six cards it boots into scary white vertical lines, how fun is that

Got it working with 4 cards but with 5th/6th card plugged in the problem persists. I will just try linuxcoin.

Boot successfully with 3rd card, but get temperature failed only on third card (flashed back to default BIOS), weird.

Doesn't work. Since it works with 2 cards, I suspect one of the PCI-E x16 extension cable is defective.

Testing them one by one.

Everything was good with my dual 5850 until I added 4 more. Ubuntu Natty simply stuck at a blank screen (without internet connection) after the logo "Ubuntu" is loaded. I tried the following (through recovery mode):

1. Reinstall the driver: tried auto install from Ubuntu, and from the package downloaded from ATI website. Also tried manual install.
2. aticonfig --initial -f --adapter=all
3. aticonfig --initial -f --adapter=0,1,2,3,4,5
4. In /etc/X11/xorg.conf ServerLayout section, change "RightOf ...screen...1", "RightOf ...screen...2..." .... to "RightOf ...screen...0..." and other layouts (also tried set resolution to 800x600)
5. In /etc/X11/xorg.conf ServerLayout section, change "RightOf..." to 0 0

None of them work, I couldn't enter login screen unless I run "aticonfig --initial -f" which only enables the first device. It doesn't solve the problem.
Of course, all cards can be detected by aticonfig (aticonfig --list-adapters), and atiflash in MS-DOS.

GPU-Z can detect all 6 GPUs on Windows.
Motherboard: GA890FXA-UD7

I think a small change in xorg.conf should be able to fix this - for the moment, I am trying to increase the number of adapters one by one. But it seems Ubuntu couldn't even start with only 2 adapters (aticonfig --initial -f --adapters=0,1)

By the way, in Windows GPU-Z can detect all 6 cards. My motherboard is GA-890FXA-UD7.

Everything was good with my dual 5850 until I added 4 more. Ubuntu Natty simply stuck at a blank screen (without internet connection) after the logo "Ubuntu" is loaded. I tried the following (through recovery mode):

1. Reinstall the driver: tried auto install from Ubuntu, and from the package downloaded from ATI website. Also tried manual install.
2. aticonfig --initial -f --adapter=all
3. aticonfig --initial -f --adapter=0,1,2,3,4,5
4. In /etc/X11/xorg.conf ServerLayout section, change "RightOf ...screen...1", "RightOf ...screen...2..." .... to "RightOf ...screen...0..."
5. In /etc/X11/xorg.conf ServerLayout section, change "RightOf..." to 0 0

None of them work, I couldn't enter login screen unless I run "aticonfig --initial -f" which only enables the first device. It doesn't solve the problem.
Of course, all cards can be detected by aticonfig (as seen by aticonfig --list-adapters), and atiflash in MS-DOS.

In windows I have to trick the OS by using some dummy DVI adapters - that's why I don't want to switch to Windows.

Many thanks. Do you mean I can connect the "power on" pin to a ground pin on the same 24 pin connector (or any ground pin on any component)? I know nothing about PSU. Just want to make sure it is safe, so I wouldn't short circuit my PSU.

My PSU refuses to provide power to GPUs if 24pin and 8pin cable are not connected to motherboard. Is there anyway to trick the PSU so I could have a dedicated PSU for GPUs?

Update: I was wrong. The problem was I set something wrong in RBE. Powerstate voltage should not be changed, otherwise GPU BIOS would treat it as invalid value - that's why one card dropped to 400MHz (and in fact it was both cards dropped to 400MHz. The "working" one was only an display error).

After everything was fixed, my cards could stay stable at 970MHz/1.25V, the whole system is draining 497W +-3W from wall. (1200W PSU, 80+ Silver)

725MHz/1.088V

Income from increase in production rate is far greater than increase in cost of electricity, though the power consumption grows exponentially against voltage level and core speed. Temperature should be not a concern - @870MHz the temperature does not even exceed 55 C, with 90% fanspeed.

For now I changed them back to 1.088V/870MHz, waiting for my new 1200W PSU.

*More PSU are needed to compensate the increase in power consumption from GPU.  In my calculation I did not put that into consideration. To find the optimal strategy I have to collect more data.

Problem solved. It's just a tragedy of using cheap PSU. Unplugged one card and OCed the other to 1000MHz/1.25V, running at 398MHz, no problem at all.

Tried Sapphire Trixxx on Windows, same problem, overvolt does not apply to the second card.

I wonder how these people on the forum made it to 400MHash/s. Even with only one single card, my GPU simply freezes when I push it to 365MHash/s at 1.2V 925MHz core

I suggest you to buy a power metre, which can be found in most computer shops (or shopfront of electricity company), cheap and accurate, and you can avoid wasting tons of money on expensive PSU.

My first post

I have seen many people on this forum successfully flashed BIOS of multiple 5850 and got them both working at >1.1V, but I never had any luck even after spending hours on it.

Two Sapphire HD 5850 1GB cards, on PCI-E 2.0 x16@16 slot
Gigabyte 890FX UD7 Board
Thermaltake Lite 700W PSU

RBE + ATIflash works for each of them, but not both - when BIOS of both graphics cards are flashed to 950MHz/1.19V, one of them (shown as GPU 1, the one which my monitor is not plugged into) would then automatically drop to 400MHz core speed and 900MHz memory speed, voltage stays at 1.0V .

I have checked the power consumption of the entire system with a watt meter which the result stays stable at 340W (so there shouldn't be any power shortage). I have also tried different PCI-E slot for both graphic cards but never got both of them successfully working at  950MHz/1.19V. I have also tried a BIOS with lower core clock and same voltage level, but the problem remains.

In Ubuntu aticonfig can't even read temperature or set clock speed on the modded cards (both GPU0 and GPU1) and performance is dropped significantly (both cards seems to be working at 400MHz, mining rate is only around 150Mhz)