OK. I am really tired if everything I tried during this week so I am applying a 3 LTC bounty to the first (if anyone!) who can tell me how to get my rig working.
1. Rig specs:
- Mobo: Gigabyte 990FXA-UD7
- GPU: 6 x Gigabyte R9 270
- PSU: 2 x EVGA Nova Gold 750W with add2psu
- CPU: Sempron 145
- RAM: Kingston DDR3 1600Mghz Hyper-X Blue
2. Rig built
1. 3 GPU + everything else connected to my first PSU. These GPUs are using as well powered molex, which are as well connected to the first PSU.
2. The other 3 GPU are connected to the second PSU, 2 of them using not-connected powered risers and one being directly put on the mobo slot.
Why I have not connected the other 3 GPUs with molexes - because I was told that the molexes need to be connected only to the first PSU and I strictly must avoid any mixing of the 12V line between the 2 PSUs.
I was also given an advise to increase the RAM up to 4 GB per card which sounds strange to me as everywhere I read 4 GB for the whole rig must be more than enough. 24 GB sounds crazy for me.
3. The story so far:
1. First I tried to use BAMT. I tried BAMT 1.3, and if I am not wrong as well at least some of the BAMT 1.1, 1.2 and 1.5. The result was when the system boots up I was getting errors messages about problems with GPUs and I needed to shutdown. Reading advises from forums, I restarted few times, but this didn't solve the problem for me.
2. Then I tried BAMT SMOS 1.3 and BAMT SMOS 1.3 Bee edition. I tried both in usual and failsave mode. The result was: System booting up OK, and then, when i start mining with any kind of settings, both copying settings from folks with same GPUs and lower values, everything was freezing in just 2 minutes, the remote panel for monitoring was not loading as well, and I needed to shut down the rig.
3. Then I tried installing other Linuxes. I tried installing: Centos 6.5, Ubuntu 12.04.4, Xubuntu 12.10, Xubuntu 13.10, Linyx Mint 13. As a result I could even not install them at all, with ending up with some kind of strange message: "initframes unable to find a medium containing live system" (Just CENTOS was giving me some other error for a corrupted image). I have tried at least 5 different USB sticks and installs. After googling this error I found out an advise to put the USB into the USB 2.0 slot, which didn't help, as also neither help changins BIOS settings for botting in IDE or AHCI mode. I finally found an article, stating that Linux 64bit can not be installed via USB stick on this mobo.
Then I bought a DVD-ROM. I recorded the Linux on a DVD and tried to boot it, it went farer then before, but then ended up with the same error.
4. Then I tried Windows 8.1 64 bit. For the purpose I bought 2 SSDs for two of my rigs. I installed first the express installation of the Catalyst drivers from the AMD site. Then, whatever settings of the cgminer I was using, I was getting a system crash and freeze in few seconds. Then I read advise to stop the Catalyst Control Center. So did I and I went a step forward. I was able to get around 30-60 minutes of mining but ending up again with crash. I have tried again tons of settings on the cgminer, both literally copied from folks with same GPUs and with much lower TC and intensity values. The result was that when I waked up in the morning the rig has been shut down from the power.
Then I read a suggestion to make a clear instllation of Windows, install ONLY the Display driver without anything else. So did I. The result was, Windows 8.1 recognizing only 5 cards whatever I did with my 6th card. However, I made another step forward, I succeded to mine very stable at around 69 degrees, 4.5 hours, 465 kh/s. However, when I woked up, the rig has been shut down from the power again. I tought it may be the PSU or sth, however, my PSU should be plenty for these GPUs, even more, that night only 5 GPUs were running, on the first PSU 2 GPUs + Mobo and anothr 3 GPU on the second PSU. But anyway. My 6th card was not recognised so I tried various staff again - installing Catalyst 13,1 from techpowerup, 13.11 beta from AMD, installing only the Display driver, installing older (2.7) version of SDK or not installing SDK at all. Whatever I did it didn't recognise the 6th card again.
I am going to run out of ideas and options for what to try. I just want to point out that I am believeing it is not hardware related as I have tried (well, not all but most of the described above options) on 3 rigs.
I am compltely serious about the bounty announcment and I am ready to send it right away now to a trusted escrow member. I am ready to release it once I get my rigs running at 480 or at least 470 kh/s per GPU for at least 72 hours with no failure and no HW errors as I learned that anytning can break up later even if it starts well in the begining.
I am most curious why the SMOS and BAMT are freezing instantly and I would most like to use them instead of other options but Windows will be OK as well if i can get my system really running.
Edit: Regarding recognizing the 6th card under Windows I forgot to mention that I have already tried:
* dummy plug
* A1 to B17 shortening
* Right clicking on not recognized device in device manager, uninstall, and rescan for hardware.
When doing my first install what worked out of the box was the pin shortening, however, it worked when I used the default auto detectable AMD drivers with express install, which later was causing my system to freeze after 5 minutes of mining. When I tried to install just the Display Driver without the other extras or any other version of the driver, whatever I tried, the 6th GPU was not recognised.
In SMOS and BAMT at least all the 6 GPUs were recognised and were starting to mine for a minute or so before everything freezes and crashes.