Thanks to this thread I have successfully used 4 GPU (Gigabyte 1070) for mining on ASRock H81 Pro rev2. It took maybe 8h mostly installing OS and update (so slow). If you are crazy enough like me and want to try Linux and thinking it's going to be easier let me stop you right now. I failed miserably trying Linux but your mileage may vary.
code 12 is your best friend because it means your hardware is most likely installed correctly (all plugged correctly) and working. Code 43/45 are your enemies: double check cables and connections.
Summary solution:
1) Install
Windows 10 1709. Install all latest windows update. Disable all audio device on the video card by sorting the view by connection in device manager. At this point you should already see all your video cards.
2) If you have any GPU that has error
code 12 do HackFlags (400) fix. Reboot.
3) If GPU has error
code 43 make sure all connectors are attached tightly. Ex: power issue (connector not properly inserted for example), I suspected heat but looking at afterburner app it's not the case.
4) If GPU has error
code 45 make sure riser cable/adapter are working fine. Ex: riser issue, (code 43 or 35 are mostly hardware related)
5) If you still see error code 12 sort view by connection and disable some un-used device like a USB device (make sure it's not being used; I used a USB hub and disabled another un-used one).
5b) It doesn't hurt to run search "troubleshoot" and select "Hardware and Devices", I would not enable auto driver update feature when asked, and select device not in this list option and proceed, it might fix something. In my case it was able to find my last GPU when it was hidden (but second time this happened it didn't help).
6) Reboot and see if all GPU are detected correctly. If so start mining and see if you get
VIDEO_TDR_FAILURE (BSOD).
6b) If TDR BSOD, try TdrDelay registry workaround first (reboot). MSDN:
https://docs.microsoft.com/en-us/windows-hardware/drivers/display/tdr-registry-keys. Reddit mentioned about this:
https://www.reddit.com/r/nvidia/comments/79nncg/driver_38813_faqdiscussion_thread/. You may download a tool (in that link) to change the TDR registry. I would just skip to next step first.
6c) If #6b failed try
mining with one algorithm only (forget dual mining). This is what made my rig stable. I was BSOD very quickly (few minutes into mining). This method worked for 4 GPU. When I added a 5th one TDR issue re-surfaced again. I am still debugging this issue. [update] I used that that tool (in previous link) to disable TDR. Let's hope it can last till tomorrow. Ideally it's better to increase delay because at least in worst case machine will reboot and you can auto start your miner [update, overnight no issue but this morning it froze without any crash; had to reset]. If you disable TDR and it fails it will freeze your machine (this is good for me right now).
If in doubt use Display Driver Uninstaller (
DDU,
http://www.guru3d.com/files-details/display-driver-uninstaller-download.html) in safe mode (run msconfig and select safe mode, it will ask to reboot, once in safe mode run DDU). You only need to do this if you have already installed latest driver. I had everything detected by OS and working with driver provided by Windows 10 (so until step #6 I didn't install any driver). It's possible that you don't even need to install latest display driver depending on the situation. Once I got BSOD I ran DDU and installed latest driver but the problem was elsewhere for me.
Next rig I will use better riser:
https://www.amazon.com/Ubit-Express-Graphics-Extension-Ethereum/dp/B076KLJR2Y/ref=sr_1_1?s=home-garden&ie=UTF8&qid=1515290286&sr=8-1&keywords=pcie+riser+ubit with LED hint to help identify hardware issue.
[Update]KERNEL_SECURITY_CHECK_FAILURE - FAIL_FAST_LIST_ENTRY_CORRUPT BSOD (nvlddmkm involved) after being stable for less than 3h.
I set up a batch script that launches the miner and put it in the startup folder and disabled password login. So after startup it will start miner automatically. Hopefully it can survive the night (update: crashed once in 12h).
[Update2]Added a 5th GPU and TDR issue came in full force. See #6 for info. Essentially I disabled TDR with a tool (mentioned in link reddit link above), but you may try TdrDelay and other settings and have auto start on reboot to avoid down time. Only down side to this approach is that if you BSOD very often it's best to fix the issue in a more permanent way to avoid hardware damage. I am in the process of trying out different drivers (390 is out) to find the most stable one.
[update3: trying drivers]Device Installation Settings: disable auto device update before doing DDU and installing older driver. Disable device driver update (somehow windows still updated nvidia driver with this setting disabled only) and disable windows update in group policy.
Completely disable windows update -
https://www.windowscentral.com/how-stop-updates-installing-automatically-windows-10Disable device update itself -
https://www.itechtics.com/disable-automatic-driver-updates-windows/Disable device update search -
https://technet.microsoft.com/en-us/library/cc730606%28v=ws.10%29.aspx?f=255&MSPPError=-2147217396I am using
nvidia driver 385.69 now no BSOD (it's possible other version can also work). Make sure to disable update or it will get auto updated to newer driver (it will eventually happen if you don't disable it).
Another issue I got which is puzzling is my 5th GPU stopped working (afterburner shows 0 for all sensors) after sometime even when not mining. What I ended up doing is set
power limit to 80%(my other settings are temp limit 80%, core 100, memory 350). All temperatures were
below 70c (67-68c). If it can run for 1 month without issues I can call it stable but for now I am happy.
Don't give up!
