Topic: ASRock H81 Pro will not accept more than 2 GPU's (Code 12) - page 3. (Read 4189 times)

Bought another gpu, so rearranged the existing gpus to fit, and changed the slots. Now only 3 cards are detectable. 4th one if detected throws up a code 43 error 

hackflags, uninstalling using ddu, swapping cards, resetting bios, adjust bios settings to no avail. Going to try reinstalling windows now

I hope the asus z270-a board for my 2nd rig isn't going to be such a pita

hi, Guys!

It is Great News!
Windows Update on Jan/3/2018[KB4056892] might fix this probrem!

You no longer need HackFlags nor Disable Audio Devices!

hi, geck.

Sadly, iGPU will be the obstacle.

If iGPU is enabled on my rig,
Code 12 will be indicated even if the registry is turned on and all audio devices are disabled.

Followed your steps to the T with th exception of installing 1703 and all 4 gpus were detected.

I also did not install the intel graphics driver and have left the pci-e 16x slot vacant. I have a feeling i am going to have troubles when my fifth and sixth card arrives. Will probably need to disable the igpu but this would be troublesome as the cards are inside a server case. But we will cross that bridge when i get ther 😬

Thanks so much for taking the time to write it all down!

I would like to share my experience with that problem! I had the same problem, the 3rd gpu was giving me the same error. I did pretty much everything i found online, increase the virtual memory, updated the bios, change the registry, updated the windows to 1079, switch all the hardware, and im not exactly sure what was the problem but the last thing i did is to disable the audio attached to each card and now everything is working properly. 3 more cards to go will see how will go. Thanks everybody

You sir are my hero!

Well I did a little bit digging for this issue and I can confirm 100% that it's cause by Windows Update 1709, and this helped my issue.
I had 6x 1060 GB cards, and I had two not working.

After I applied that one started to work, and just went to Device Manager disabled HD Audio for the second graphic and it magically started to work!

hi WickedPigeon.

Perhaps, FallCreatorsUpdate(1709) triggered a malfunction of GPU driver installation.
If you are going to install 1709 cleanly, please try disabling HD audio before installing the GPU driver.

Thanks for all the great help on this thread! I'm back up. I was doing fine for a couple weeks until I did a Windows update. Then everything went down and I couldn't get but 1 or 2 GPUs to show up. Drove me crazy. Tried all the suggestions on this board. Nothing worked. Until...

Here is what I have:
ASRock H81 Pro R2.0 BTC
Intel Celeron G1840 Haswell Processor. 2.8 GHz. LGA1150. 2MB Cache
Gskill (2X4) 8GB DDR3 1600MHz
Kingston A400 SSD 120 GB
5 x MSI Nvidia GeForce GTX 1070 Ti G1 Gaming 8GB GDDR5
EVGA 1000W G3,  
6x 1x To 16x Extender Riser USB 3.0

No changes to the bios except "on board" video.

Ended up formatting SSD. Unplugged the Internet cable. Installed Windows 10 Home, V1607 (because that's what I have in the original, flash media).  Disabled Windows Updates - Via Control Panel > Administrative Tools, you can access Services. In the Services window, scroll down to Windows Update and turn off the process. Then I uninstalled the Windows Update Assistant. With one riser in the x16 Pci slot (I believe that is slot 1), updated Nvidia graphic driver. Rebooted. Shut down, added next card and rebooted. Once I had all the cards in (I'm only using 5 cards), I re-installed Nvidia driver with "Clean instillation" checked. And it's finally working.

I believe this is a new Windows update problem.

hi Guys,

Today,
I tried to clean installation again on this MOBO. (BIOS 1.20, Using default settings)
The successful procedure is below.

1. Using "Media Creation Tool", prepare the latest USB installation media. (Version 1709)

2. Install the OS without connecting to the network.

3. Please still be offline after installation. Make sure that all of your GPUs are visible in Device Manager. (The NVIDIA driver has not been installed yet.)

4. Disable all HD audio devices. (Optionally, set HackFlags and reboot.)

5. Connect to the network. Install the driver automatically in device manager, or install Nvidia genuine. (Keep all HD audio devices disabled even after driver installation.)

6. If HackFlags does not set and Code12 still indicated, try HackFlags and reboot.

But I do not know if this is effective with AMD's GPU.
Can anyone verify it.

ty

All the other options didn't work for me.  2 cards were fine the 3rd had a code 12.  The above work around worked perfectly.  I just deleted some HD audio under the pcie list and it worked.

hi, guys.

Sorry. That's right. I made a mistake in expression.
Since this is a "32bit value", the "numbers of zeros" after "0x" has significance.

Correct screenshot is here.
https://bitcointalksearch.org/topic/m.24696485

If you are doing exactly the same edits and your rig does not run yet.
I want to know the cause. Please tell me about your rig composition.

ty

Hi do you (Or someone else) have a screenshot of the HackFlags setting in Regedit?

I couldn't set it to 0x600 but it looked more like 0x0000600 (Can't remember, my pc didn't start up after that). 
And I am quite unsure if I did it correct.  (seems not, otherwise my rig would start up),   

Thanks,

Hi, Scippylisp.

About the screw of PCI bus number.
There may be no problem if the numbers are not overlapped each other.
These are automatically numbered, it may be a consecutive number or not be so.

Physical Slot2(PCIEx16) of this MOBO is different from other PCIEx1.
It seems that it is directly connected to the CPU.

You can find it under "Intel (R) Xeon (R) processor E3-1200 v3 / 4th Gen Core processor PCI Express x16 Controller".
Not "Intel (R) 8 Series / C220 Series PCI Express Root Port".

However,
It seems that "the behavior of PCIE assignment to memory" was surely changed in version 1709.
Combining version 1709 with "this MOBO with less memory" seems to make the problem serious.

By the way,
6GPUs on my rig running with windows10 1709 now.

OS: Windows10pro 64bit 1709 (Clean install from USB media. Then edit HackFlags turn on)
MB: ASrock H81 BTC Pro R2.0 (BIOS v1.20, Setting is "load default".)
CPU: Intel Celeron G1840
RAM: DDR3 8GBx2 (16GB)
GPUs
 PCIE1: GTX1060 6GB
 PCIE2: GTX1080
 PCIE3: GTX1070
 PCIE4: GTX1070
 PCIE5: GTX1070
 PCIE6: GTX1070

(All GPUs connected via riser card. and Disabling HD audio devices in device manager.)

When turn off HackFlags and enable audio devices,  Three GPUs indicate code12.

Hey all,

Thank you for making this thread. It has helped me troubleshoot the same problem. I have 6 GTX 1070s, but only 4 work. Yes, it definitely seems to be the latest Windows update. I updated on Wednesday and have had the problem since that time. I did NOT have a problem before the update.


I should state going in that I'm no tech wizard, but built the rig myself. So apologies if I get the language wrong.

Many of the workarounds here have not worked for me (even though I feel I've cleaned up my system a bit).

Here is one thing I am noticing though, maybe others have the same setup.


BEFORE the update, I am pretty certain that the PCI slots on the mobo were the same as the device properties. In other words, the card plugged into slot 1 would read "PCI Bus 1" in the properties.

NOW, this appears to be screwed up. It appears as if 1 and 2 are swapped, 3 and 4 are swapped, & 5 and 6 are swapped. So the Windows updated created a conflict where it now thinks PCI slot 6 (for example) is actually PCI slot 5 and so on.


In trying to troubleshoot this, I noticed when I went into the device manager and sorted by "resources by connection", I noticed that the larger slot (PCIE 2) was not being resourced by one of the PCI root drivers. As such, I believe this is why both cards plugged into slot 1 and 2 work.

However, I believe that slots 3/4 and slots 5/6 are only able to take one card each because they have been swapped by the Windows update. Both slots are trying to draw off the same PCI root driver/resource. This explains to me why Windows recognizes that the card is plugged into the slot but can't use it.

So based on that, perhaps any workaround would have to involve manually assigning a PCI port to the driver that is not currently using a resource.

Either that or wait for the new Windows update or roll back to 1703.  


What do you think?

I had the same problems yesterday-today with a rig where I wanted to add 2 card to the existing 2 cards.
long story short:
killed the win 10 v1709 and fresh install v1703, deactivated updates, antivirus. now it's like it should be.
imho don't waste time finding workarounds, just use 1703 and deactivate updates.

Hi guys, I'm looking for help with this MOBO rev1.0  using GTX970's. I have 3 connected via risers all power is connected to risers and mobo.

The HKEY I need to add, please could you walk me through this to make sure I have it correct. I pretty sure it is though it hasn't worked.

I have followed the advice so far with little success.

Is there a list of Bios settings?

Thanks in advance of any help.

hi, djtappage.


Yes.
If TOLUD setting to "Auto" (like ASRock H81 BTC Pro R2.0)
"The amount of resources that can be allocated for GPUs" will Adjusted by physical RAM amount automatically.

So, If the rig have less physical RAM.
We must makeshift with fewer resources.
(like disabling some HDMI audio on video devices.)

A simple way to reduce the problem is to add physical RAM.
But, I have seen the case of driving 3GPUs with 4GB RAM. It seems possible.

The Haggyy's rig with 3GPUs and 4GB physical RAM finally started running.
https://bitcointalk.org/index.php?topic=2307023.0;all

Aside from this.
"HackFlags" increases limit of "the number that Windows can enumerate PCIE devices".
"devices not enumerated" also indicate CODE 12.
Perhaps, I think that initial limit value has changed to less value in FallCreatorsUpdate. Or Update have bugs.

be careful.
The problems of TOLUD and LimitRegister are independent each other.

I didn't go through all posts BUT I had same issue. I started with wind10, tried all options listed on internet - nothing helped. The best result was that I could run 2 GPU (gtx 1080) on my windows but still with error code 12 on third one. When I added 4th, then the windows was freezing at the login page. I think it is Windows issue and one of the recent patches.

Solution:
Install Ubuntu 16.04 - it took me one hour to configure the system for mining. I know there is no MSI afterburner software but it is quite easy to use nvidia-smi commands instead.

Hi Guys, having same issues over here with this

only purchased all hardware last week and still havn't got to the bottom of this issue. Tried multiple fixes so any suggestions i would gratefully appreciate it in advance

My Hardware setup is as follows:
asrok h81 Pro BTC r2.0
Intel g3220
4gb ddr3
windows 10 1709 build
Latest bios Flashed via internet
4 pins are connected both sides on board
3x Nvidia gtx 1050ti's - firmware 388.43 (drivers)
3x gpu risers (https://www.amazon.co.uk/Ptsaying-Powered-Riser-Adapter-Extension/dp/B073F1QHZ5/ref=sr_1_1?ie=UTF8&qid=1512829644&sr=8-1&keywords=gpu+risers)

My issue is code 12 in device manager. I have managed to get 2 cards working at the same time but never 3. They are registering there connected however the third card will always show code 12.
https://ibb.co/kDp2Zb

heres what ive tried so far,
tried older drivers for graphics card
ive tried removing cards individually to determine if the riser's are faulty - which there not and nor are the cards
tries the suggested methods of modifying the bios i.e setting to gen 1 gen 2, None of which seem to make any difference also disabled hd audio
Tried increasing page file.
Tried adjusting registry with HackFlags method.
tried installing ubuntu however i couldnt get this to recognize 1 card let alone 2 or three. (im not a fan of ubuntu) so i gave up with this but i want it to work on windows anyway.
tried following the https://forums.geforce.com/default/topic/470858/code-12/   - however my 3rd card doesn't show on the lists
tried DDU to remove drivers and starting again (45 restarts and counting)


Ive done a few more things but im starting to run out of ideas

one question i do have - does physical RAM effect the amount of resources that can be allocated for a gpu? - bascially is 4gb ddr3 enough to run 3 gpu's? or should this not matter

Im sure its the later but i feel like im missing something really obvious here

George

Hi, guys.

Yes. Specific operations in the device manager are as follows.