MSI Z170A GAMING M5 - 7 GPU [SOLVED]

[hcwh]

Hi,

I have 7 x Nvidia GTX 1070 with this motherboard and have it running with 6 fine (not overclocking yet).

I am using an M.2. SSD and have 5 cards and 6 USB risers on a 1000W PSU and 2 cards and the M.2. SSD on a 500W PSU.

Sometimes when I reboot one of the cards fails to load (in device manager):

Windows has stopped this device because it has reported problems. (Code 43)

If I disable / enable the device then it works again.

I have run Display Driver Uninstaller then re-installed the Nvidia drivers but this does not help.

I have swapped the USB cable riser around and used different 3d cards, the hardware is all fine.

Any advice?

To get 8 GPU working in Windows 7, you need to make some changes to the registry to address the code 43 warning.  Problem I had is when I went down to 7 GPUs and back up to 8, the registry hack no longer worked.  I didn't spend too much more time on it before switching to Linux which doesn't have that problem.  Check out the NVIDIA forum post below and try that and it may help.  Supposedly not necessary for Windows 10.

https://devtalk.nvidia.com/default/topic/738070/cuda-setup-and-installation/8x-gpu-gtx-issue-under-windows/post/4208565/#4208565

[1715088317]

1715088317

[1715088317]

1715088317

[1715088317]

1715088317

[1715088317]

1715088317

[hcwh]

Anything special you needed to do to get the M.2 to PCIe converter to work on the Gaming M5?  What BIOS version are you using and what are your BIOS settings for PCI and monitor?

my bios is the latest version, i enabled 4g decoding, the gen1,gen2 setting is auto,  disable integreted video card, plug hdmi monitor  to the first Pcie 16x, m.2 closest to CPU works, the other m.2 doesn't work. I haven't try  Windows, I use nvOC, a customized linux, can be found in this forum.
now 8 cards works very stable, 7 1050ti and 1 1060.
btw, CPU is g3900.

Thanks.  How much system RAM do you have?

I have more m.2 to pcie converters arriving this week along with another 4GB RAM so I'll be able to do more testing very soon.

What issues (if any) did you encounter plugging HDMI monitor into other cards?

Did you disable other things like sound, serial i/o, etc?

Do you have any automated startup routines to undervolt and/or overclock (using nvidia-smi and nvidia-settings)?

[car1999]

Anything special you needed to do to get the M.2 to PCIe converter to work on the Gaming M5?  What BIOS version are you using and what are your BIOS settings for PCI and monitor?

my bios is the latest version, i enabled 4g decoding, the gen1,gen2 setting is auto,  disable integreted video card, plug hdmi monitor  to the first Pcie 16x, m.2 closest to CPU works, the other m.2 doesn't work. I haven't try  Windows, I use nvOC, a customized linux, can be found in this forum.
now 8 cards works very stable, 7 1050ti and 1 1060.
btw, CPU is g3900.

Thanks.  How much system RAM do you have?

I have more m.2 to pcie converters arriving this week along with another 4GB RAM so I'll be able to do more testing very soon.

What issues (if any) did you encounter plugging HDMI monitor into other cards?

Did you disable other things like sound, serial i/o, etc?

Do you have any automated startup routines to undervolt and/or overclock (using nvidia-smi and nvidia-settings)?

my ram is 4GB, it's big enough. we don't need to plugin hdmi to other cards, the first card is fine. I don't disable other things, mysystem is nvOC and I believe it runs nvidia-settings on bootup to setup overclock and fan spead.

[themonkii]

Running into an issue with Windows 10 not properly detecting my cards, or not installing the drivers. Seems fine with 4 or 5 cards, 6 or 7 and it goes to junk. Also giving a Code 43 in some cases.

It was working the other day, all cards detecting, and showing up in Device Manager, but had to fool around some to flash a card. (card is working) so I know it wasn't the flash that is causing my issues.

I have to say, this board is finicky with 5+ GPU's....

Any ideas?

[ltcsprite]

Running into an issue with Windows 10 not properly detecting my cards, or not installing the drivers. Seems fine with 4 or 5 cards, 6 or 7 and it goes to junk. Also giving a Code 43 in some cases.

It was working the other day, all cards detecting, and showing up in Device Manager, but had to fool around some to flash a card. (card is working) so I know it wasn't the flash that is causing my issues.

I have to say, this board is finicky with 5+ GPU's....

Any ideas?

Only way I was able to get 7 running smooth on these boards is running them on SMOS instead of win10.

[themonkii]

Running into an issue with Windows 10 not properly detecting my cards, or not installing the drivers. Seems fine with 4 or 5 cards, 6 or 7 and it goes to junk. Also giving a Code 43 in some cases.

It was working the other day, all cards detecting, and showing up in Device Manager, but had to fool around some to flash a card. (card is working) so I know it wasn't the flash that is causing my issues.

I have to say, this board is finicky with 5+ GPU's....

Any ideas?

Only way I was able to get 7 running smooth on these boards is running them on SMOS instead of win10.

I can say with certainty that I DID have all 7 firing, and mining under Win10 w Claymore 9.6 - Then I screwed it up when I tried to flash one of the cards. Seems, based on what I read, having a modded card in slot 2 causes issues with the board not getting past POST - (board debug code may be 92-96, 97,98,99) - This has been a problem for me. I'll test tonight, and report back.

[ltcsprite]

Running into an issue with Windows 10 not properly detecting my cards, or not installing the drivers. Seems fine with 4 or 5 cards, 6 or 7 and it goes to junk. Also giving a Code 43 in some cases.

It was working the other day, all cards detecting, and showing up in Device Manager, but had to fool around some to flash a card. (card is working) so I know it wasn't the flash that is causing my issues.

I have to say, this board is finicky with 5+ GPU's....

Any ideas?

Only way I was able to get 7 running smooth on these boards is running them on SMOS instead of win10.

I can say with certainty that I DID have all 7 firing, and mining under Win10 w Claymore 9.6 - Then I screwed it up when I tried to flash one of the cards. Seems, based on what I read, having a modded card in slot 2 causes issues with the board not getting past POST - (board debug code may be 92-96, 97,98,99) - This has been a problem for me. I'll test tonight, and report back.

Yeah I had 7 running in win10 but at some point the board always went nuts and they would stop working, code 12, code 43, all kinda problems I gave up and switched to SMOS.

[KevinUK]

Hi,

I have 7 x Nvidia GTX 1070 with this motherboard and have it running with 6 fine (not overclocking yet).

I am using an M.2. SSD and have 5 cards and 6 USB risers on a 1000W PSU and 2 cards and the M.2. SSD on a 500W PSU.

Sometimes when I reboot one of the cards fails to load (in device manager):

Windows has stopped this device because it has reported problems. (Code 43)

If I disable / enable the device then it works again.

I have run Display Driver Uninstaller then re-installed the Nvidia drivers but this does not help.

I have swapped the USB cable riser around and used different 3d cards, the hardware is all fine.

Any advice?

To get 8 GPU working in Windows 7, you need to make some changes to the registry to address the code 43 warning.  Problem I had is when I went down to 7 GPUs and back up to 8, the registry hack no longer worked.  I didn't spend too much more time on it before switching to Linux which doesn't have that problem.  Check out the NVIDIA forum post below and try that and it may help.  Supposedly not necessary for Windows 10.

https://devtalk.nvidia.com/default/topic/738070/cuda-setup-and-installation/8x-gpu-gtx-issue-under-windows/post/4208565/#4208565

I'm on Windows 10 but tried anyway, it didn't work though. I can occasionally get 7 cards to show without errors but even then I mostly get the error:

CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.

I've had it working perfectly with 7 cards a couple of times. Maybe it's time to try SMOS 

[themonkii]

Hi,

I have 7 x Nvidia GTX 1070 with this motherboard and have it running with 6 fine (not overclocking yet).

I am using an M.2. SSD and have 5 cards and 6 USB risers on a 1000W PSU and 2 cards and the M.2. SSD on a 500W PSU.

Sometimes when I reboot one of the cards fails to load (in device manager):

Windows has stopped this device because it has reported problems. (Code 43)

If I disable / enable the device then it works again.

I have run Display Driver Uninstaller then re-installed the Nvidia drivers but this does not help.

I have swapped the USB cable riser around and used different 3d cards, the hardware is all fine.

Any advice?

To get 8 GPU working in Windows 7, you need to make some changes to the registry to address the code 43 warning.  Problem I had is when I went down to 7 GPUs and back up to 8, the registry hack no longer worked.  I didn't spend too much more time on it before switching to Linux which doesn't have that problem.  Check out the NVIDIA forum post below and try that and it may help.  Supposedly not necessary for Windows 10.

https://devtalk.nvidia.com/default/topic/738070/cuda-setup-and-installation/8x-gpu-gtx-issue-under-windows/post/4208565/#4208565

I'm on Windows 10 but tried anyway, it didn't work though. I can occasionally get 7 cards to show without errors but even then I mostly get the error:

CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.

I've had it working perfectly with 7 cards a couple of times. Maybe it's time to try SMOS 

The "Cannot allocate big buffer for DAG error...." could also be because you haven't increased the Virtual Memory to 16GB

[KevinUK]

I've got the virtual memory set to 16GB, I've tried everything I've read about it. 

Plus the miner has worked for me with 7 cards, just very rarely and until I had to reboot.

[themonkii]

I've got the virtual memory set to 16GB, I've tried everything I've read about it. 

Plus the miner has worked for me with 7 cards, just very rarely and until I had to reboot.

hrmmm, and I assume that all cards are over 2GB?

I've been at it about a week now with this board. I'm gonna make another effort tonight, so stay tuned. I'll report back here with my findings. All that said, its been a real ball ache, I could only imagine the BS involved with trying to use the M.2 connections. 

[KevinUK]

I've got the virtual memory set to 16GB, I've tried everything I've read about it. 

Plus the miner has worked for me with 7 cards, just very rarely and until I had to reboot.

hrmmm, and I assume that all cards are over 2GB?

I've been at it about a week now with this board. I'm gonna make another effort tonight, so stay tuned. I'll report back here with my findings. All that said, its been a real ball ache, I could only imagine the BS involved with trying to use the M.2 connections. 

Yep, 1070 4GB cards.

[themonkii]

I've got the virtual memory set to 16GB, I've tried everything I've read about it. 

Plus the miner has worked for me with 7 cards, just very rarely and until I had to reboot.

hrmmm, and I assume that all cards are over 2GB?

I've been at it about a week now with this board. I'm gonna make another effort tonight, so stay tuned. I'll report back here with my findings. All that said, its been a real ball ache, I could only imagine the BS involved with trying to use the M.2 connections. 

Yep, 1070 4GB cards.

So, just as an update - I did get all cards working properly last night. 7 GPUS, all firing now. What a nasty adventure. Here is what I can confirm worked.

Set PEG 0 and PEG 1 to GEN 1, 4G Decode "Enabled", Windows OS Configuration: Windows 8.1/10 "Disabled", Windows 7 "Disabled", Fast Boot "Disabled".  Reboot/shutdown.  Install one card at a time, rebooting between each card (not sure why, but worked for me) starting with x16 slot #2, then #5, 7, 1, 6, 4, 3.  If cards are modded you may have issues where screen won't post (board debug code may be 92-96, 97,98,99): in that case try installing a non-modded card in slot 2, and the modded ones in the remaining slots.  Alternatively, try OS Config...Windows 8.1/10 "Enabled" + with or without Fast Boot.

I used hacked drivers, but what is above is what did it for me. Hope this helps!

[isoneguy]

good lord, stop buying the z1xx chipset and buy the z2xx chipset, biostar's z2xx models all work out of box for me.

no rain dancing required...

[Renanxx1]

Guys I've tried many options and nothing work here in my z170a m5.

When I active the 4G decode the computer gets black screen.

[KevinUK]

I had mine mining for 8 hours on 6 cards, the PC then died and needed a hard reboot. It rebooted and came up with 7 cards working for roughly the same length of time.

I hard rebooted again and with 7 cards I get the CUDA error (never get the CUDA error with 6 cards).

I've also had it mining a few minutes and then get:

GpuMiner cu_k01 failed 6, the launch timed out and was terminated

It's all very random. I've tried -r 180 to try to prevent the 8 hour or so hard reboot issue but it didn't help.

I think I'm on GEN 2, I'll try GEN 1.

[themonkii]

Guys I've tried many options and nothing work here in my z170a m5.

When I active the 4G decode the computer gets black screen.

You must have 4G decode set, and have UIFI Mode set BEFORE you install Windows. Doing it after the fact will cause it to not boot.

[BennyT]

Hey guys.. got this board working like a charm with 7 cards. All 1070's by the way with kaby lake. Using 6 of the PCIE slots and one M2 slot. I don't have an 8th card to try maxing out the PCIE slots. Took a lot of trial and error so I know the pain that it can be. If anyone needs pointers, glad to help.

[armazingerz]

Hi,

we are trying to set up a 7x 1070 rig, but we are stucked when detecting the gpus, we can't get over 4x 1070, we get a #12 code error, like it has not enough resources. This is our set up:
- MSI Z170A GAMING M5
- OS Windows 10
- CPU Intel G4400
- 4GB RAM
- 7x 1070
- Latest drivers

We also have the same problem with 7x 1080 rig.

[MONKEYSTOMP]

Update have 2 rigs currently running 6 1070 running stable. Can't get a 7th card stable, I think it's due to my psu issues with rails. 6 is stable. I was crashing because I think I had a bad gpu. That one card  would run at about 72 degrees when the other 5 would be at 55. Swapped that out and 6 ran fine. Setting a 3rd rig up on a z2 today hope it goes smoothly.