2 months of struggle. Can't get 6 PGU's to work. Tried literally everything, I've found on forums

[Simonas.st]

Good day!

I’m using MSI z390-a pro.
4x 1660 TI
2x 1660 Super

BIOS settings done:
2 gen pcie
Windows support UEFI
4G memory/Crypto Currency mining to enabled

BIOS drivers updated
Nvidia drivers updated to 470.74
Hiveos 0.6-210@210928

Interchanged risers, GPU’s, even bought 16 to 16 risers for 2 GPU’s. Nothing works.
5 GPU’s working fine, when I put 6th in, it doesn’t boot. After many tries it boots up and don’t mine. After: cd /hive-config
rm nvidia-oc.conf autofan.conf it starts to mine, but immediately throws errors: GPU temps missing, OC failed, load average too high, and etc.
Then I disconnect one GPU, doesn’t matter which one, and after some reboots it starts to work, and works fine.
Tried to reach out to hiveos help, without much success.
I don’t know what to do next. Hope someone in community has a minute to help me out. Would be much appreciated.

[Grea]

Have you:

• tried a different OS with the 6 GPUs? Easy enough to trial a couple to confirm it is or is not the current “HiveOS” load and confirm it is Hardware/Software Config related.
• no mention of how much power is available and it’s rig distribution
• Assuming you have disabled all the non-essentials for mining to free up post boot resources. This was the frustrating issue getting a 6th GPU running stable on my Asus z-87 rig.
• Bumped the motherboard RAM up to 8GB and made sure whatever you are running passes a real memory test?

Hope you can track it down. Good luck.

[Simonas.st]

Hey!
I haven’t tried different OS, because all my computers Mac’s. On rig I have only Hiveos installed. No windows.

I have 2 PSU’s.
Motherboard and 2x 1660supers running from 650w corsair
4x 1660 TI running from 750w corsair

Yes, I’ve disabled everything except LAN, as I run rig from cable, not wifi.

Hive says I have 2.3GB free RAM. I have 4GB crucial installed.

Does it say anything to You?

[Grea]

Guessing running stable with (5) GPUs has been proven in both situations:
(4) on the 750w Corsair
(1) and the mainboard on the 650w Corsair

and

(3) on the 750w Corsair
(2) and mainboard on the 650w Corsair

Running OS from a USB?
Logging turned on or off?

FWIW: I only have (1) nVidia GPU running, so I am not a great source to spot brand related particulars

[Simonas.st]

Running stable was only tried
(4) on the 750w Corsair
(1) and the mainboard on the 650w Corsair

Right now, I’m setting up
(3) on the 750w Corsair
(2) and mainboard on the 650w Corsair

Running Hiveos from SSD.
I’m not sure, where could I find logging setup?

Thank You for helping me out.

[Simonas.st]

Also, after trying to run 6th GPU, I’ve decided to go back to 5 and leave it for now. But it didn’t work either.
I’ve only managed to work it with downgrading nvidia drivers to 460.91.03
But now, I’m experiencing high load average and dropped hashrate.

WTF?

[Grea]

Always good to be sure the baseline gear is solid and has some extra capacity. Even with (5) GPUs, were you able to confirm each of your power supplies would run the number of GPUs you wanted, the 650w (2 GPUs) + MB & 750w (4 GPUs)?

In the past experience my high LA events were miner based conflicts/config issues. I am sure there are lots of other causes.

T-Rex definitely has a few commands that needed updates across the Linux OS updates. If you are running T-Rex, you may have crossed some of those thresholds.

Personally, I don’t mess with drivers outside of stable kernel loads or selecting beta loads for my specific GPU&miners. Again, I am worthless in your pursuit

You might find some detailed help on your target miner discord channel, or even Hive Overclock discord.

[Simonas.st]

I’m not discord friendly. I’ll try. Maybe even worth to write in msi forums.

BTW, I’ve tried to switch miner from Trex to Phoenixminer, and it helped with hashrate and load averages.
Maybe it would work with 6 gpu’s? I’m afraid to try it though.

[Grea]

Downtime of a rig is far more detrimental than chasing 5% increases in hashrate, I don’t blame you. Better to have cheap test rig or a Windows PC for that other GPU until you have some spare test gear to quickly isolate possible HW issues

@Simonas.st

5 GPU’s working fine, when I put 6th in, it doesn’t boot. After many tries it boots up and don’t mine.

If the BIOS lets an OS see all 6, you should be able to get it working. I don’t think the 5 vs. 6 has anything to do with your miner software. Seems low level to me.

After confirming the power supplies you have can power the 4 & 2+MB legs you built, I’d put a fresh load of HiveOS, MMPOS, or even a solid Windows load on that system at some point to see if you can detect 6 GPUs. Eliminate the potential for power issues, risers, etc., first.

fwiw: I have 7 RX series GPUs running on dual 860i Corsairs, 4 & 3+MB configurations. Measured them at the wall to get the right balance 610w and 454w respectively.

[Simonas.st]

Measuring at the wall, sounds like a good idea. I need to buy that measurement tool. Other than that, I don’t know how to make sure it’s enough power.

[Grea]

Split the power legs of the rig.

Shut off the power to 750, run the few of the GPUs+MB the 650 alone. If it is stable with MB+2 or even MB+3…bring up the rest on the 750.

I don’t know jack about your GPUs or your MB+CPU power draw. You might need your power supplies on different house circuits depending on what else is drawing off those circuits

[Simonas.st]

Alright I will make some experiment with PSU switching. Will post results here later. Thank You. Much appreciate it.

[Simonas.st]

Just a though. Should it work with 6 PGU’s if I would under power it to like 50W each? Theoretically?

[Grea]

If the wattages on WhatToMine are correct for your GPUs…I’d likely try this first if shooting for (6) is the goal. It should keep both below the 50% max range and physically logical:

650 Power supply
100W = MB with no CPU mining
180W = 1660 Super 90w Each

Then add these:
750 Power supply
320W = 1660Ti 80w Each

Limitations of course is powering risers with your available power supply cabling. Not to proud to say, a few of my ATX powered rigs are molex powered risers

[bigtoes18]

What’s your cpu? Some cpu don’t have enough pcie lanes to handle more than 6 pcie connected devices. Try to disable the integrated graphics in your mb bios to release one pcie lane.

[Simonas.st]

Hey!
I don’t have integrated graphics.
My CPU is: Intel Core i3-9100F

[bigtoes18]

Try to set all the pcie setting in the motherboard as Gen1, you may need set it in both of the two different locations. It seems your cpu can only handle 8 Gen 2 pcie lanes. You may somehow used 3 of them already. And also, try to disable UEFI support. Lastly, you may already did this, but update the motherboard bios to the latest version may help. Some other minor tips: plug in gpu one by one. Try new risers. I am also suspecting that using a 750 w power supply for 4 1660 Ti may not enough. You can also try to use the pcie spillter card to change one of the pcie ports to 4 of them. Maybe it’s just a bad pcie ports on your motherboard.

[tlsolutions]

The integrated graph is inside your Intel CPU, not on the motherboard.

You have to use it for troubleshooting, otherwise you can’t check what is happening to the system itself between turning on and booting Hive with the sixth card plugged in.

In the BIOS set the CPU to use native clocks only, no automatic under or overclocking.

Try starting up the rig without a flight sheet.

Try fix your HIVE overclocks.

I had high LA problem on my 3060 v2 LHR cards, tried a lot of different drivers, miner versions, HIVE versions. Turned out it was the effect of a badly choosen memory overclock, wich was too high. After I had set it lower, they became stable and running non-stop since then.

[sirvok]

i also have msi z390-a pro, first time having a problem when i try to connect 6 GPU :
3 rx 5600 xt, 3 rx 580/570/470 4GB and it was a nightmare

manage to run all 6 GPU run using various setting / oc as state in many forum

• 2 gen pcie
• 4G memory/Crypto Currency mining to enabled
• windows UEFI settings
• using ssd / usb with rufus (UEFI) not using belena etcher (someone said not using image with .xz extension but with zip extension in the HiveOs History Image)
• disable all port sound / serial / parallel port
• powerup the PCIE power in motherboard (someone said this mobo lack of power)
• test the GPU one by one
• disable on board GPU, using the GPU 0 as display
• also clear the oc setting in hiveos seem helping

eventually this mobo successfully mine with 6 GPU , my advise is don’t change anything after you have successfully mine with 6 GPU

because i’m having a problem after takeout 1-2 GPU and trying to add a new GPU
(sold 2 of my gpu) and as for now this mobo having a problem AGAIN limited to 4 GPU , the problem is so frustrating

random pcie connect / disconnect , the only thing that i have not try again is build another SDD / USB using rufus with the zip image in hive history image website (too lazy to do that).

as for now i’m gonna change this mobo with another mobo.

[Simonas.st]

Alright, so it’s not working with 5 gpu’s anymore. It has worked fines like 24 hours, when I’ve changed miner. BUT, today I was away for a whole day, and I was getting miner reboot, OC failed, etc messages all day. Came home, unplugged everything, left one card. So smooth… How can it be so difficult?