Solved: PhönixMiner Four GPU and One Black Sheep

[IceTea91]

Morning,

I am mining with four GPU at the time, 2 x 3070 and 2 x 3080 Founders Edition.

One of the 3080 works perfect, it goes with following settings:

Power limit: 73
Temp Limit: 76
Core Clock: -325
Memory Clock: 1400
Fan Speed: Auto

Hashrate at 101 with 233 Watt

The second one I wanted to run with the same settings and i get a failure which causes into a Bluescreen with the Note: Memory_Management

Turns out if I set it to

Power limit: 73
Temp Limit: 76
Core Clock: -325
Memory Clock: 1000
Fan Speed: 65 (I cant make it Auto since it will use 100 % FanSpeed all the time, no matter how cool the card is)

Hashrate at 96 with 209 Watt

The card will run with phönixminer BUT after lets say 30 to 60 Minutes the Cards reduces the Memory Clock itself slowly step by step.

Hashrate will go down in Steps from 96 to 84 and in the same time the Watts will go up to around 300.

When I look into Afterburner and i "refresh" the card by selecting another and go back to the "black sheep" it will show me that it resets the settings itself to the default

Power limit: 100
Temp Limit: 76
Core Clock: 0
Memory Clock: 0
Fan Speed: Auto

I am unsure what to do, I am struggeling since 4 days trying to fix that problem with google but i cant find anything to solve it.

Would be happy to find someone who can help me out. Ty already.

EDIT: Virtual Memory is set to 44000

[1714623221]

1714623221

[1714623221]

1714623221

[1714623221]

1714623221

[Wingsbtc]

1. Is the MSI afterburner the latest version? Check that out first if it's not them update
2. Remove other GPUs and leave that one stubborn GPU in your motherboard to play with, the fact is settings are always different for GPUs even if they have the same name, memory can be different, hynix, Samsung, micron all have different settings so which one is yours?

[ummasek]

I would stop using AB. Use proper command line to use oc on PHXM.

Reseting OC is typical for driver error. Mostly had it after too much OC.
Try to lower mem OC to 800 and see what happens.
Could You tell me what brand and certyficate PSU you`re using??

[sxemini]

Memory overheat, so the card throttle down maybe? Check the memory temp. I can´t find the thread here at the moment.
Or to much mem oc and the card reset the driver after a while. Not all cards can do the same mem clock speed.

[ummasek]

Use gpu-z to watch live all sensors.

in my opinion, oc drives to driver restart, witch restarts PL to 0 (300W draw), witch overheats GPU and speed is lowering step by step.
those two 3080s, are the same brand/model ??
Just a fact: 2 different models, and +1000 is just 2 diffrent clock rates.

[FloridaKid]

I'm suspecting the temperature of the GPU, if it's too high the card will throttle and high clock will perform very worse, listen to yourself OP you said after 30 to 60 Minutes the performs or clock speed will reduce itself, that's more like overheating issue

[FP91G]

Remove this video card from the farm and test on a separate computer until you find the optimal settings.
You need to read reviews about your model. You may need to change thermal pads to reduce the temperature of the memory and processor, or add additional cooling.Test with standard settings without overclocking.

[IceTea91]

So after your input my first thought was to test separating the system so i removed everything but the black sheep. I overclocked with 1400 mhz and it crashed immediately so this should be about the card itsself. After that i let that run for about 90 minutes, the temperatur was about 59 C° in average. The Hashrate was not falling, nothing else was suspicios. So i lowerd the memory mhz from 1000 before to 800 and the power limit from 73 to 64.

I switched the Riser and the USB Cabel just in case while i was installing my other 3 cards again. Its running for now, we will see how long. I can work on long time observations with gpu-z at saturday or something since the rig is not running in my apartment.

And I will looking forward to get rid of afterburner (which is the newest Version) as someone mentioned before. Got to do some reading for that first ;-).

By the way both 3080 are the same type of GPU, its the Founders Edition.

PSU: LC Power LC1800 V2,31 Mining Edition

I am using about 800 Watt while running the system with those 4 cards at the moment. I was about to add 2 or 3 more when i can get something which isnt too overpriced.

[Inkdull]

800 watts for 2x 3080 and 2x 3070s??? Bro that's too small to handle those four GPU, that's exactly where the problem is coming from, since you like overclocking the GPU will draw more power, forget what your software are showing you, to get the accurate power draw you need Kill-a- watt matter

[FloridaKid]

That's true, imagine each GPU drawing 233 watts the the math yourself that's over 900watts, definitely not enough power to run your GPU, you need a higher PSU like 1200 watts or two power supply using add2psu

[IceTea91]

That’s a misunderstanding, sorry but English is not my first language

What I meant to say is that my system is using 800 watts. The PSU can give out to 1800 Watts. So there is plenty of room for power spikes. I hope I could explain it better this time

[ummasek]

PPL, you`re reading what op`s writing ?
Good PSU.
It`s OC problem in my opinion - aggain.
Check GPU-z monitoring on both. It`s strange, but not impossible to owe diffrent OC gpu`s of that samne brand.

[FP91G]

So after your input my first thought was to test separating the system so i removed everything but the black sheep. I overclocked with 1400 mhz and it crashed immediately so this should be about the card itsself. After that i let that run for about 90 minutes, the temperatur was about 59 C° in average. The Hashrate was not falling, nothing else was suspicios. So i lowerd the memory mhz from 1000 before to 800 and the power limit from 73 to 64.

I switched the Riser and the USB Cabel just in case while i was installing my other 3 cards again. Its running for now, we will see how long. I can work on long time observations with gpu-z at saturday or something since the rig is not running in my apartment.

And I will looking forward to get rid of afterburner (which is the newest Version) as someone mentioned before. Got to do some reading for that first ;-).

By the way both 3080 are the same type of GPU, its the Founders Edition.

PSU: LC Power LC1800 V2,31 Mining Edition

I am using about 800 Watt while running the system with those 4 cards at the moment. I was about to add 2 or 3 more when i can get something which isnt too overpriced.

What is your video card memory? If micron, then you need to set negative values.

I would recommend that you use the Hive OS, because it is very easy to configure video cards and you do not need additional software. Do not lower the power limit, try testing your problem video card at standard settings

[IceTea91]

Solved:

It was the temperature. Just like the people above guessed it was about the heat which was generate by the card. I am running the card now with 800 memory overclock while I increased the fanspeed from 65 to 90 and it works for 4 days without a single up or down in the reported hashrate. Maybe if I wanted to go for more memory overclock I have to go for the advice with changing the thermal pads but i am fine for now.

The update is just for people with similiar problems.

I will change to Subject to solved.

Ty all for your help, I will take some tips with me, for example to get rid of Afterburner.