ETH Mining problem

[1337leet]

Hey guys,

I have an ETH Miner running with several RTX 30xx cards.
I run t-rex Miner on Windows 10.
I have no monitor plugged in, just teamviewer to monitor the rig.

It usually runs perfect for 2-7 days and then it „crashes“.
But I can‘t figure out why.

The system has 840W power usage atw and I run a 850W Seasonic Focus GX850 on it.

So every 2-7 days I get a message from my pool that my rig went offline.

When I get this message the System reports 0 MH/s and it uses 730W atw.
But I can not get access via Teamviewer. It will not connect me to the System.
The system will not come back before I restart it.

When I restart the system it will work fine again for 2-7 days.

So what‘s the problem?
- If a card crashes the miner would restart
- If the PSU cuts off there wouldn’t be 740W power usage
- If the internet went off the miner would restart after some time

I have no clue what else I should check.

[1714971234]

1714971234

[1714971234]

1714971234

[FP91G]

You can configure Windows to restart automatically every 2 days.
But in practice, you need to look for the reason.
It is necessary to reduce the overclocking of video cards. Remove 1 video card from the farm and test. If after 2-7 days a similar problem occurs, then remove another video card, and replace it with the one that was removed before.

[rdluffy]

In my opinion, you have two things to consider

You're pushing too much on your PSU, you're using almost 100%, it's not good, better to use at least 1200w PSU Gold rated
PSUs are better at 70 to 80% of load, you can use 100% but for short periods of time

Windows 10 have some issues and if a card crashes, this card will go back to default values, and if you are using 840w with underclock, the moment your card (s) go to default, it's pushing more watts and your PSU will turn off to protect itself

Another solution is to use HiveOS, it's way better than windows 10, more stable, and with the new updates, you can fix the core clock for Nvidias RTX 3xxx and consumes less power, and become even more stable

[1337leet]

When one card crashes the miner should not go to 0 MH/s and or would restart the card.

And when a card crashes I should still be able to connect with teamviewer.

It looks like the whole system freezes but the cards continue to run (740w power consumption)

I have no clue.. it just doesn‘t make sense.
And I need to set the msi afterburner settings manually after each restart, so an automated windows restart wouldn‘t help me.

[1337leet]

In my opinion, you have two things to consider

You're pushing too much on your PSU, you're using almost 100%, it's not good, better to use at least 1200w PSU Gold rated
PSUs are better at 70 to 80% of load, you can use 100% but for short periods of time

Windows 10 have some issues and if a card crashes, this card will go back to default values, and if you are using 840w with underclock, the moment your card (s) go to default, it's pushing more watts and your PSU will turn off to protect itself

Another solution is to use HiveOS, it's way better than windows 10, more stable, and with the new updates, you can fix the core clock for Nvidias RTX 3xxx and consumes less power, and become even more stable

I know the PSU problem. But it is rated for 850 output which is more like 900 atw. And seasonic has some beefy buffers. They say there PSUs can handle 10-15% more than officially rated.

But even if it shuts off then the whole pc would turn off or restart.
I don‘t start the miner automatically on restart- so the system should not use 740w.
And if the system restarts I would be able to connect with Teamviewer.

[1337leet]

I guess it has something to do with Windows or the miner freezes.

That would explain why there is still a lot of power draw and why there is no internet connection or why I can not connect with Teamviewer.


I already ordered a Seasonic Prime TX 1000.
Let‘s see if the system is more stable with more power.

[FP91G]

When one card crashes the miner should not go to 0 MH/s and or would restart the card.

And when a card crashes I should still be able to connect with teamviewer.

It looks like the whole system freezes but the cards continue to run (740w power consumption)

I have no clue.. it just doesn‘t make sense.
And I need to set the msi afterburner settings manually after each restart, so an automated windows restart wouldn‘t help me.

I have msi afterburner saves the settings for each video card and even after restarting it runs the saved parameters.
You need a monitor to test your farm. You can even use the oldest monitor like the SyncMaster 795, but you should see what is happening on the screen or nothing is happening there.
The whole farm may freeze for various reasons. This can be a problem in both the power supply and the video card, so you need to look for the problem using the elimination method.

[rdluffy]

I guess it has something to do with Windows or the miner freezes.

When I changed to HiveOS, all my problems were solved, no more freezes or crashes
Windows 10 updates are horrible and there's always something to fix, and even if you stop the updates, the system continue to update after some time

But try HiveOS, it's simple, you can burn on a flashstick or in an small SSD, in a few minutes you're mining, and after the config, your system can run stable

[Wotan Wipeout]

"Several" 30xx GPUs means ? 6 ? on a 850 Watt power supply?
I would go for more than 1000 Watt, even when you dont need it always.
If possible go for a 1200 / 1250 Watt.
Stay in using less than 80%.

[FP91G]

I guess it has something to do with Windows or the miner freezes.

That would explain why there is still a lot of power draw and why there is no internet connection or why I can not connect with Teamviewer.


I already ordered a Seasonic Prime TX 1000.
Let‘s see if the system is more stable with more power.

I don't understand your logic. You can turn off 1 video card and check the performance of the mining farm with less electricity consumption. If this helps, then of course it is worth buying a more powerful power supply unit or a second power supply unit for 500-600 watts + a synchronizer for starting power supplies.
And you spend more than $ 500 on a new power supply unit without knowing that it will help.

[arielbit]

Check your risers and riser cables

[sxemini]

This happens if you have to much OC, the driver do a reset and that is the reason why you can´t connect to the system via teamviewer.

[1337leet]

This happens if you have to much OC, the driver do a reset and that is the reason why you can´t connect to the system via teamviewer.

Why?
I can‘t even connect. It‘s not like I get a black screen.

[1337leet]

Check your risers and riser cables

Good point, but what should cause the mentioned problems?

[1337leet]

And you spend more than $ 500 on a new power supply unit without knowing that it will help.

I wanted to build an other system and needed it anyways.
But it has 3 weeks delivery time

[1337leet]

Maybe I really need a monitor so I can see what the system looks like after crashing.

At the moment I can not connect, so I won‘t see what‘s going on.

[philipma1957]

That psu can not do 840 watts 24/7/365

It freezes the system nuff said.


Now 1st question is what are several 30xx cards?

is it 3x 3090 cards
is it 3x 3080 cards
is it 4x 3070 cards
is it 4x 3060ti cards
is it 4x 3060 cards


I do not know if you simple have great clocks but 1 card to many or


I can tell you right here right now windows drops power settings goes to TDP and it does it at random.

Been doing it since 2012 when I mined btc with amd 5770

does it with nvidia
does it with amd.

So to help you

tell us all your gpus
tell us your clocks

we can figure if you are clocked perfectly fine but have an extra card and the psu turns off due to thermals.
or we can figure if you are getting runaway tdp on 1 or more cards and the psu turns off due to thermals

I wish I had money for every over taxed system I read about on all the posts I have read I would be rich.


No ATX runs 24/7/365   at 840 of 850 for a long time.

[arielbit]

Check your risers and riser cables

Good point, but what should cause the mentioned problems?

Maybe I really need a monitor so I can see what the system looks like after crashing.

At the moment I can not connect, so I won‘t see what‘s going on.

recently i have a two or three failed risers on separate rigs, different timeframe.

sometimes the card(with riser problem) is not hot to touch unlike the other cards in the rig....sometimes cards with different light color when idle and running shows where the card with issue is..

on the mining pool i have a view of the number of workers, if the number of workers goes down i check what rig it is. if it is a reoccurring or suspicious to me if there is a problem, i remote view for 24/7 that specific rig so i can catch whatever is the issue.---note: this is on a separate monitor-not my main monitor.

tip: you can jam pack your pool monitor/ surveillance / remote view in a single monitor by window-sizing, putting one on top of the other (get creative).

[1337leet]

I run the following cards:

Zotac RTX 3090 Trinity @ +1150 Mem, 302W, 121 MH/s - on a riser
EVGA RTX 3090 Kingpin @ +1000 Mem, -200 Core, 285W, 119 MH/s - on a riser
Zotac RTX 3070 Twin Edge @ +1200 Mem, -100 Core, 130W, 61 MH/s - on a x16 Board Slot


In total that’s only 717 watts on 12V. The PSU is rated to 840 on 12V.
The rest of the system is about 50 watts.

So about 767 watts in total. 840 watts atw.

As soon as I lower the power limits the cards will not run stable or drop by 15 MH/s if I reduce 1% PL.

[FP91G]

I was not too lazy and counted the number of 8 pin connectors for connecting these three video cards. I counted 7 8 pin connectors.
I also don't understand which raisers are used and how they are connected.
For RTS 3090 6 pin raisers should be used. For such a mining farm, I would use a server power supply with the required number of connectors.