Mining rig freezes every 2-3 days

[dan_729]

Mining Rig:

• PSU: Corsair RM1000i (1000W)
• Mobo: B450 Tomahawk MAX
• Processor: Athlon 3000G
• Storage: A400 240GB SSD
• GPUs:
  4 × Strix 3070@123W
  Core Clock +0
  Mem Clock +1200
  1 × Strix 3060 Ti@125W
  Core Clock -502
  Mem Clock +1200
• Other: 5 SATA powered risers
• RAM: 2 × 8GB
• OS: Windows 10 20H2 (latest)
• Mining Software: T-Rex 0.19.11

When it happens, the GPU fans continue spinning but not generating as much heat as when they are mining. The RGBs are also still present although nothing is responsive (i.e. mouse/keyboard are unable to wake up visual output).

I want to disconnect this rig from my monitor and let it run by itself - but it keeps freezing and I don't know what to do. Any help would be greatly appreciated.

[philipma1957]

drop  memory ram to 1000 see what happens

drop power to 120

do -400 core for the 3060 ti

do -300 core for the 3070
if it crashes more
consider phoenix

if it crashes with phoenix try doing smos and use phoenix

what pool?

I have two boards that drop out.

Over the 9 years of gpu mining I have had 5 or 6 boards with your issue.

sometimes it is psu
sometimes it is riser
sometimes it is gpu
sometimes it is mobo and just skip one connection pcie

sometimes no matter what you do the mobo won’t stay up.

[dan_729]

I will try that, althought I don't think the GPUs are at fault because they have been performing without Rejected Shares ever since I set them up a month ago.

I am using Ethermine pool.

[skablast]

Setting the miner to restart each day isn't an easier solution ? oops I saw now that you are on Trex & Nvda.
I don't know if Trex allows it. Phoenix for sure

[dan_729]

Hmm I will try setting tasks on the OS and see if that works

[adaseb]

Phil is correct. You can try 100 different thing and you will still get these freezes once in a while. Just be glad it’s not more common because it would be a hassle then having to reboot them all the time.

This is the only area where I prefer ASICS because those are plug and play. You just shut them off to clean the air out and that’s it. But with GPUs they freeze from time to time. I think the longest I ever got on any rig without any issues like hardware errors or dead GPUs was like 4 weeks.

[dan_729]

One dead GPU every 4 weeks sounds terrifying 

[Ceyflix-Rez]

Freezing when mining have become a normal thing when running on windows 10 OS, it happens even if you are using two or three gpus only on a motherboard, all I can say is try other mining software apart from Trex and also consider mining OS built on linux, they are more stable

[FP91G]

I would not look for a problem in the hardware first thing. This is either overclocking in memory or underreporting in the kernel.
Memory +800 +1000
Core 0
This mode is for tests only. If this does not help, then you should change the operating system. Better Hive OS and similar.
Linux systems are more stable, and if the problem is only in one video card, the system will disable it.

[rdluffy]

Some weeks ago I was having the same problem, but my rig was freezing once a day, and I'm used to mining and building PCs, I'm mining since 2017

Windows is good to mine with your main PC, like you have 1 or 2 cards and use the PC everyday, but if you have a rig, please change to a mining OS

I changed to Hive OS, but you have Simplemining, NicehashOS and a few others, it was the best thing I did, no more worries about my rig, it's running 24/7, no more babysitting my rig

Windows can update itself and reset some important settings, can disable your miner, drivers errors, restarts, blue screen etc

It's easy to try, put HiveOS on a USB stick or a spare SSD, and try for a couple of days

[logicalray77]

Sounds like a power issue. Check your risers power distribution. Preferably one riser on a single sata cable but we know it`s not always possible. Never power more than two risers on single sata cable.

My calculations your rig pulls about 800watts, that is high for only a 1000watt psu.

Disconnect one card and see how it goes.

[philipma1957]

Sounds like a power issue. Check your risers power distribution. Preferably one riser on a single sata cable but we know it`s not always possible. Never power more than two risers on single sata cable.

My calculations your rig pulls about 800watts, that is high for only a 1000watt psu.

Disconnect one card and see how it goes.

Yeah this is a good point.

He could disconnect 1 card or drop them all to 115 watts.  Yeah his hash will drop with way but 5 cards at 115 watts is 575+75 other = 650-675

Psu age from hard use. So 5 x 124 = 620 + 100 = 720 watts which could be hard on the psu.

[tg88]

I believe your problem is related to SATA cables, I had problems with the RM1000i before when I tried to connect more than 1 gpu per sata cable ... try to switch to molex that supports 2 gpus without problems.

[dan_729]

Thanks all of you! I will be doing some tests with your inputs for the next few days and see what I get 

[BitcoinBlaster51]

Just want to mention as a side note before getting into details. Virtual Memory settings on Windows may help. For ETH I've noticed VM size needs adjusted according to this formula:
Initial VM size = DAG size x Quantity of GPUs
Maximum VM size = Projected DAG size x Quantity of GPUs + Extra VM

In regards to the OP, I have been experiencing a similar issue described here: https://www.reddit.com/r/gpumining/comments/80r6j6/does_anyone_elses_windows_10_start_lagging/ Saw how the Reddit user Reed1981 mentioned using the "ethi" command for Claymore's dual Ethereum miner. This command adjusts the mining intensity, it's scale is from 0 - 8. Default is 8.

Although, I have been mining ETH with Ethminer through miningpoolhub.com. So I searched for a parameter list of Ethminer here: https://gist.github.com/jkcdarunday/e89dde8fa71e510596f45480c5529376
Noticed talk regarding OpenCL (Open Computing Language) configurations here: https://forum.ethereum.org/discussion/comment/31821/#Comment_31821

So far, I've adjusted:
--cl-local-work 64 (down from 128)
&
--cl-global-work 524288 (8192 * 64)

The default is 8192 that I found within the Ethminer parameter list.
--cl-local-work best operates with powers of 2 so I just scaled down one increment.

So far I will adjust these values if my rig continues its current issue. You may be able to attempt changing such values with T-Rex although I am not familiar with T-Rex.

Mining Rig:
PSU: EVGA 1600W G2 SUPERNOVA 80+G
Mobo: ASUS B250
Processor: INTEL I5-7500 @ 3.40Ghz
Storage: Kingston SUV400S37120G SSD 120GB
GPUs: 6 × Gigabyte GTX 1070 8GB
Core Clocks +125
Mem Clocks +270
PCI-e power adapters: 6 pin to dual 8 pin (6+2) VGA Y-splitter extension cables
Other: 6 risers powered by the extension cable above
RAM: 1 × CRUCIAL 8GB DDR4 2400Mhz SPLT CL16
OS: Windows 10 Pro v2004
Mining Software: Ethminer-0.18.0-cuda10.0-windows-amd64

[MRKLYE]

Make sure your power supply is on multi-rail mode.

Found that single rail mode on those power supplies under a decent load would tend to result in weak voltage readouts. Could be enough to cause a hang.

Also try and set up a script or something to clean out the miners RAM and virtual memory every day or so. It'll help avoid crashes hopefully. Good luck.

[BitcoinBlaster51]

Mining Rig:

• PSU: Corsair RM1000i (1000W)
• Mobo: B450 Tomahawk MAX
• Processor: Athlon 3000G
• Storage: A400 240GB SSD
• GPUs:
  4 × Strix 3070@123W
  Core Clock +0
  Mem Clock +1200
  1 × Strix 3060 Ti@125W
  Core Clock -502
  Mem Clock +1200
• Other: 5 SATA powered risers
• RAM: 2 × 8GB
• OS: Windows 10 20H2 (latest)
• Mining Software: T-Rex 0.19.11

When it happens, the GPU fans continue spinning but not generating as much heat as when they are mining. The RGBs are also still present although nothing is responsive (i.e. mouse/keyboard are unable to wake up visual output).

I want to disconnect this rig from my monitor and let it run by itself - but it keeps freezing and I don't know what to do. Any help would be greatly appreciated.

An increase in virtual memory seems to have helped. Maybe try increasing VM and see what happens after a few days. That's what I'm currently doing. If the machine behaves normally for more than 2 or 3 days, I'd say it's safe to assume VM. Could use HWinfo64  
Update: It's been well over 3 days as of now and the machine runs better. Although I have noticed insufficient virtual memory within Event Viewer due to the ETH DAG size continually growing.