3070 ti randomly drops hashrate and power drasticly on ergo, win10 nbminer 39,7

[Blackwing215]

I have an issue with ergo on version 39.7, win10 21h1, driver 496,46

After hour to several hours of mining hashrate drops from about 180 to 60-70 MH/s and power from about 185 to 65-95 W. And when this happens - other algos are broken in the same manner, even 39.5 version of miner. Restart fixes this, until next random drop. I haven't found any correlation with overclock (if it stable) or running stock. Here are some screenshots: My overclock settings and power consumption pattern after drop

NBMiner header

Console output after drop

Normal console output

I have never had any issues with nbminer on eth, etc and raven. Now i'm running raven on nbminer 39.7 without any issues in +100 core +1500 mem overclock, totally stable for a couple of days.

[Blackwing215]

Oops, i was too fast to say raven is stable too, while i was writing that issue - it did the same thing. Hashrate droped from 39,5 to about 18 Mh/s, power - from 280 to 130 W. In both cases GPU load is 100% (according to MSI Afterburner)

Raven on NBMiner 39,5 has been working for week of uptime without problems tho

[scattergame]

I met the same issue before. I noticed that when this situation happens the mem clock is fixed at 5001, which is extremely low. To solve this quickly, you can try to lock cclock frequency @ 1500 - 1550 with msi afterburner.

[PatrickGlatt]

Had this issue too on dagger. used an auto restart batch for nbminer to restart it every 2h. Mabye set it to 1h.

==================
@echo off
title NB Autorestart
SET ProcessName=nbminer.exe
SET RestartDelay=7200
REM Delay is in seconds!!!

:loop
taskkill /F /IM %ProcessName%
timeout /t %RestartDelay%
goto loop
===================

[Blackwing215]

@scattergame recently i had this issue with ETH too with locked cclock 1150MHz - so it looks like it is not 100% solution

@BluefireXD unfortunately simple restart of miner doesn't work for me at all - only restart of machine. And since this is my desktop, which i'm using not only for mining - autorestart is not an option, when you can just use previous version of miner)

Optimal solution for me was using version 39,5 of miner. I haven't tried it with ergo, but for raven and ETH it works fine. Thanks for your suggestions tho!

[vladtenet]

I am having this exactly same problem with drivers 512.15 in two separated machines with 3080ti and 3070ti When I was using drivers 472.12 it never happened. Sometimes if I lock the mem solves, and sometimes I need to restart the machine. Did someone find a solution?