Claymore restart alone : WATCHDOG: GPU 1 hangs in OpenCL

[joboy84]

Hello, 

A few times per day, claymore (v.10) need to restart from alone and I don't understand why.   

Here is my setup :

Bios : Biostar TB250-BTC Ver. 6.x. 
GPU : Sapphire RX580 Nitro+ 8GB
OS : win10 pro (fresh install)
Driver : Win10-64Bit-Crimson-ReLive-Beta-Blockchain-Workloads-Aug23 
Ram : 8Gb

GPU-Z :
   
Graphics Card :


Sensor :


Claymore settings : 

-allpools 1 -ftime 10 -retrydelay 3 -tstop 85 -tt 65 -cclock 1150 -mclock 2200 -cvddc 850 -mvddc 850

 

Getting 0 rejected for ETH and approx 2% rejected on DCR. 

This are the infos from the logfile : 

15:25:57:738   ee8   em hbt: 0, dm hbt: 0, fm hbt: 78,   
15:25:57:738   ee8   watchdog - thread 0 (gpu0), hb time 62 
15:25:57:738   ee8   watchdog - thread 1 (gpu0), hb time 156 
15:25:57:738   ee8   watchdog - thread 2 (gpu1), hb time 75296 
15:25:57:754   ee8   WATCHDOG: GPU 1 hangs in OpenCL call, exit 
15:25:57:754   ee8   watchdog - thread 3 (gpu1), hb time 75391 
15:25:57:754   ee8   WATCHDOG: GPU 1 hangs in OpenCL call, exit 
15:25:57:754   ee8   OC v7, Reset control for GPU 0, close miner right now if you want to use default control from Catalyst   
15:25:57:926   ee8   OC v7, Reset control for GPU 1, close miner right now if you want to use default control from Catalyst   
15:25:59:035   ee8   Restarting OK, exit...   

 
Sometimes by restarting the computer I get this message :

"Default Radeon WattMan settings have been restored due to unexpected system failure" 

What does it mean ? How can I solve this problem ?

Thank you

[treanski]

you overclock to much, you undervolt to much or your modded bios sucks

[xxcsu]

Claymore settings : 

-allpools 1 -ftime 10 -retrydelay 3 -tstop 85 -tt 65 -cclock 1150 -mclock 2200 -cvddc 850 -mvddc 850

 

Your card is way overclocked / undervolted  , not too many cards can handle 2200 / 850 for mem  and stay stable for long run .
get the mem speed lower first and see whats happening...

[joboy84]

To be honest, I'm not very comfortable with OC/UV. This is all new for me.

I have searched for some tutorials but haven't found one where the whole thing is explained.

I took a look on well known websites and copied their values. (BuriedOne and 1stminingrig)

Buriedone has this conf in MSI afterburner :

Core Voltage : -96
Core Clock 1340
Memory clock 2250

But when I set up the same settings in MSI afterburner, my hasrates are very low (1-2 mh/s)

From what I have understood and read, the core and memory voltage are used to make the cards consume less energy. The initial value is 1000.

I don't really understand the value at Buriedone, using MSI afterburner. Is the equivalent for Claymore 1000 - 96 = 904?

Where I'm lost is with the core clock and memory clock...

1stminingrig :

-cclock 1130 -mclock 2150 -cvddc 850 -mvddc 850 (900 mh/s decred 30 mh/s eth)

They have a 512kb Bios and I have a 256kB. Don't know if this has a influence.

And in some posts I saw this values :

-cclock 1150 -mclock 2250 -cvddc 850 -mvddc 850 -> 800 mh/s decred 26/27 mh/s eth
-cclock 1130 -mclock 2200 -cvddc 850 -mvddc 850 -> 890 mh/s decred 29 mh/s eth
-cclock 1200 -mclock 2200 -cvddc ? -mvddc ? (put 850) -> 910 mh/s decred 30.5 mh/s eth
-cclock 1200 -mclock 2250 -cvddc ? -mvddc ? (put 850) -> 930 mh/s decred 31.5 mh/s eth

Reminder : My values :

-cclock 1150 -mclock 2200 -cvddc 850 -mvddc 850

The 2 last ones are great but at this time I don't know if the cards will crash because they have higher cclock that mine. What do you think ?

@xxcsu : You said that not too many cards can handle 2200 / 850 for mem clock. But in all of this exemples I found, they are at 2200+ in mem clock. I will try to put it at 2100. If I put 850 up, that would mean my card draw more kw ? If I put the cclock up, could this stabilise my card with 2200 mclock ?

Thank you

[xxcsu]

Every cards overclocking potential is a little bit different ...
with gpuZ you can check the ASIC quality
i would say if the asic quality is almost the same for all of your cards in the same rig , you can tune them together . If there is a big different with asic quality between your cards , and you want the max out from your cards , then you have tune your cards one by one for a long term stability .

Big different with asic quality , i mean some of your cards have 90% ASIC quality , some under 70%

[joboy84]

ok, just checked and for one card the ASIC quality is 71% (didn't put the second card online at this time)

I put only one card online for the moment for monitoring.

Actually I undervolted at 920.

The GPU memory clock is at 2200 (maybe I can get it a try a little higher)

Power draw is at 135/140w (power meter), card is running at 61°C and I got less than 0.5% rejected shares for dcred.

Hashrates : 31.15 mh/s Eth and 935 mh/s for decred. Not a single rejected share and no memory errors !! Seems to be a perfect match

System is running stable since 4 hours. I will let this card running the whole night. Tomorrow I will put the other card online (alone) the whole day. Then, by night I will put both GPU's online again.

What I don't understand is the core clock. What will it do if it's lower or highter ?

I saw some people saying that the core clock doesn't really affect the values for ETH. But for some other algorithm higher values are better. Actually I have it at 1300.

On the packaging of the cards they mentionned that in OC mode the card will be at 1411 Mhz...