strange issue with T-Rex in NiceHash

[sepheronx]

So I am running nicehash through t-rex plugin and it was all going perfect. Something like 4900/4917 and then all of a sudden, I leave, come back and it was at 10/10 as if it restarted itself. Then my MH/s was low, like 40 MH/s. So I restarted it, ran 3070 at default to let the DAG file build itself, then I did my custom OC (-500 core, +800 MEM, 54 Power limit) and it went from 44MH/s down to 30MH/s. So I set to default settings which crashed t-rex and then restarted itself. Let itself build its DAG file again and then set my custom OC again. I see it is running now about 58MH/s as it was prior to initial crash.

Can someone explain to me what has happened? Is there something I should be concerned about regarding my GPU? Did I screw it up with its OC? Or is this something that is rather common? So far it is 19/20 (1 Failed). Yesterday and day before, I didn't get a single fail till at least 1000 job.

 

[P4-630]

Seems like failing GDDR6....

j/k
I hope not.

 

[sepheronx]

Seems like failing GDDR6....

j/k
I hope not.

If it is, I will have to obviously RMA it.

 

[hat]

It sounds like it might be unstable? Try a higher power target, or a lower memory speed, maybe both.

 

[sepheronx]

It sounds like it might be unstable? Try a higher power target, or a lower memory speed, maybe both.

I find it strange.

Am I the only one with a 3070 with Samsung vram that cant OC at all? Near stock settings I get failures while everyone else gets +1100 or higher.

Did I get a bad card?

I increased power to 60%. Hopefully this will fix the issue.

 

[P4-630]

I find it strange.

Am I the only one with a 3070 with Samsung vram that cant OC at all? Near stock settings I get failures while everyone else gets +1100 or higher.

Did I get a bad card?

I increased power to 60%. Hopefully this will fix the issue.

Personally I had luck with vram overclocking on my last 2 cards, but the core would barely do +35.

No graphics card is the same even when they have the same brand of vram.

 

[sepheronx]

Personally I had luck with vram overclocking on my last 2 cards, but the core would barely do +35.

No graphics card is the same even when they have the same brand of vram.

I never really had luck OC anything tbh. I'll test the GPU tonight at stock settings to see if I get any artifacts (afterburner stress test). If I do, I guess I gotta start RMA process.

 

[phill]

If I'm honest I never bother with clocking the VRAM or even the Core on my GPUs because the majority of the time, they just don't add anything to it. With them mining, a 100MHz extra can make a good difference with the card I'm using but I have to ask if it's really worth the extra heat and hassle that these cards seem to have. Right now just trying to keep the Memory Junction Temp below 90C seems impossible whatever I do right now... This is at the moment, is stock RAM timings...
If you can run the card with a lower power limit and don't worry about stressing the RAM so much. Check the temps of the RAM chips too, if they are getting too hot they will lower performance, a lot.

 

[sepheronx]

If I'm honest I never bother with clocking the VRAM or even the Core on my GPUs because the majority of the time, they just don't add anything to it. With them mining, a 100MHz extra can make a good difference with the card I'm using but I have to ask if it's really worth the extra heat and hassle that these cards seem to have. Right now just trying to keep the Memory Junction Temp below 90C seems impossible whatever I do right now... This is at the moment, is stock RAM timings...
If you can run the card with a lower power limit and don't worry about stressing the RAM so much. Check the temps of the RAM chips too, if they are getting too hot they will lower performance, a lot.

3070 doesn't have the sensor so no way to check the mem temps.

3070 doesn't have the sensor so no way to check the mem temps.

So I adjusted the clocks again, down to 500 on mem

Nothing wrong with GPU though. Ran it at all stock settings, ran some games (cyberpunk and Tropico 6) and it all looked fine. Besides Cyberpunk, some strange issues that doesn't seem GPU related, like texture pop ins when you are looking/moving a certain direction. Tropico 6 stresses that GPU hard and it seems fine, no artifacts. Ran MSI kombustor with artifact scanning for over half an hour and no artifacts were found.

So I think it's safe to say the GPU is running perfectly fine. From my understanding, the VRAM doesn't run hot like GDDR6X does.

 

[phill]

If you run it at stock, does that lower down the hash rate very much at all? Can you touch the back of the card? Not massively hot or anything?

 

[sepheronx]

If you run it at stock, does that lower down the hash rate very much at all? Can you touch the back of the card? Not massively hot or anything?

Never tried touching it tbh. I'll try it.

Hash rate at stock is about 51MH/s

Never tried touching it tbh. I'll try it.

Hash rate at stock is about 51MH/s

so I tested by placing my hand on the backplate and its hot. Not like "Oh ouch that is hot" but more like "I Cant keep my hand on this for long" hot.

 

[phill]

Get a fan or two over it, it might be its just getting too hot and the VRAM is just throttling.... Definitely worth a go...

 

[sepheronx]

Get a fan or two over it, it might be its just getting too hot and the VRAM is just throttling.... Definitely worth a go...

I will have to re-arrange my CPU cooler (H60) as its tubes are touching the top of the backplate.

Although, if it was hot, like 90C hot, then I wouldn't be able to put my hand on the backplate. Even if it was about 70C. It shouldn't be thermal throttling at least till 90C, no?

Right now I have it at +600 for mem, and still -500 for core at 57% power. I have so far 5100/27/0 in Phoenix. So 27 rejected shares (18 Stale) out of 5100 isn't bad.

So rejection rate is below 1%. What do you think, do you think that it sounds like issue is resolved?

 

[phill]

I don't think its anything to worry about. So far out of 48 hours mining, around 4360 shares accepted, I've 2 stale.. Only if you start getting a lot more stable/rejected shares would I worry about it. That said, get a fan over the top back part of the card, even with the cables around the back plate, I think some air over it is better than none

It says my Memory Junction temp is 96C (average of 94C over 48.5 hours) but I can put my hand on the back plate and it's hot and you know it but it's not so hot you couldn't touch it. Hopefully some heatsinks I've ordered will come and I can sort it out a little better

 

[sepheronx]

I don't think its anything to worry about. So far out of 48 hours mining, around 4360 shares accepted, I've 2 stale.. Only if you start getting a lot more stable/rejected shares would I worry about it. That said, get a fan over the top back part of the card, even with the cables around the back plate, I think some air over it is better than none

It says my Memory Junction temp is 96C (average of 94C over 48.5 hours) but I can put my hand on the back plate and it's hot and you know it but it's not so hot you couldn't touch it. Hopefully some heatsinks I've ordered will come and I can sort it out a little better

I'll have to more or less take the whole thing apart as space is a problem. So far, at least touch wise, it feels as hot as my other machines 390X which its memory temps are 60C right now.

Other than that, no idea what else. I cant get fans on there. May replace case fans with higher pressure ones in order to move more air. Other than that, cant say.

I mean, I can just quit mining, but then gaming may be an issue too with the card if it is this way, no?

I think I will just let it mine till I earn enough to buy a 3060ti and I'll let that card do the mining while I use my 3070 for every other use. I'll try to get fans blowing on the back plate, but from what I read others doing same thing, it didn't really change much in temps, if at all. Maybe ill give the heatsinks a try as well. Please let us know how well it worked out for you.

Phil, I need to ask you this question.

So lets say, after 5 days of running crypto mining and I had my GPU VRAM overclocked by various amounts during that time - 1200 in one day, 1100 for two and now two at about 500-700.

I moved it back down to its base clock (no OC on it) to see how well it does. 51MH/s average. I will let it run at that. I am not greedy so I am OK with $6CAD per day in mining. I mean, its better than nothing, right?

My question though is, what are the chances I ruined the GPU VRAM? I mean, I want to still use this card for gaming, and in this time frame, I could have caused damage, no? What do you think?

I have been building and playing with PC's for decades. But Overclocking was only really ever done on CPU for me or core on the GPU. Memory OC'ing is something I rarely if ever touched. So my experience on OC'ing is rather slim.

 

[RJARRRPCGP]

I find it strange.

Am I the only one with a 3070 with Samsung vram that cant OC at all?

Reminds me of my GTX 1660 Super (Turing) being the opposite, begging for a VRAM OC. I actually find that surprising with a late Nvidia. Albeit my VRAM OC was only +206, but it resulted in a good benchmark score.

 

[trog100]

just my thinking but i recon a max power setting less than 60% is too low..

cooling the baclplate wont do much unless its removed and some thermal pads are fitted in between the back of the card and the backplate.. i assume you wont want to do this..

if in doubt maybe leave the side off your case..

trog

 

[sepheronx]

Hey, quick quest, so I want to move my earnings to my coinbase account.

Now, I have read that there is a minimum you need to have in terms of BTC before you can send it. Is this still the case? Something like 0.001 BTC. I have 0.00083653. Does this mean that I cannot send it yet? Also, I presume that I will get a charged fee?

 

[hat]

Yeah, there's a fee. It's like 2% of the transaction or something. There's also a minimum transfer amount, not too sure what it is but it ain't huge.