GPU hash rate being cut in half after a few hours of mining.

[TXmining]

I have a Radeon VII (2 of them), but one is decreasing hash rate to about half after mining for a couple of hours. The other continues fine, temps seem ok but I noticed the the watts being used declines about 30% on the card with declined hash rate. I’m using NiceHash and Asus B250 MB running Windows 10. Anyone have a solution for this?

[Philipma1957cellphone]

This belongs in alt coins not here I asked mod to move it.

How did you clock the cards?

How do you monitor cards?

It could be the card is over heating and down clocks to cool off.

[TXmining]

This belongs in alt coins not here I asked mod to move it.

How did you clock the cards?

How do you monitor cards?

It could be the card is over heating and down clocks to cool off.

Using MSI Afterburner to clock and GPU-Z to monitor. Temps are usually in the 60s but may hit 70c but the other card continues to operate at those temps. I did notice that after I pause/stop mining the CPU clock on that card is still going at 100% and memory clock is about 20% almost like it’s stuck on a different process. After I restart the system it begins operating normally but only lasts for a couple of hours.

[rdluffy]

Are you doing OC on this card?
Try to decrease to OC, and see how it goes

Put a fan directly on this card to see if it's something related to temps

Update the drivers...

[JayDDee]

I presume the cards are identical, same mem etc.

A history of the problem always helps. When and how did the problem start?

It's interesting that only one card has a problem.
The first thing to do is identify any differences in configuration, ie FW,
monitor attached.

If so move the monitor to the other card and see if the problem moves with it.

You could try different algos or miner SW to see if you can reproduce it
with different variables.

You could also try swapping slots to see if the problem follows the card or stays with
the slot.

Swap anything else you can think of that's in the path, risers, power cables.

The symptoms seem like a mem leak but that should affect both cards. If testing doesn't
give any clues I would suspect a HW issue with that card.

[TXmining]

Are you doing OC on this card?
Try to decrease to OC, and see how it goes

Put a fan directly on this card to see if it's something related to temps

Update the drivers...

I’ve got it clocked for efficiency so lower power, cpu but increased memory clock. 950 power, 1750 cpu, and 1100 memory. I tried default with same results. I believe I have most current driver, and the other VII has no issues.

[TXmining]

I presume the cards are identical, same mem etc.

A history of the problem always helps. When and how did the problem start?

It's interesting that only one card has a problem.
The first thing to do is identify any differences in configuration, ie FW,
monitor attached.

If so move the monitor to the other card and see if the problem moves with it.

You could try different algos or miner SW to see if you can reproduce it
with different variables.

You could also try swapping slots to see if the problem follows the card or stays with
the slot.

Swap anything else you can think of that's in the path, risers, power cables.

The symptoms seem like a mem leak but that should affect both cards. If testing doesn't
give any clues I would suspect a HW issue with that card.

They are both new, recently installed and I don’t have any monitors connected to the card, using internal graphics for the monitor. They have same specs and I already tried different slots, power cables etc. I tried the fans right on it which improved temps but still same issue. And it goes back to full hash rate after I restart pc. But simply restarting the miner doesn’t work, the entire system has to be restarted.

[JayDDee]

If it's a new card, it's defective, replace it.

[bluspirit]

If hashrate decrease it probably means that the card has memory errors. If this also happens with stock settings then the card is defective.
Did you check the GPU memory errors with HWinfo?

[TXmining]

If hashrate decrease it probably means that the card has memory errors. If this also happens with stock settings then the card is defective.
Did you check the GPU memory errors with HWinfo?

I have not. Any instructions on how to check that?

[bluspirit]

Googling "hwinfo", download the software and use it while you are mining with stock settings (no OC/UV), to check if the card has "GPU memory errors" (in sensors status).

[philipma1957]

sometimes a card has

 Hynix ram

 some times a card has

 samsung ram.

do they have the exact same ram.

gpuid software can let you know.

[JayDDee]

Different mem would be apparent from the start and wouldn't explain the drop
over time.

It's a new card, I would have returned it by now.

[philipma1957]

Different mem would be apparent from the start and wouldn't explain the drop
over time.

It's a new card, I would have returned it by now.

one could handle the clock.

the other does not. while you may say it means one is defective. it could mean one is great and the other is average.

I do agree if it is new he should return it as it is the same make and model. it should match the stats of the other.

if he gets a replacement he can then tell if the cards preform equally.

[TXmining]

Googling "hwinfo", download the software and use it while you are mining with stock settings (no OC/UV), to check if the card has "GPU memory errors" (in sensors status).

I don’t see any errors but now the card stops mining after about an hour with an error on the miner stating gpu error restart miner. So I guess it’s a defective card?

[bluspirit]

Googling "hwinfo", download the software and use it while you are mining with stock settings (no OC/UV), to check if the card has "GPU memory errors" (in sensors status).

I don’t see any errors but now the card stops mining after about an hour with an error on the miner stating gpu error restart miner. So I guess it’s a defective card?

Weird, but I think so.
Ask for replacement.

Different mem would be apparent from the start and wouldn't explain the drop
over time.

It's a new card, I would have returned it by now.

one could handle the clock.

the other does not. while you may say it means one is defective. it could mean one is great and the other is average.

I do agree if it is new he should return it as it is the same make and model. it should match the stats of the other.

if he gets a replacement he can then tell if the cards preform equally.

Totally agree.