Bitcoin Forum

Hi all,

need some help with a strange issue with one of my RX580 cards.
The card is in a 6 GPU rig, AMD only GPUs (but different 5x0 models).
It's a Powercolor Red Dragon V2 8GB RX 580, modded bios.
It's been running just fine since August, no issues, very little memory errors, stable undervolt at 825mV, 2100 memory clock.
The suddently today, the rig starts rebooting over and over. I pinpointed the problem to this card (rig runs ok without it!) - when it starts mining in Claymore Dual 10.2, after DAG creation, it just gets about 600 million memory errors and the rig reboots.
What I tried with to no avail:
- different 8-pin power cable
- different USB cable for the riser
- put it on a riser which was previously used by on of other cards
- reduced the undervolt and memory clock
- disconnected other cards (except iGPU from where I run the monitor)
- disabled it in Device Manager
- removed the drivers with DDU and reinstalled with this card only
- flashed the bios to stock

What happened when running this card only after reinstalling the drivers is especially weird. The card - even when disabled! - seems to be running the GPU clock at 1350MHz (memory stays at 300) and is heating up, the fan eventually kicks in... while the card is supposed to be doing absolutely nothing.

Any ideas apart from putting it into another system (which I don't have at hand atm)?
The entire rig started acting weird then I reconnected the other cards - had to run DDU a couple of times, disconnect all GPUs and reconnect (+ patch drivers) them one by one as I usually do with AMD rigs. When I connected the problematic card the system got unstable again, freezing etc.

- Win10 1709
- Aug 23 Blockchain beta drivers
- Gigabyte GA-H110-D3A mobo with Mining mode bios (PCIe Gen 2, 4G Encoding on etc)
- Corsair HX1000 PSU

Cards die. Sounds like it gave up the ghost! RMA it.

After flashing on Stock bios, try to install it solely in PCI-e x 16, exclude other GPU's let it mine and observe, if still the same, then it seems its already faulty and will agreed the one posted above me, better RMA it.

Yup, had to RMA the card, looks like the memory died - GPU-Z showed RAM at 0GB...

Now the other Powercolor RX580 is starting to show memory errors at settings that were just fine for 6 months, oh uh. Bad batch of cards in summer when there was a big shortage (like now...)???