Quote from: Beruang123 on June 15, 2019, 04:54:21 AM

Quote from: kerney666 on June 14, 2019, 06:07:58 PM

Quote from: netgeek3 on June 14, 2019, 02:20:52 PM

Quote from: arlekin on June 12, 2019, 11:07:20 PM

I have the same problem on 4 rigs with 4 various Vega 56 cards (ref with samsun g and asus/powercolor with hynix, with timings and without timings ). The card runs for an hour or two and then the hash rate drops to zero or to 1400 h/s with a DEAD message. I have observed in several rigs, but for some reason problems always with the first PCI slot (PCIe 3:0.0). If you do not stick the card into the first PCI slot (PCI 16x), then the problem disappears. Now I just do not use 16x pci slot Sad

I am using all 1x slots and am having the same behavior. I am seriously thinking that this is a driver compatibility issue with Win10 for a couple of reasons:

1) It is always (in my case) the first card enumerated on the PCIe bus
2) It doesn't matter what motherboard I'm using (have tried on 7 different)
3) Starting ANY other applications on the same system slows down ONLY the affected GPU, and it shouldn't affect anything, historically.
4) In digging through logs and in the rare cases of a BSOD, there are consistently "stuck thread" issues
5) This phenomenon is not isolated to TRM - I am having the same issue with SRB, Phoenix, Claymore and others

I did not have this problem with earlier builds, but since the latest Windows and AMD updates, I can't get rid of it. I have tried rolling back through every iteration all the way back to Blockchain drivers without success.

I should mention that I am using a mix of V64 and V56 cards in my rigs, but all have Samsung memory.

Hi! Just curious, did you have time to test 0.5.2 yet? It has a first wave of changes that only aim to stabilize the miner. No real guarantees, only some changes that theoretically should be nicer to the hardware. Otherwise, I agree, were dealing with some very delicate issues, and the software/driver/hardware interaction here is not all hunky dory.

Apparently the new release didnot fix the issue.. still have the same error

Ah interesting! You see, I didnt make the same small stabilizing adjustments in the main 4MB variant kernels. Obviously I should have, my mistake, I simply forgot. Will try fix them as well very soon and will let you know when a release is available.