Quote from: premiero on May 25, 2017, 10:51:22 AM

It's 99,9% graphic memory realted and heat combined issue. I have 3 rigs based on Asus cards, same mainboards as You have. 2 Rigs 6 x Asus RX480 8GB DUAL and 1 Rig 3xRX480 + 1x Asus R9 290. RM860 / AX1200 / OCZ1250 PSU's. I spent lot of time to diagnose the issue and i can say one thing - avoid to use any of Asus 480 8GB cards. Initially, just like any other i copied 1750 to 2000 memory starps, set all of my 480's to GPU:1010@980mv / MEM:2000@1000mV giving me 28,3 MH/s each while -dcri set to 6 and -ethi set to 16. Than it started. Hangs, freezes, bluescrrens, no signal at all etc. Raising GPU/memory voltage to stock value didnt help at all, neither lowering intensity. So i started to mine with only one card, one by one, to check what's wrong. Out of 15 GPUS: one got RMA, because it even fails at stock bios (no mod at all) and stock clocks (gpu hangs within minutes at PCIEx16 slot, I had to wait almost 3 months for new card!), 4 GPUS cannot handle correctly 2000 MHz memory speed (altough there was no Memory errors at HWINFO), i had to lower it to 1950. Another 2 cards was very temperature sensitive (temps higher than 69 caused hangs - it's very strange cause i tested few other cards and thhey did work well even at 85). One GPU cannot handle even 1950 MHz and i had to set it to 1900 MHz. They are much more stable ATM. GPU temperatures varies from 64 to 73, VRM hit 84. Since then, one rig is rock stable (do not hang at all), other 2 rigs still hang up (and i do not know why, because I have no more "GPU x hangs in OpenCL call,..." i suspect it's network related issue) but now it's like once per 4-7 days, rather than 5-10 times per day. I ended up with stats like below (rejects are due to weak wi-f signal).

115.974 MH/s, 12615/2/1 (0.02%): 28.115+28.110+28.131+31.616 MH/s
164.959 MH/s, 1729/1/0 (0.06%): 27.388+27.393+26.572+27.386+28.089+28.128 MH/s
168.027 MH/s, 2468/0/0 (0.00%): 28.119+28.132+28.128+28.126+28.131+27.388 MH/s

PS1. Claymore 9.1 is more stable for me than 9.3 (single mining)
PS2. There is some corelation between loosing Internet connection and hanging GPU's.

Yeah, after few modification, I loose GPU x hangs in OpenCL but it freeze without explanation every few hours. What I found now is UberMix 3.1, memory timing on one GPU bios forum.
UberMix v3.1 memory timing: 777000000000000022CC1C00AD615C41C0590E152ECC8608006007000B031420FA8900A00300000 010122F3FBA354019

After this, my rig works (and still works) flawless without any problem for 5h. I used original bios and change timing 2000 with this one. It is rock stable 175MH/s with sia coin 3160MH/s. I set undervolt -48mV, powertune 0%, core clock 1100MHz and memory clock 2000MHz. First it was set to 2050MHz and it make over 180MH/s.

There I found also another timing:
v2.3 -Less Extreme: 777000000000000022CC1C00AD615B41C0570E152DCB7409006007000B031420FA8900A00300000 010123A46DB354019

I tried it work very good for now.

panda86

Posts

Merits