After upgrade to 11.6 from 11.2, one of my Gigabyte Windforce OC GTX 1070 (GPU #8) keep crashing every 1~3 minutes.

04:33:45:433   18bc   ETH: GPU0 32.220 Mh/s
04:33:46:371   1748   GPU0 t=48C fan=69%
04:33:50:903   f78   GPU 0, GpuMiner cu_k1 failed 4, unspecified launch failure
04:33:50:903   f78   GPU 0, GpuMiner kx failed 1
04:33:50:903   f78   Set global fail flag, failed GPU0
04:33:50:903   f78   GPU 0 failed
04:33:50:903   a48   GPU 0, GpuMiner cu_k1 failed 4, unspecified launch failure
04:33:50:903   a48   GPU 0, GpuMiner kx failed 1
04:33:50:903   a48   Set global fail flag, failed GPU0
04:33:50:903   a48   GPU 0 failed

In 11.2 this GPU can run between 15~21 days straight w/o problem. Then I might need to restart miner/PC when similar issue above occurs (my guess is GPU #8 cannot survive 4 epoch changes since I use -eres 4)
In 11.6 the runtime has shortened to 1~3 minutes even everything else (OC setting via afterburn, claymores' command line arguements) stay the same.
I have tried to set -ethi manually between 8 and 0 to reduce intensity and increase stability, but it crashes nonetheless.

FYI,
This is a mixed rig w/ 1070, 1060, RX 580. I started 9 instances of claymores to run each GPU individually. When the GPU #8 crashes, the other 8 run normally w/o any issue.
I have another WF OC 1070 on the same rig, same setting everyhting. It runs smooth w/ 11.6
After the GPU #8 repeated crashed about 10 times, I decide to run it w/ 11.2. It runs very stable w/ 11.2 as before
So I have 8 GPUs run 11.6 stable, 1 WF OC GTX 1070 runs 11.2 stable.
The reason I upgrade to 11.6 for GTX 1070 is to reduce stale shares.

Any suggestions?