Having some weird issue.
0.3.7 is far less stable than 0.3.6 when initiating.
Running 7-7 conf on 4 4GB cards.
When initiating the miner, one of the GPUs will have a thread failing and producing tons of HWs.
For some reason, seems to happen more often on XMR. When starting to mine graft, it does not appear so often.
WHen launching it on 0.3.6, no issues.
Yeah, with 0.3.7 we solved a lot of init issues, but a few weird new ones appeared as well. Are you under Linux or Windows? All known bugs right now are actually driver bugs. One example is the win driver that allocates the same mem space for two kernels in some multiple Polaris gpu settings. We triggered it when solving an earlier bug. We should hopefully solve it in the next version by addressing some code sizes.
The bug youre describing sounds more like a Linux driver bug weve seen though. On every fourth restart of the miner, one command queue will go nuts and just fail running the kernels, but there is no error returned from the opencl api. On the next restart, all is fine again. Three more restarts and the same thing happens. Havent seen this on win though, so curious what os youre running...