I saw that happening when I was trying to figure out how much to overclock my gpus. It means your overclock settings are unstable, try lowering them.
Yes that may be. The main loop currently goes like this:
Look if any card has completed some work successfully.
If yes --> Offload old work, send it to tester against target, count sols found and Issue new work
if no --> Wait for a short time and look again.
Well while this works most of the time this loop has one flaw: If due to OC or other circumstances the OpenCL code fails in its execution (it may be enough one bit flipping in memory to make this happen) or the driver is forced to cancel the current work the condition on
successful execution does not get fulfilled and the miner may deadlock. I will try to come up with something more stable or that is able to recover better from this situation, but I see a pattern that this happens more often when OC is in the game.
When it started - usage jumps to around 14 GB after some time begin to drop and when stabilized - usage is only around 50 MB.
This is happening on amd rig. On nvidia rig with 8 cards - there is not such problem. I had to borrow 8 GB of ram from nvidia rig to put it in amd rig of 5 cards because noticed that when I start miner on amd rig usage jump to almost 8 GB of ram. Now with 16 GB of ram miner uses almost all of it for some period of time.
