Similar results here - 2 Win7 machines, each got zombies from time to time. In fact, each got a cluster of them at different times - haven't seen that behavior in a while. Both machines are fine with 3.8.4 at the moment.
Thanks. Yes that's the typical windows fail I've been trying to work around: When there's some kind of usb communication lag it affects everything on the bus. To be clear, when you get zombies, do they eventually re-hotplug for you? They do not for JMC
Yes that's correct with the later versions (zombies stay as zombies), although sorry to labour the point, but they usually do re-plug automatically with 3.5.1 and earlier. Usually but not always! I do get the occasional permanent failure with 3.5.1, but it is quite rare. Also in 3.5.1, very occasionally an AMU just falls behind and/or just stops handling traffic (Accepted shares stops increasing and/or WU slowly decreases). Just to clarify
We still seem to have differing behaviors. I just got one repluggable zombie in my 3.8.4 run, about 26.5 hours in.
Edit: just got another, also repluggable, a few minutes later.
Edit: and another, about 30 hours in. To summarize, I've had a few repluggable zombies and one cluster of several at once that appeared to require replugging the whole hub. My sense is that 3.8.4 is not reallocating any on-the-fly. My allocations now are AMU15-21, 23-25 and 28-30. I'm pretty sure the reallocation 15-n happened when I replugged the hubs (daisy-chained) and that the few reallocations since then have come from my replugging individual zombies, rather than cgminer detecting and reallocating on its own the way it did in other recent releases.
Edit: Something a bit different. At about 32 hours, the LED came on for AMU29 and I watched its hash rate dwindle to zero on the display. It was not flagged as a zombie though. I waited a few minutes then unplugged it, which caused it to be flagged as AMU0 and labelled a zombie. On replugging, it became AMU31 and the hash rate climbed back to normal. The run continues.
Edit: I eventually had to reboot for unrelated reasons. The followup run gets occasional zombies, all repluggable so far. Recently it had one unit flagged as a zombie and at the same time another with zero hashrate but not flagged. LEDs were on for both and both were repluggable.