Tried dropping the clock from 160 to 150MHz last night, made no difference.
I'm trying a selection of cables now to see if this helps.
I have noticed that when one or more boards goes SICK / DEAD that it is impossible to close cgminer (even "end process" fails) and a reboot is required. Also the Lancelot never goes SICK.
The clock speed shouldn't have any effect on the USB comms. The failure to "end process" looks very much like a driver issue.
Looking at the schematics, the Lancelot uses a FT232RL USB chip, while the Icarus uses the PL2303HXD, so there must have been a reason for the change (the PL2303HXD is a cheaper device according to one post I found via google). Ah, found
this where ngzhang cites the bad driver as a reason for the change to FT232. Anyway you should check which version of the driver you are using. My google-fu has failed to find anything much of use, though
this mentions driver version 1.50 or 1.52.
If I find anything else I'll get back to you...
Another
post with a link to the
driverThis
post and scroll down for a few more.
It seems the Avalon-2 uses this chip too
(link)And (not relevant to your problem) kano had the
same issue as I did with rpi network stack locking up! Seems the fix is just to update debian (I've not bothered so far, for the same reasons, little free disk space and a reluctance to meddle with something that (mostly) works fine)
Anyway, if you've got a spare linux box, it may be worth running the icarus on that as the gist seems to be that the linux drivers are more stable than the windows one.