Gary480six wrote: Still wish I could get back to the successful core 17 Folding production I had two weeks ago, but at least now I can be of some help to the science.
We all wish that.
Core_17 uses some features which were not needed by Cores 15/16. Given a choice between WUs that do fold and WUs that do not, you're probably stuck with cores 15/16 for a while. There seem to be strong reasons to blame the drivers for not supporting the features they're supposed to support and nobody has identified drivers which work. Even if somebody says I'm successfully using drivers xxx.xx, results seem to depend on which class of GPU you're running, perhaps on which version of the OS you're running, and possibly on other factors.
At this point, there is no reason to believe that we're looking at a bug in the Assignment Server code. Rather, as changes were made to minimize the impact of the real AS bugs, attention was focused on problems that were already present but which had remained "under the radar."
Unfortunately, when dealing with bugs which might be in the drivers, which might be in the FahCore, or which might be in the latest hardware, it takes time to isolate each problem (since there may be several) and test the validity of the fix. If all of the bugs are in the FahCore (which I doubt) then Stanford can fix them. If some are in the drivers or the hardware (which I suspect), all Stanford can do is (A) Wait for NV/ATI to create drivers which bypass the defective sections of the hardware,or (B) Wait for NV/ATI to fix their drivers or (C) Rewrite segments of the FahCore to bypass the errors encountered when the drivers produce the correct results.
Note that "(A)" presumes that issuing a recall for defective hardware is not on the list, even if the automakers have sometimes used that option.