Got this from a connection at NVIDIA, guess the old days of owning Bjorn3d.com helps out a little.
 
 
"Hey Scott,
 
The team has been working on this around the clock to unblock the folding community and we believe the root cause is a bug in the Folding@home dependent OpenMM software.  We’ve engaged with them to see if they agree and get changes done by 12/14.   
 
In case we can’t get those changes done in the app soon, we’re working on a fall-back option in our JIT compiler, based on an app profile to disable an optimization that is exposing the synchronization bug in the app.  We’re targeting to get that driver also posted publically on 12/14.
 
I’ll let you know if either of the above doesn’t happen for some reason and the new ETA.
 
Thanks for your patience!
 
Regards,
Team NV"
 
12/14 Update
 
Hey Scott,
 
Our efforts to get a resolution out by 12/14 has run into some roadblocks.  The change suggestion we made for OpenMM needs more tweaks and our driver side workaround is close to being done but not complete yet.  We plan to post a hotfix driver to unblock the Folding@Home users as soon as we have a workaround to share (within the next few days). 
 
We’ll keep you posted.
 
Regards,
Team NV

awesome!

Whenever I see a line like "we’re working on a fall-back option in our JIT compiler, based on an app profile to disable an optimization that is exposing the synchronization bug in the app" I have to wonder what this means for our production numbers?
 
**Edit - With a little coffee and thought, I realize they're saying it's NVIDIA's code optimization that forced the bug to come out of hiding, so disabling it means they'll go back to the code that is currently working, therefore NOT giving us worse numbers. But, if that optimization is allowed to go through properly, we could see BETTER numbers. So, let's hope OpenMM fixes it!**
 
I really do hope that they'll be able to get things straightened out properly, obviously, but am ready to keep the current driver installer on hand just in case.
 
d

Oh yeah I keep copies of all working drivers.
 

I will believe it when I see it.
https://folding.stanford.edu/home/core-21-openmm-gpu-issues-with-new-windows-nvidia-drivers-375-57-and-375-63/ 
Still Points to NVIDIA and Not to (OpenMM) or Their Coding.

Chris21010
that update is 2 weeks old and may not still be relevant. either way as long as we can get something working with relatively the same performance and allow people to play the newer titles will be a win/win no mater who is at fault.

Yes two Weeks and Still No New info from Folding @ Home

Stanford has never once accepted fault in their coding. I've seen work units fail seconds in, and their software says I must have been overclocking. Never-ending that 90% of the other tasks work just fine. Frankly, this is an open source problem in general. No one wants to admit that their code may broken. I'm just glad there is a fix available from Nvidia, it'll be months before Stanford fixes OpenMM. And even when they do fix it, the folding folks would still have to update the core to use a newer version. It's a big mess all around.

drougnor
Tomorrow's the day, either direction. Cross those fingers, folks!
 
d

Core 22? or New Driver?

With luck it'll be a 'simple' (depending on where in the code the bug is and what is needed to straighten it out) update to the core that will auto update without us even knowing.  That's the kind of fix PG tends to prefer - No fuss, no muss.
 
d

scott@bjorn3dHey Scott,
 
Our efforts to get a resolution out by 12/14 has run into some roadblocks.  The change suggestion we made for OpenMM needs more tweaks and our driver side workaround is close to being done but not complete yet.  We plan to post a hotfix driver to unblock the Folding@Home users as soon as we have a workaround to share (within the next few days). 
 
We’ll keep you posted.
 
Regards,
Team NV

 

The good news is the CEO of NVIDIA is being CC on the progress. 

Anyone tried the latest driver?

Workarounds are good for a temporary solution while waiting for a permanent fix. Nvidia is obviously quicker off the mark and can always remove the hack once the root cause gets patched.