As of today, your folding GPUs just got a big powerup! Thanks to NVIDIA engineers, our Folding@home GPU cores—based on the open source OpenMM toolkit—are now CUDA-enabled, allowing you to run GPU projects significantly faster. Typical GPUs will see 15-30% speedups on most Folding@home projects, drastically increasing both science throughput and points per day (PPD) these GPUs will generate.

 
Full announcement, details, and performance charts can be seen at https://foldingathome.org/2020/09/28/foldingathome-gets-cuda-support/
 
This is a public release. Nothing needs to be done as long as you are running recent or current NVIDIA drivers. If you have an NVIDIA GPU then the FAHClient will automatically download an updated Core 22 when you receive a CUDA-enabled project. It's worth noting that if you are using a computer dedicated entirely to NVIDIA GPU folding then CUDA performance under Linux is significantly higher than under Windows 10. 
 
Seeing one person's 1080 Ti getting up to 2.5m PPD already under Windows, which is insane given that card would be lucky to get 1.2m PPD a year ago. It will likely take a few months before CUDA projects replace most of the standard OpenCL projects, so the higher performance will take time to stabilize.

Wow this is great news. May be worth checking the "B" Stock for older cards to fold. 
Thanks a bunch kougar.

Thanks for the post Kougar, good info! 

That should give our team a kick in the pants thanks for the info kougar.

I had been running the beta WUs and they had been giving my 2080S a 33% uplift and the 2080ti a 38% uplift in Linux

Yep they have been growing since I posted about Cuda beta cores.  Great they are now going public.

turbomadman
Wow this is great news. May be worth checking the "B" Stock for older cards to fold. 
Thanks a bunch kougar.

Sounds like a plan :D

I will never run the beta flag it takes WU away from the testers and we do need them tested.

Aye, the beta flag is for people willing to closely monitor hardware and invest serious time into tracking down issues. I was very surprised F@H progressed to a public release so fast.
 
As a side note, the new CUDA projects are utilizing more of the core hardware, people are reporting higher power draw/temps. So if anyone experiences instability I'd recommend some stress testing with OCCT. It's free and includes a built-in artifact scanner: https://www.ocbase.com/

You can also modify your power limits as well. Which I suggest during pure folding anyway. My 2080ti is down to 180W from standard limit of 260W. 2080S down to 160W from 270W. Very marginal reduction in PPD for me so efficiency is much better now