There is a new core 22 version for GPUs under beta testing at present that is promising. Unfortunately didn't work so well for me at first but bugs are what you try and fix in beta.

SAK917 wrote:Is there any write-up or discussion anywhere about the new core kiore? What is different?

It is a new version of core 22 adding some features:

Now built on the most up-to-date release of OpenMM: OpenMM 7.4.2
Average performance (in ns/day) is now printed out to the client log at the end of the WU.
JSON viewer frames are now written every 1% by default, and project managers can now flexibly control the viewer frame interval for each project.
Checkpoints are now written every 5% by default, and project managers can now flexibly control the checkpoint interval independent of the XTC frame write interval for each project.
Checkpoint, XTC, and viewer frame intervals are now written to client log
Adds experimental support for Intel GPUs that we'll be testing soon
Slight performance improvement on some GPUs by enabling separate PME streams by default

It has had some teething problems so has been delayed in beta for a couple of fixes.

Have also been beta testing some new GPU units involved in the COVID-19 'Moonshot" including 1 suitable for older slower cards.

https://foldingforum.org/viewtopic.php? ... ad#p337997

New COVID-19 reserach units including one designed specifically for low power gpus.

DKeane wrote:Daughter will be getting my old rig (minus the newly purchased 2060 Super). I bought a Ryzen 3900X (on sale on Amazon), which initial estimates show 300K ppd. Not bad for a CPU. Runs a bit hot at 80C on a 240mm AIO setup.

These are extremely powerful cpus with a ferocious reputation for needing serious cooling, a 240mm radiator and still 80c! You can see why AMD didn't bother including a stock cooler in the packet.
These new GPU (COVID moonshot) units are frequently requiring more than a single cpu thread to run properly, how have you allocated the cores for folding?

DKeane wrote:CPU folding is set at 22 cores, with 2 left over for the GPU. Task manager shows CPU living at 98%.

The 3900X actually had a stock cooler - but I didn't use it. I think the 3950X (16 cores) doesn't include a stock air cooler.

Ah yes the 3900X is the 'lite' version with only 12/24 threads/cores, that sounds a good mix with the RTX 2060 super and really amazing computing power, I am still impressed at the capacity my low power 3700X 8/16 threads/cores has with a tdp of only 60 watts.

I wanted the 3900x but this case also holds my RTX 2080ti and heat is a serious issue so wanted the coolest powerful CPU from the series. I bought a hybrid cooler kit for the RTX and have been intending to install for quite a while now but always find something else to do instead especially since the pandemic leaves me very little free time now..

News.
Hardware wise; NVIDIA has launched their RTX 3000 series cards, first the 3080 and now the 3090 both these and the soon to follow 3070 should have a big impact on the project. Currently shortages of GPUs and issues with bios, drivers from NVIDIA and optimization from F@H means that accurate data on these new cards is hard to find and there appear to be serious teething problems as can be expected with new generations. For AMD fans a new NAVI GPU also not far away with some extraordinary claims including a TDP even bigger than the NVIDIA figures.
I am looking at getting an RTX 3xxx, but think I will wait for software issues to resolve and when a 'ti' or 'super' version comes out with hybrid cooling version after my bad experience with the RTX 2080ti I am reluctant to buy another solely air cooled high performance card and the TDP for these cards are already high.

Software wise: New optimizations for F@H including new CUDA/openCL support for GPUs is showing promise of significant increase in utilization early indication suggest in the region of 20% more output for some cards especially in Linux but also improvements in Win, the timeline for this is 'soonish'. I think with new drivers and perhaps a bios update the current issues with the NVIDIA 3000 series will be resolved.

I am not the team founder, that was RPizzle who still pops in every now and then so that credit his. But yes I had assumed with all these big corporate teams coming onboard this year we would be wiped out of the top 100 teams, but it seems we also got a large number of new folders too some with very powerful hardware, I have no idea who AJM is but wow some serious output.
Currently we have 8 team members producing over 2Mppd which is incredible.
I suspect many of the top 20 team members are using GPUs and the good news is that GPUs are about to become more productive with software upgrades without even the expected new NVIDIA and AMD hardware operating. Core 22 has seen major improvements and the newest version which has been in beta testing and seems to also be seeping out to the general population has been creating a stir. I am not sure that the improvement in ppd that have been seen are going to stay this way but new benchmarks are being set up and new was of doing things especially on the high end gpus are being looked at. Currently if you have a high end GPU and you are allocated a low atom count unit the hardware is underutilized and GPUs such as the RTX 2080ti see ppd drop from close to 3Mppd to less than half that amount. Two approaches seem to have been started, one the traditional approach is to limit small units to lower end or older cards, the second which I have yet to see working officially would be to assign multiple small workunits to the bigger cards. People have tried this manually and it seems to work but requires a large amount of user input, stopping and starting units. I have no insider information but from my perspective it seems the project really wants to maximize the impressive hardware available and is innovating rapidly.

DKeane wrote:My estimated PPD just popped. Was around 1,400,000 and it is now estimating 2,000,000 - most of this due to the 2060 Super.

Indeed, big increases for GPUs and still some teething problems still I think with the new CUDA support. I am seeing 20-30% increase in production with minimal increase to temps.
Here's the release article:

https://foldingathome.org/2020/09/28/fo ... a-support/

I have used MSI Afterburner to power limit both my RTX 2080ti and my GTX 1660ti which are aircooled, dropping the power limit and therefore temperature has a very small impact on PPD but improves stability.
I set both these cards for 80% power limit which slows them down when the card temperature reaches 78C.
I installed the MSI Afterburner for overclocking but OCing not really such a thing anymore but works perfectly for power limiting.

Note if you have an NVIDIA card and the CUDA is not being detected or not working:
1) Use GPU-z to check CUDA ticked.
or
2) If you have a CUDA dev kit installed try removing that and run default.