I wonder if it would be feasible to create some kind of exhaust vent for some of your warmer running systems. Just off the top of my head, I am thinking about those standalone air conditioner units that sit on the floor with rollers, then have a tube to exhaust out the window. Position the computer near a window, run a large flexible tube from the exhaust fan(s) to a window, and have the tube exit out at the window. I guess if you have your windows open all day that probably wouldn't help much, but if you run air conditioner it might help. Just thinking out loud...

With all the different exhaust vents on computers though, not sure this would work or how efficient it would be. But if it helped control the heat inside the room, that system and all the other systems may drop a little in operating temperatures. Also keeping the curtains pulled when you are not at home can help keep some of the radiant heat from the sun from building up.

"I have been thinking about a 360mm closed loop water cooler, but I am not so sure they are durable/reliable enough for continuous use 24/7."

My Ryzen 7 2700 system uses a 240mm closed loop liquid cooler. Running 100% cpu usage and it is currently at 36 degrees C (room is 72 degrees F). Crunches 24/7.

Thanks for sharing your experiences with AIO coolers!
Granted they have been developing greater quality for the pumps and they're are now even alternatives to Asetek.

My story is I had a Corsair H100i 240mm cooler and it worked fine about 6 months, then I got significantly higher temperatures suddenly. Reaseating the coldplate to the Cpu with new paste didn't help. Max fan speeds (push-pull config) didn't help either. I got it replaced by Corsair. The new cooler lasted 3 months before the same symptoms reappeared. (100% load 24/7). For it replaced again but sold the new unit while it was still wrapped in plastic wrap.

Ever since, I haven't tried or really regained trust for water cooling AIOs.

That's a bit misleading. From the link, the GFLOPS/core seems reasonably plausible, although I would expect 2700X to score higher per core than the Threadripper.

The higher score per computer for 1950X vs the 2950X seems to be because the people running the 1950X are using more cores on average - 29 cores (out of 32) for the 1950X vs 19 cores for the 2950X. It doesn't say much about the CPU, but rather about how people are using it. E.g. I frequently run WCG on my 2950X at 8, 12 or 16 cores, because I have other stuff running too. So the only thing the "per computer" score tells us is that a larger proportion of the 1950X's are dedicated crunchers, while many of the 2950X are used for other things as well.

For the 2990WX, the number of cores (216) and GFLOPS per computer is just plain wrong, must be a bug somewhere. The 2990WX has twice the number of cores and slightly lower per-core performance than the 1950X/2950X, so we would expect a bit less than twice the performance per computer.

But the per-core GFLOPS appears plausible for the 2990WX too. If you're using these data to decide on which CPU to get, go with the perf-per-core and multiply by the number of cores.

There is a thread that discusses this issue in the MIP section of the forum:
https://www.worldcommunitygrid.org/forums/wcg...d_thread,40374_offset,100

Evidently, Rosetta has a Level 3 Cache problem when many instances of MIP are run on the same processor. If you want full credit while running MIP, lower the amount of MIP instances.

This is the equation I use to determine how many instances of MIP I should run on a given processor:
Max_core_count_MIP = floor( cpu_L3_cache_size_MB / 5 );

the 2990wx has 64 MB of L3 Cache therefore --->
Max_core_count_MIP = floor (64/5) = 12

the 2950wx has 32MB of L3 Cache
Max_core_count_MIP = floor(32/5) = 6

So only run 12 instances of MIP on the 2990wx, and only run 6 instances on the 2950x