Folding Forum

I don't know how to dump a WU in v7 but in v8 you just have to pause folding and then click the trash can to dump. But it's not something that should be done unless it's an exceptional circumstance. If you just want to fiddle with the Pis, then pause the WU and continue it later, unless "later" is a long time.

Do your ETAs usually say you're able to finish before the timeout?

There's probably no point in running BOINC at the same time on those devices. It will make them less likely to complete before the timeout.

Overclocking can easily make a system unstable even if it seems stable. FAH pushes CPUs very hard and a seemingly "stable" overclocked system might not really be as stable as you think. It's not that FAH dislikes overclocking, it's that most stress testing isn't as effective as FAH in discovering problems.

I believe in v7 you can dump a WU by deleting the slot it is in. Then recreate the slot. Pause it before deleting.

Now that the dust of fiddling with overclocking (at 2 GHz) and so on has settled, it very much looks like F@H knows how to choose WU's that can complete in time for the timeout. All 4 Pi's currently run F@H WU's on 3 cpu's and 25% of cpu's goes to Boinc.

Two of the 4 Pi's now run with moderate overclocking of 1800 MHz and appear to be doing fine. It's 20% extra speed. The TPF estimate on the WU at hand did drop after I applied the new settings. I found a few articles with an identical table of different overclocking settings to go by. You need to give the CPU slightly more voltage with an over_voltage-setting. Some push the Pi4 cpu to 2.2 GHz and perhaps test it only a limited amount of time. The notion that F@H is a more severe test of stability than so-called stress tests was mentioned, and it was also observed that for instability to actually occur/show it may take a while. Not all Pi4 CPU are created equal, either it seems. A silicon lottery
Apart from stable running there is also the need to have adequate cooling. The 40x40 mm Noctua quality fans aim at keeping the CPU's temperature at 60 degrees Celsius maximum controlled by a single transistor on a GPIO pin of the Pi and a python PWM control script. They're quiet at 95% maximum speed and built to last.

Thanks all for helping with advice! I hadn't seen this forum before.

P.S. Estimated PPD now about 12750. I never looked at these numbers before and for some it may be ridiculously low, but it is contributing.