krisztian.erdelyi, Welcome to the forums.

Boinc manager does have a feature that will get you close to what you are asking for. In Device Profiles, Basic Options you can set the time frame in which you want to crunch, and the % of processor usage. This will give you two separate timeframe specifications. The only problem is that one of them is 0% participation, not 20%.

Or have you considered under Advanced Options setting "Do work only after computer is idle for: xx minutes." This works.

Or consider this. Boinc only crunches the spare CPU cycles so there should be no noticeable affect on your other computing. GPU computing can be an exception to this.

Good luck with your crunching.

More clearly, in design ALL sciences run in the 'yield to everything' lowest priority mode [Nicest in nix world], which means when -you- need the power, the BOINC science apps make way, with an ever so slight latency that any normal person wont ever notice. **

The [WCG] recommended setting in the web or local preferences is "Suspend when non-BOINC usage is above" is 50% [and of cource on the client 'run according preferences'], which means if you really need that much power to run whatever non-BOINC, then the client will pause all tasks for 10 seconds. In 10 seconds it then checks again in 10 seconds if it is still the case, and continues to pause for another 10 seconds, or resumes. In my case I do this not so much that I notice, but more that WCG's Clean Energy then starts crashing on my system because file write then suffer too much delay, causing time-out fails.

Such granularity, either percentages while in use or when idle, has been asked from the developers often times [Berkeley, not WCG] and it's always been 'we can not be convinced this adds anything of value', just a whole lot of scheduling code that can break things.

The upside is, much less cycles get wasted by letting the system and OS manage this in stead of the user micromanaging and see that 'System Idle process" clocking up hour after hour (Look in WinOS task manager for how much CPU time goes to waste).

If you've been busy, you can actually see in the client for each task [under properties] how much time on the clock the task was allowed to run [Elapsed], and how much it was actually computing [CPU time]. The greater the gap between the 2, the more you [or a rogue virus] have been using.

Typically at the end of the day, the nightly 'efficiency' on my systems is greater than 99% (your plan was to crunch 24/24 I understood), and when using, it can drop to 95 - 90 or even 80, but that's for single tasks, not all tasks, as the OS takes care that the always higher priority user applications get all the time on all the threads the user app needs, leaving other threads to crunch on science and use any spare cycle [99.9% or better], automagically. Summa summarum, my 'used' 8 core typically turns in 7.5 CPU days per day of crunch results, and I don't notice it's running.

** Some notice when gaming or heavy video streaming, so there is the <exclusive_app> config setting to cause BOINC to pause when the system is used for these specific activities. Load them, BOINC pauses, end their use [exit], and BOINC resumes, without having to remember.

Using some tools like Tthrottle with lower temp., might do just that...lower CPU usage during real working hours, while more 80-100% throttling of CPU science when not working on computer...though you'd have to play more with max temps! ;)