I have several devices running BOINC 24/7, and I want to make sure I am maximizing their productivity and efficiency by having each machine run as much as it can.

What do you have your resource share setting set to?

Hi there,
I've read (and confirmed) that BOINC CPU tasks run at Windows' lowest process priority (I can't speak for Unix/Mac/Android). If you want to surf, change documents, or any other normal task (which all have normal/average priority), the BOINC processes can only use spare processing power after other processes have taken what they can. So BOINC comes last in the queue when it comes to processing power - 100% means 100% OF WHAT'S LEFT OVER !
(BOINC GPU tasks need a CPU-thread with which to communicate, and this gets next-lowest priority so that other BOINC tasks don't get in the way of this communication).

But there are other limitations. Some BOINC processes use a lot of RAM, and Windows begins to panic if it gets to more than about 80% total usage of available RAM. Everything suffers here. So it's important to choose projects that fit comfortably within your available RAM.
A few (very few) BOINC projects transfer large amounts of data. If you're on regular broadband this isn't a problem. If however your data-traffic is restricted, you need to keep an eye on your data-traffic usage and avoid projects that could cost too much traffic.
And lastly a strong word of warning.
Desktop computers are generally designed to perform continuously at their maximum. But only if air-ways are kept clear.
This is not necessarily true of tablets, smartphones, laptops - even if you keep air-ways clear. BOINC can damage devices which are not designed to run continuously at 100%, so check your manufacturer's limits and find a suitable app to monitor these limits. There are many other posts in the forum if it's necessary to restrict your devices.
I hope you will succeed in contributing as much as possible to these worthwhile goals

JohnMD is describing how much of your computer you want to dedicate to BOINC running on that computer. I suspect what you are asking is the per project setting of Resource Share. If so, see the "List of project preferences" section of https://boinc.berkeley.edu/wiki/Preferences:



Let's say, you have 5 projects you are running on a few hostsm not all of them on all hosts, and you gave them the per project resource share values of 10,20,30,40,50. This means for every 5 units of time/resource project5 will get, project1 will get only 1 units of the same, if a host is only connected to project1 and project5. In other words, 10+50=60, so project1 will get 10/60=17% of the resources on that host that you allowed BOINC to use in BOINC settings either via BAM! or on the Settings dialog on your host, and project5 will get 50/60=83%. Note that it is NOT the case that project1 will get 10% and project5 will get 50%, leaving the host 40% idle. And, if you have a host that runs all 5 projects, then 10+20+30+40+50=150, so the same project1 running on that host will get 10/150=7% of resources, and project5 will get 50/150=33%, with the rest of the projects sharing the rest of the resources given the math that applies to them.

Yes, this needs a bit of highschool math...

Tuna

And, what project you give the highest share/weight/priority, or how you rank your projects for the different levels of attention they should get from your hosts, is totally up to you. It usually comes down to which projects you like. And, that is sometimes determined by how you feel about their mission/goal, and sometimes depends on how much credit they give per unit work.

Tuna

Good

I experimented with these values and discovered I like 0 for every project and every host. I control the overall task distribution by allowing or disallowing New Tasks as I see fit on each different host. Heat control is the standard for every host also with different apps/settings - as was mentioned below that is very important - especially when adding new projects (like SIDock, SRBase, PrimeGrid) that seem to run especially warm. I have ballooned several droid batteries (they didn't pop fortunately) as a testament.

I did start off with weighting the projects and doing the math - the problem is the variances between the projects (how many WU's they send for instance), the differences in hosts CPU/GPU properties not to mention OS overhead make weighing too dynamic to put a formula to - seems easier to let the chips fall more or less "naturally" and its kind of an experiment to see how the different projects behave (which adds dynamic all its' own).

Might be dumb but good enough for double-bogey!