It'll be a local BOINC client decision. The clue is in the "Not requesting ..." - that's a client decision.

The project server isn't involved, so you probably asked for 101 tasks at a moment when you had 999 runnable, or something like that.

The question is: why? The purpose would be to stop BOINC wasting computer power on housekeeping, when you could be doing science. And it helps the scientists get their research back in a timely fashion.

There is a hard limit in the client code of 1000 tasks per host.

work_fetch.cpp module

https://github.com/BOINC/boinc/blob/5dead4107b4c379769b65174136ad2d709b56248/client/work_fetch.cpp#L664-L665

---

There is a hard limit in the client code of 1000 tasks per host.

work_fetch.cpp module

https://github.com/BOINC/boinc/blob/5dead4107b4c379769b65174136ad2d709b56248/client/work_fetch.cpp#L664-L665

Presumably, not a horrendous task for someone who rolls their own to increase that?

Presumably, not a horrendous task for someone who rolls their own to increase that?

Make sure you read the full context of that change. That makes it even easier!

- Make max runnable jobs (1000) into a #define

(from https://github.com/BOINC/boinc/commit/40f0cb44f4fcd11eb2789408dfc868de63e42242)

https://github.com/BOINC/boinc/blob/master/client/client_state.h#L613

There is a hard limit in the client code of 1000 tasks per host.

work_fetch.cpp module

https://github.com/BOINC/boinc/blob/5dead4107b4c379769b65174136ad2d709b56248/client/work_fetch.cpp#L664-L665

Presumably, not a horrendous task for someone who rolls their own to increase that?

Absolutely, back when we were doing Seti, we used to just pop in an extra zero for the limit since we were compiling our own custom client. That made the limit 10K since we could easily crunch more than that in a day with the custom app.

---

Yes, MW still has its own server hard limit of 900 tasks per host. Can't get around that on the client side.

But if your host is running 4-7 gpus on multiple projects, you really need to get around the default client limit if your gpu tasks in general are crunched in under a minute for all projects.

---