Is there a hard limit (programmed) on the number of simultaneous tasks running on the BOINC Client or on their management by the Manager?

I'm focusing on the BOINC software itself, not any particular hardware limitations. Assume the hardware has sufficient horsepower, RAM, Storage, bandwidth, cooling(!), etc. For example, a dual EPYC 9965 computer has 768 threads. Some massive 8-way Intel boxes have more. Assuming the hardware is up to the job, can BOINC Client/Manager run and manage e.g. 760 actively running tasks (I've save a few threads for the BOINC software itself ;-) ?

Is there a document that lists any hard limits on the BOINC Client/Manager software?

Thanks

---

Is there a hard limit (programmed) on the number of simultaneous tasks running on the BOINC Client or on their management by the Manager?

As far as I know, there isn't a hard limit. I don't know what happens if you run one of the projects which produces several gigabytes of data per task with those numbers though.

In reply to Dave's message of 20 Nov 2025:

As far as I know, there isn't a hard limit. I don't know what happens if you run one of the projects which produces several gigabytes of data per task with those numbers though.

Thanks. There would be numerous potential bottlenecks on a high thread-count box. But I was focusing on the BOINC software. Assume projects running are relatively high compute, low bandwidth, low data-transfer tasks. OTOH, these monstrous machines can have 6 TB of RAM or more.

Is there a document listing BOINC software limitations?

In reply to wprion's message of 20 Nov 2025:

Is there a document listing BOINC software limitations?

There are none cause we're limitless.
https://youtu.be/JNvEyOCQ1f0?si=SlYvp9RhkRiwEhsJ

---

BOINC maintainer.
For any insight, check my BOINC Development Blog.

There are none cause we're limitless.

I really appreciate being able to get your input from inside the project Vitalii. So much better than my mostly but not always informed speculation.

There's the BOINC WIKI which is the official documentation for the software.

If you want there's also the BOINC software source code.

Edit to add some project have a hard limit on tasks that a single computer can receive at any given time for various reasons.

On a fully loaded 16c/32t CPU I measure 0.2-0.3 % CPU load for the BOINC client itself.
100 % would be a full CPU thread.

Hence, in the unlikely case that you identify a single BOINC instance being the limiting bottleneck, consider to run additional BOINC instances on that computer.

On a fully loaded 16c/32t CPU I measure 0.2-0.3 % CPU load for the BOINC client itself.
100 % would be a full CPU thread.

Hence, in the unlikely case that you identify a single BOINC instance being the limiting bottleneck, consider to run additional BOINC instances on that computer.

That's a normal amount of load for the BOINC client.

Remember, the BOINC client is NOT what is going to consume the CPU resources, it's the various project science apps that do the actual work and consume CPU cycles.

There's a hard limit on the number of tasks the Boinc server software can send to a host in the Boinc code. The limit is 1000. So you could theoretically have a host with computing capabilities greater than 1000 WU's and not be able to be fully utilized.

#define WF_MAX_RUNNABLE_JOBS 1000
// don't fetch work from a project if it has this many runnable jobs.

In reply to Keith Myers's message of 29 Nov 2025:

There's a hard limit on the number of tasks the Boinc server software can send to a host in the Boinc code. The limit is 1000. So you could theoretically have a host with computing capabilities greater than 1000 WU's and not be able to be fully utilized.

#define WF_MAX_RUNNABLE_JOBS 1000
// don't fetch work from a project if it has this many runnable jobs.

A Dual socket Sierra Forrest system will have 576 cores/threads, so hitting that limit isn't that far away at all now.

---

Grant
Darwin NT.

In reply to Keith Myers's message of 29 Nov 2025:

There's a hard limit on the number of tasks the Boinc server software can send to a host in the Boinc code. The limit is 1000. So you could theoretically have a host with computing capabilities greater than 1000 WU's and not be able to be fully utilized.

#define WF_MAX_RUNNABLE_JOBS 1000
// don't fetch work from a project if it has this many runnable jobs.

Let's say, it's default hard limit.
But don't forget, this is a limit of number of jobs in progress.
If devicce receives 1000 jobs, then report back 500, so it can acquire 500 more additional jobs again.
So this is not so hard limit as it sounds.
And some projects has created their own mechanisms to assign maxx number of jobs dynamically or by the user's preference.

---

BOINC maintainer.
For any insight, check my BOINC Development Blog.