The project is growing, and before more people join we want to check that everything holds up under heavy load: task creation, downloads, result uploads, validation and the live dashboard. To do that, we need a lot of computers running at the same time, for a few hours.

We're looking for team members and team captains who would be willing to bring their team along for a short stress test. No date is set yet: we'll pick one together, at a time that suits the teams who volunteer.

If your team could take part, or you'd like to help organise it, please reply in this thread (or send me a private message) with:

your team name;
roughly how many computers or cores you could bring;
the days and times that would suit you.
All tasks during the test are real work and are credited normally. Thanks a lot, it really helps us get ready for the next step!

Team Borg can bring 344 Zen5 cores if someone suggests a testing period (date range).

It would be good between PG challenge ends and FB begins, 7-8th

Thank you very much for your help.
Could you start at 9 p.m. and continue until tomorrow?

If you wanted a real test, I would contact the admins for Formula Boinc or Boinc Games and schedule a very small competition in between one of their normal challenges.

already done, just waiting...

I am moving my CPUs back over to here now that PG challenge is complete.

Welcome back, and thanks for bringing your CPUs ! So far everything is holding up well: plenty of work in reserve and no errors.

Allow more tasks to download than 2x # of threads if you actually want to stress the server. I had a PC running other work than this project for some reason. It only takes 1 hour of not receiving a task for it to move to other projects.

Thanks for the feedback, it's a fair point. On our side the server never ran dry during the test: the queue stayed above 20,000 ready tasks the whole time. So what you saw comes from the per-host limit. We currently allow 2 tasks in progress per thread, and with tasks of about an hour that is only a small buffer: if your client can't reach us for a while, or decides to fetch from another project first, it switches quickly. Now that tasks checkpoint and resume properly, we're looking at raising that limit, and your experience helps us decide.

My 9950x is closer to 40min/task. Of the 32 threads, are are running memory yoyo tasks and and it still currently has 7 tasks ready to start. It'll be out of work again in a few minutes.

You're right, and your numbers make it clear: at about 40 minutes per task, 2 tasks per thread gives your 9950X barely an hour and a half of buffer. The server has plenty of work, more than 26,000 tasks ready, so the bottleneck was our per-host limit. We've just raised it: up to 4 tasks in progress per thread, and up to 8 tasks per scheduler contact instead of 2. Your clients should fill up on their next contacts; clicking "Update" for Eternity II speeds it up. Thanks for pushing on this!

In reply to benjamin's message of 9 Oct 2026:

We've just raised it: up to 4 tasks in progress per thread,

I am not seeing this change on my machines. Still stuck at 2 tasks per thread.

I am seeing the same as zombie. The task count per PC is no more than 2x the # of threads. Some have around 1.5x which is well within my 0.25 day queue request.

You're both right, sorry about that: the announcement went out before the change was actually applied on the server. It's live now (since about 20:57 UTC): up to 4 tasks in progress per thread and up to 8 per scheduler contact. Clicking "Update" for Eternity II should fill your queues right away. Thanks for checking!

Thanks again to both of you for testing and reporting. The limit is now 4 tasks in progress per thread, and that's only a first step.

On Sunday evening we'll raise it much further, to between 35 and 40 tasks per thread. With tasks of about 40 minutes to an hour, that's roughly one day of work per thread on average. Your machines will be able to fill a normal BOINC queue and keep crunching through a server or network interruption, or while your client works on other projects.

Tasks keep their 7-day deadline, so a one-day queue leaves plenty of margin.

Awesome, thank you. I have 90+ tasks ready to start

Ok, on Sunday I will enlarge my queues to grab a huge number of tasks to help with the stress test. I assume Sunday your time, please give a heads up on when UTC.

Additionally, I have two computers that due to the resource demands of your tasks interfere with the users work. I can download task only to abort them on those machines if that sort of activity would help in the stress testing.

Thanks Werinbert, that's very kind!

Here's the plan for Sunday. First I'll rank the best candidates from the month-long screening that ends that day, then launch a large campaign of new tasks for them. That part goes quickly. But raising the limit to 35–40 tasks per thread will cause a big wave of requests, because every host will want to fill its queue at once. So I'll release it on Sunday evening if everything checks out, or on Monday morning (European time) at the latest. I'll keep you posted here as soon as it's live.

About your two other computers: please don't download tasks just to abort them. Aborted tasks have to be sent out again, so they only add load without testing anything useful. Those machines help most by not taking work from us. If you'd still like them to contribute without disturbing their users, BOINC's computing preferences can limit the number of CPUs used, or suspend computing while the computer is in use. On Windows, our latest version (2.37) also runs at low priority, like other BOINC apps, so the desktop stays much more responsive than before.

I'm stress testing this project almost since the very beginning :)
I saw this thread a bit late, but if there's anything I can help with in the future, feel free to message me in PM or reply here.