Hi,

i am running many WUs Orbital fitting and not, version 1.90 , and since i use a 28 cores (56 threads) when it's the turn of Milkway, not even after 3/4 mins of computation my entire system FREEZES, i got plenty of RAM and i run LHC@Home and Einstein no problem, Only Milkway does this for me.

I tried it on Linux Debian 12, and on Windows 11. Same errors, Same problems.

As soon you start 3 WUs each one of 16 cores, everything freezes, and is not responsive at all.As far as i can understand, Milkway basically takes up all the available memory in the system, and never cleans it up, so is like it takes 1gb, then 2gb, then 3, until memory is completely full.

It makes also other WUs crash (due to memory).

This problem was never fixed, same as the one that you need to run everything at 100% or Milkway WUs stay stuck at a fixed percentage.

Please Fix it :D

As soon you start 3 WUs each one of 16 cores, everything freezes, and is not responsive at all.As far as i can understand, Milkway basically takes up all the available memory in the system, and never cleans it up, so is like it takes 1gb, then 2gb, then 3, until memory is completely full.

You are pretty much the first one reporting that issue and apparently all your other computers complete Milkyway WUs without any issues, except few errors that all of us have.

In fact, even that system running Linux has a RAC of over 11,000, only two errors caused by bad WUs and a total BOINC credit of 1.6M.

So it's just when running Windows 11 and according to BOINCstats, so far this computer does not have any credit on Windows, it basically does not exist in the list of your computers.

---

The WUs don't even compare in the statistics?.

Do you want some screenshots of "computation error" , or a video recording that when Milkway WUs start, every computer freezes and i need to reboot?. Literally, every PC i use do this, Windows/Linux.

The fact that it doesn't even show in the logs/statistics is even more serious, because that now that i think about it, if you use preferrences from milkway website (not boinc) and select "execute maximum 1 task of Milkway it goes completely ignored, and the system still will compute every milkway WU available.

Also, tried to use 50% Cpu, same error, 10%, Same error. Deleted WUs (and you don't see them deleted?) SAME ERRORS.

i am stopping the project to be honest.

About people not having the same problems, about WU stalling, here some old posts.

https://milkyway.cs.rpi.edu/milkyway//forum_thread.php?id=4584#69615

https://boinc.berkeley.edu/forum_thread.php?id=15116#112892

That's why you need to use 100%, and, when you use it, everything just straight up freezes. Same if i use less cpu, or less time.

It "works" if the WU is a single one, a 4CPU one, but 16, or more, it straight up don't work.

Do you want some screenshots of "computation error" , or a video recording that when Milkway WUs start, every computer freezes and i need to reboot?. Literally, every PC i use do this, Windows/Linux.

Whatever helps to tell what's happening on your computers. Until now the only thing anyone can see, is all your computers except the Xeon 8173M running Windows crunching successfully Milkyway WUs.

About people not having the same problems, about WU stalling, here some old posts.

This is a known issue, but this does not cause excessive RAM usage by the application or computer freezes, the application just stops crunching.

That's why you need to use 100%, and, when you use it, everything just straight up freezes. Same if i use less cpu, or less time.

Well, if everything freezes even when you use less CPUs or less CPU time, than this bug forcing you to use 100% of CPU time obviously doesn't really matter in your case.

It "works" if the WU is a single one, a 4CPU one, but 16, or more, it straight up don't work.

This might be true for the Xeon when it runs Windows, but apparently not when it runs Linux, it completed hundreds of tasks with 16 threads/task. Also this does not add up with "Literally, every PC i use do this, Windows/Linux" as no other system you have has more than 4 or 8 threads. Running single thread is most efficient anyway, so if that works, why run the less efficient 16 threads/task?

Long story short, you need to provide more and correct information if anyone should help you.

---

bonjour
J ai le meme probleme sous linux mint avec i9 14900 et 96 gigas de ram.
J ai du forcer mon pc en extinction car il ne repondait plus en calculant une unité avec 16 threads.
J ai le probleme qu 'avec milkyway et pas les autres projets.
Alors J'ai viré milkyway de mes projets tant que le probleme ne sera pas réglé.


hello
I have the same problem under Linux Mint with i9 14900 and 96 gigas of ram.
I had to force my PC to shut down because it no longer responded by calculating a unit with 16 threads.
I have the problem only with Milkyway and not the other projects.
So I removed milkyway from my projects as long as the problem is not resolved.

---

Pascal,

I am not an Intel user, but I have seen threads on forums in other places that talk about various issues running BOINC projects on Intel systems with p-cores and e-cores, including tasks stalling; the usual recommendation in those threads seems to be that if the system is primarily used for crunching, disable the e-cores!

If that is a "regular" problem, it would be interesting to see what happens if you try using (say) 3 or 4 cores per task (and don't run enough tasks to take up all the p-cores!).

Just passing this on, recognizing that it might not actually help...

Cheers - Al.

If that is a "regular" problem, it would be interesting to see what happens if you try using (say) 3 or 4 cores per task (and don't run enough tasks to take up all the p-cores!).

I'd recommend trying single core tasks, current CPUs will complete the WUs in resonable amount of time anyway. But yes, I've read too, that Intel's P- and E- cores can be a major PITA for BOINC (and other similar loads), one of the many reasons why I bought AMD CPU for my current PC.

---

As mentioned earlier in this thread, it is more efficient and more productive to only run 1T tasks and just run more tasks on more cores. I've never had any memory issues on any of my hosts.

---