First off, let me clarify that I don't speak tech. I've noticed a bunch of changes lately with MW and it's leaving me scratching my head. I was running an optimized application, have been since I started the project, and recent work units ran quickly, but when the results were uploaded they were invalid. I read somewhere on the site the optimized apps could be causing a problem and the best course of action was to reinstall the stock application. I did that and my work units were being reported correctly. Was I on the right track? I don't run GPU.

My current question concerns MW grabbing all my cores. I have a decent quad core running on this machine. I run several different projects and generally have one project running on each core. Tonight, I downloaded some MW work units (6 nbody and 6 regular)and immediately the nbody units jumped into hyperdrive(running high priority) and took all four CPU cores for one work unit. I'm wondering why? Is this a change to the project or is it an error?

Thanks for taking time to read this.

Regards,
Zen

Zen,

I have read that the opti apps are not working anymore. For now you have to deal with the stock apps.

MW is using multi-core now, this means that it will use all available CPU's to crunch for one WU.

---

Greetings from,
TJ

Actually, only the N-Body tasks are multi-threaded and use more than one core when running.

The MW 0.50 CPU tasks are single threaded and use one core when running, unless BOINC has more than one task queued up for running that is.

Your comments are not exactly true.
I have de_seperation_10_3_fix20 etc. hogging all 4 cores as well and not using them to full capacity. I checked my cpu monitoring program and noticed that 1 core is at 16% with a max around 73%. Another core runs 17-78%, core 3 min 25% max 91% core 4 min 6% max 56%

This is not good usage of my system. There should be a better way to crunch the data than to hog all my cores and not use them at 100% ALL the time.

I've been running BOINC with S@h for a few weeks, and today I've tried out Collatz, PrimeGrid and Milky Way - MW insists in hogging both cores when running, when I expected it to run on 1 core and still allowing other apps to crunch away. I'm to stop processing MW WUs until a time when the MW app is less power hungry!

I've been running BOINC with S@h for a few weeks, and today I've tried out Collatz, PrimeGrid and Milky Way - MW insists in hogging both cores when running, when I expected it to run on 1 core and still allowing other apps to crunch away. I'm to stop processing MW WUs until a time when the MW app is less power hungry!

It should be a more efficient use of your CPUs to use all of them for the same task. The BOINC scheduler should take care of running different projects at different times. They don't need to run at the same time to get a chance to run.

How about one WU taking up all 16 cores (2 physical CPUs)? That's what it's doing to me!

---

How about one WU taking up all 16 cores (2 physical CPUs)? That's what it's doing to me!

That's what happens when it's a multi-threaded app. When it runs it uses all available CPU's.

These WU that call all the cpu's seem to finish pretty quickly. I just watched one scheduled for an hour of crunching finish up in 7+ minutes. The power of 64 bit multithreading I presume.
My general cpu use meter shows me staying between 90 and 95% use for these, less efficient maybe but these WU seem to go so quickly it doesn't worry me.

---

;

I, too, panicked at first when first I saw 8 cores* utilised (hogged) when I only really wanted one core to be used. Then I saw the way the app ripped through the n_body WU so I waited to see what happened over a few days.
The way I understand it, let MW and other projects & BOINC sort everything out based on the way you have set your BOINC processor usage and work cache. If you set processor usage to 100% the n_body WUs will pass through almost unnoticed on e.g., an *i7-2600K at minimum safe overclock (known true). I also run an Intel C2D T5750 @ 2.00ghz, Boinc processor usage:- 25% that runs an n_body in an hour or two so these WUs hardly get in the way of other 30-day+ deadline tasks like SETI etc.
HTH, Ray
(see other recent thread about my comment on initial elapsed time for n_body WUs)
(BOINC/MW etc WU scheduling needs shepherding at times :-P )

---

Oh my goodness. I just saw a multi threaded N body simulation that the BOINC client had estimated at 38 hours crunching finish up in 12 minutes.
I'm sold.. we all need to make the jump to 64 bit and force all the software vendors to support it. Not just for BOINC but everything...what an astounding performance increase.

---

;

There is an issue however. It may only take 16 minutes to finish with 2 cpu's, but it loads another one and it becomes priority and my other projects never run. The time remaining is starting in the 300's and goes down real fast, but this causes priority.

---

There is an issue however. It may only take 16 minutes to finish with 2 cpu's, but it loads another one and it becomes priority and my other projects never run. The time remaining is starting in the 300's and goes down real fast, but this causes priority.

Give it time. BOINC will work it out. The other projects will get their share again. While MW is running the other projects are building 'debt' which BOINC will pay off by giving them time.

You are SOOOooo missing the point. IF you crunch 38 hours of work in 12 minutes, even IF it jams the system and stops all other processing for that 12 minjutes, you've shown a 50,000% net profit on your crunching time. hello?

---

;

You are SOOOooo missing the point. IF you crunch 38 hours of work in 12 minutes, even IF it jams the system and stops all other processing for that 12 minjutes, you've shown a 50,000% net profit on your crunching time. hello?

And you aren't doing any homework and considering how BOINC works. This is not the office of arguments.

bump :)
Here is my computer, I have intel xeon 2 CPU with each 8 cores. My system actually FreeBSD/AMD64 8.2, however since running boinc under wine is faster, I run it under wine (boinc 7.025, the new one 7.028 from Einstein project won't pass the internet connection somehow :( ).
If 1 task MW takes 1 core, it usually finishs in 2-3 hrs. However, if it runs 6-12 cores it finished less than 2minutes!!
The same thing happen with my old mac X 10.6.8. If MW runs with 2 cores, it finishes in less than 3 minutes. But if 1 core takes 2-3 hours.
So, the question is how to force each task to run 6 cores (even only 1 unit active at one time), is it possible ?

There are 2 different MW apps:
- separation is using 1 core
- nbody is using multi core

Actually you are running both types, so you are seeing the different behaviour you describe. You can choose with one you want to run with a setting on the project preferences page.

OK, thx, let me try if it always run using more than 1 core now.

That's what happens when it's a multi-threaded app. When it runs it uses all available CPU's.

I believe there is a hard limit of 16 threads/WU. The multi-threaded WU's never use all of my 24-core machine.