Hi guys,

this may sound not too intelligent, but:
Is there a way to assign one project to one processor?

I always seem to get too many applications from one project or the other. Just depends on which project is faster. I get two apps and the other project won't request other tasks. If I increase the work buffer I only get more tasks from one project. I don't think it's really weighted.

Look at my statistics in the signature. I'm sharing 50/50 since two months after I started with grid computing. And SETI is even more bound to leave me taskless.
I tried Lattice when I didn't get tasks either from SETI or Rosetta.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Is there a way to assign one project to one processor?

No, there isn't, sorry.

I always seem to get too many applications from one project or the other. Just depends on which project is faster. I get two apps and the other project won't request other tasks. If I increase the work buffer I only get more tasks from one project. I don't think it's really weighted.

It is - over the long term.

Look at my statistics in the signature. I'm sharing 50/50 since two months after I started with grid computing. And SETI is even more bound to leave me taskless.

Okay, 28000/15000 isn't quite 50/50 ;-) but you have to count in pending credits, as those are not counted in your signature. Also, if one project has no jobs available (as happens quite often with SETI lately), the other one gets more, but BOINC counts up the debt and will pay it back later, when enough tasks are available. So, you'll have to be patient. It might take several weeks more to even out.

I tried Lattice when I didn't get tasks either from SETI or Rosetta.

Tough luck :-) they only have work very sporadically.

Depending on your BOINC version, there might be bugs that make it difficult to keep the resource share you've set, especially if you are using a GPU for crunching.

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

No, there isn't, sorry.

;-(

It is - over the long term.

Not in 1.5 years it's not.

Okay, 28000/15000 isn't quite 50/50 ;-)

50/50 in time. ;-)
Member at SETI since November 2008.
Member at Rosetta sind December 2008.
But the uneven credit counting is another problem...

Tough luck :-) they only have work very sporadically.

Yeah. But as you can see I didn't have to use it very often.

Depending on your BOINC version, there might be bugs that make it difficult to keep the resource share you've set, especially if you are using a GPU for crunching.

6.10.18
And no GPU.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

6.10.18
And no GPU.

Then you could try 5.10.45. I'm running it since it came out and have no problems with it ;-)

Gruß,
Gundolf

...and I have Windows. :-D

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Is there a way to assign one project to one processor?

No, there isn't, sorry.

There is a way. Set the "switch between applications" value to something that's higher than the nominal run-time of any task. In seconds, of course.

So for example, if the most time taken by any task is 8 hours 13 minutes, you take 8.5 hours as value and fill in 30600 seconds as switch time. With only two projects having work, this usually makes sure that one CPU/core will be on one project all the time.

---

...and I have Windows. :-D

Same here: XP with BOINC 5.10.45 and NT4 with BOINC 5.8.16 (the last to support NT4).

Gruß,
Gundolf

There is a way. Set the "switch between applications" value to something that's higher than the nominal run-time of any task. In seconds, of course.

I have it set to 99999999999999999000000000000000000000000000000000000000000000000000000000000000000000000.00 since some months.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Are you sure that value is accepted? Illegal values tend to be interpreted as default values (60 seconds in this case).

Gruß,
Gundolf

With the aforementioned value global_prefs_override.xml says:
<cpu_scheduling_period_minutes>99999999999999999000000000000000000000000000000000000000000000000000000000000000000000000.000000</cpu_scheduling_period_minutes>

Too many 9's gives:
<cpu_scheduling_period_minutes>1.#INF00</cpu_scheduling_period_minutes>
...and sets the value in the GUI to 60.00 minutes.

I've set it now to 500. But I'm pessimistic that this will help.

---

My reasons to participate in grid computing:
1. I'm using true renewable energy (German accreditation; no certificate purchasing)
2. Production and disposal (will) amount to about 95% of the energy "used" during the lifetime of my PC
3. Helping

Are you sure that value is accepted? Illegal values tend to be interpreted as default values (60 seconds in this case).

Gruß,
Gundolf

Minutes?

Are you sure that value is accepted? Illegal values tend to be interpreted as default values (60 seconds in this case).

Gruß,
Gundolf

Minutes?

Yep, if I say seconds, I mean minutes ;-)
(have been mentally at another delay:-)

Gruß,
Gundolf

I think axl and I have the same idea. I would like one core(processor) to run on a project. I have a quad and set Einstein and Rosetta 'On multiprocessors, use at most' to 1. BOINC updated and then allowed Climate prediction to run on only one core while Einstein runs on three. Does this preference actually not work or is there some other underlying method to devote a core to a project.

"On multiprocessors, use at most" value is for BOINC 5 only.
"On multiprocessors, use at most" percentage is for BOINC 6 and above.

And then, with that option you tell BOINC how many CPUs or cores it should use at maximum. If you have a quad core and set it to 25% it will use 1 core only. The other 3 cores won't be used by BOINC.

---

I should have edited to say that I did try tinkering with the 5.10 and the 6.0+ settings. When I do set processor usage to 25% it only will run one task between all the projects. I was hoping that these options are tied to the project and not the account.

I was hoping that these options are tied to the project and not the account.

They are used by BOINC, the client, to decide how many applications can use the CPU at once. With it you tell BOINC the amount of CPUs or cores it can use.

So on a quad this is:
25% - 1
50% - 2
75% - 3
100% - 4

When you set it to use only 25% of the cores, only one application will run at a time. The other 3 cores are no longer used by BOINC or any project application, until you change the processor preference.

Also, the value is an integer number, meaning that even if you fill in 47% only 1 core will be used, or when you fill in 79% that only 3 cores will be used.

---

That's exactly what I figured. I was hoping that I could run a concurrent task on four separate projects rather than having the work shift around from project to project.

---

Well no, as what's the use of that? Projects are all individual (but for their alphas and betas). If all were the same, you would only need one project to attach to and be done with it.

So since all projects are individual, they all have their own work. You can't run Seti work for Einstein (or in Einstein's time), nor can you (for instance) show the screen saver of Einstein while CPDN runs.

BOINC brings all those projects together and under your reach to help them out.

---

If I could run a separate core on four project I could see how much work each one could do and track points, moving to the highest paying project. I wouldn't have to suspend a project to force BOINC to run the tasks I want it to run. Resource share is a novel idea but it doesn't really work. I can set a project to 100,000 and one to 100 and it will still do the 100, I've tried. Well at least with MW at 100,000 and collatz at 100, it will always run collatz even though those tasks take longer. This is why I want a core dedicated to each project. Otherwise it's get tasks, complete them all, upload, switch projects and repeat. I can't be the only one that has thought of this or is trying to make a point. Or perhaps it's personal reasons that I would like to split my work.

In the case of the collatz/MW scenario, MW only gives you 10 or so tasks and they have been having some trouble lately keeping their servers up. If I set the share on MW high I would expect it to run MW forever until it can't get anymore tasks and then fall back on collatz.

---

BOINC is a tool for helping out science doing their work in a relatively cheap and easy way first and for all. The developers aren't interested in how you can get the most credit at any project. The tool needs to be simple enough so everyone can work with it, on as many projects as they want to, without them having to interfere with the running of it.

If you want to micromanage BOINC purely to get the best credit in as little time as possible, then that's your choice. Then you'll just have to figure out a way to do so that suits you. It's not a reason to go change the program.

---