I have connected one of my hosts a few day ago to climate prediction. Later I realized that it doesn\'t meet the minimum requirements.
I have an AMD duron with 1.3Ghz (instead of 1.6) running linux.
The expected time was 3520hours (less then 6 month).
The host works (full time and dedicated) on 2 projects, ideally sharing 50%. That would bring climate to above 7000h of elapsed time, close to one year (the expected result date is about one year ahead of now (just in time?)
Now, what happens is the following:
- Climate prediction is hijacking 100% of the CPU
- The remaining time grows instead of decreasing.
I wonder whether this is due to the inadequacy of the machine. I don\'t want to wait one year to see what happens.
I have already crunched close to 90 hours, and got a first credit award.
I see two possibilities: dedicate this host 100% to climate prediction (but only if there is chance to succeed) or kill now climateprediction (and may be connect, but restart from scratch, from a more powerful host of mines.
Any advice?

---

The machine has enough memory and that helps a lot. Running Linux should also help, compared to Windows. It will also help if Preferences in Your Account are set to \"yes\" for \'Leave in memory when suspended\' and \'Do work when computer in use\'.

It is recommended to look through the Read-me Sticky posts for avoiding problems, making backups, etc. Even if you don\'t read everything, it\'s useful to know what\'s there, so you\'ll know where to go if necessary at some future time.

CPDN is probably \"hogging\" all the CPU time because boinc thinks the machine can\'t finish the Model in time, so it gives everything to CPDN in a valiant attempt to make the deadline. (Don\'t worry about the deadline; CPDN accepts and credits results, even if late.)

If you want to finish the CPDN Model (it\'s quite an accomplishment!), then set your other Projects, in Manager\'s Projects tab, to No New Work. (That way, when their current work finishes, no new work will be downloaded.)

Especially when one\'s machine is marginal, it\'s very important to make periodic backups of the entire boinc Directory, after CPDN and boinc are exited. Weekly should be good enough (whatever you feel comfortable re-running in case of necessity for restoring after a crash). Instructions for making backups are in the Stickys.

By the way, estimated time to completion will change during one\'s first Model, both up and down, because boinc is trying to determine how fast your machine can do the job. (Boinc is designed for short-running Work Units. CPDN drives it a little crazy.)

Welcome to CPDN, Maurizio!

[Edited for typo.]

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

Yeah, 1.2GHz here; I feel your pain. My upgrade went bad due to MB voltage problems. :( I\'m 9% into it already but based on my realistic calculations, mine will take 9-10 months if it finishes. However, I have dual Athon MP CPU, so performance is OK just running one BOINC task all the time.

In your case, I would choose one project or the other. But I believe it is possible for your system to finish one work unit within alloted time. It just depends on how much BOINC runs.

In your case, I would choose one project or the other. But I believe it is possible for your system to finish one work unit within alloted time. It just depends on how much BOINC runs.

Well, I would actually choose climate (for this machine, I have some other working on other projects). The machine is also 100% dedicated, it\'s my old PC that I have replaced.

But before choosing, I would have liked to finish the current WU of WCG running on the same machine. It is only 4 hours to the end, but Climate is not leaving it any further CPU time. What is strange is that I tryed to suspend climate to let the other WU end. But also with climate suspended, it remains in the \"Waiting to Run\" state. I have 3 days left to complete these 4 hours, any idea how to (of course without killing climate)?

---

Dear oh dear.
This is all in the READMEs here.

The part you want is Suspending. In this case, Suspending a model.

1) In Projects, click on the climate project name, then click the \"No new tasks\" button to the left.
2) In tasks, click on the climate model, then click the Suspend button to the left.

The reason that the climate model is \"hogging the processor\", is that BOINC is in shock at the huge amount of processing time needed for the climate model, so it\'s giving it all of it\'s attention. Soon it will realise that it CAN process 2 projects together, and will start allowing the other project some cpu time.

The secret to running a climate model alongside work units from other projects is PATIENCE. Lots of it!

---

Backups: Here

The part you want is Suspending. In this case, Suspending a model.

1) In Projects, click on the climate project name, then click the \"No new tasks\" button to the left.
2) In tasks, click on the climate model, then click the Suspend button to the left.

[....}

The secret to running a climate model alongside work units from other projects is PATIENCE. Lots of it!

Thank you for helping.
I did wat you suggest: my idea was just to suspend the CPDN for the few hours to complete the other task, and then let it go again.
What happens, is that I have CPDN in supended state, and WCG in \"Waiting to run state\". Apparently BOINC is still confused, also after CPDN is suspended.

If I succeed completing the WCG WU, then I have no problem waiting one or two month to see how CPDN stabilizes, and if more tasks may run together.
No problems to be patient.

---

I guess I got it!!!

I just had to restart BOINC client, and now the WCG task is processing normally, while CPDN is suspended.

---

My BBC model due to finish tonight spent about 120 model years on a 1.33GHz machine. It met the minimum specs for running a model on memory - 512Mb. The memory much more critical than the speed. It was doing 6sec/TS. But then the CPU probably fractured and to keep the machine running it had to be underclocked by 25%. I let it do that for a while (about 7.5sec/TS) and then let it race to the end on a new computer. But if I\'d left it on the old slow computer, I think it would still have finished.

The cpdn server ignores the deadline, unlike the servers from other projects.

The longer your model is likely to run, the more important it is to take regular backups. More time = more chances of a model crash.

---

Cpdn news