I'm using a laptop - 2.66 ghz, 512 RAM and a 40 Gb hard drive. I've just started running CP on Boinc along with Seti (80% resources for CP and 20% for Seti)and the current estimated completion for my first CP unit is 1100 hours. Is this normal? Seti units only take about 1.5 hours.

Tom

---

Hi Amphlaon, welcome to the project. And your first big shock.

Yes, it is normal. Approximately anyway. The time will shrink as it progresses and gets a better idea of total time.
Other projects, as far as I can see, take a list of results and compare them against a list of criteria,
looking for a match.

CP, on the other hand, takes a set of parameters, and spends several weeks building a model of the weather
over 45 years, in 3 separate time ranges.

A lot of people think it's like comparing apples and oranges, but it's actually more like comparing apples and horses.

Some info in the FAQ to the left of here, and LOTS more in the Climate Science section.
And, as you will see, this is just the start. Experiment 2 will take a LOT longer per model.
As does the Sulphur cycle model currently in Alpha testing, which has 5 phases instead of 3.

And a hint: raise your laptop a bit above the bench surface - it will run cooler.

Les

5 phases!!! geez ive never been able to make it through 3 consectuively.farthest ive got was 57 ts away from the end of seg one. then i started testing the aplha versions of boinc and they liked to forget where they put there stuff.

Reformatted my computer after deleting parts of windows. the os was fine but the apps wernt. calculator, note and word pads disk cleaner, defrager all down the drain. now instead of 10 sec per ts its 7.5 sec per ts, go figure.:]

---

boinc beta, predictor alpha/beta, lhc alpha/beta, cpdn alpha/beta

Cp is a full time job. The figures I mentioned were based on the assumption that you ONLY run cp, and do so full time,
24 hours a day. Running for part of a day, and / or using processor time to run other projects, will lengthen the
time to complete a model. And I have a feeling that using BOINC to swap projects contributes to failures in CP.
But thats just MY view, based on reading about lots of problems from lots of people.
My computer is monogamus, and runs continuously. It takes 36 days to crunch a pair of models.
I've had 11 successes, one failure, and a second which was my booboo.

What we are running normally runs on a 64 bit supercomputer. Over a million lines of fortran, converted to about
half a million lines of 32 bit fortran to run on desktop computers.
This is 'big boys toys' stuff. So keep crunching, and hope. Not sure if praying helps with this, but it's worth a try.

Les

Thanks for the info, looks like I'll need to invest in a supercomputer.

Tom

---

I'm saving up too.
For now though, just a desktop. At present the fastest is an Intel P4, (because of the compiler used), running Linux.
However there are hopes that a new compiler, (which may be a long way off), will not favour either Intel or AMD,
and that it may support 64 bit cpus. 64 bit machines would speed up crunching a fair bit.
And laptops, I think, use a processor with a 400Meg fsb, whereas a desktop P4 uses 800Meg fsb.
This speeds up data flow, which is very intensive in CP.

And then there are the multi-chip machines coming up in the near future, which may well bring us back to
small supercomputers. And big morgages.

<a href="http://www.meto.gov.uk/research/nwp/numerical/computers/index.html"> This</a> is the Met Office computer page, which has a link to their previous machines.

Les

I suspect that desktops do have the edge. My desktop with 2.66GHz P4 running with 540 MHz FSB under WinXP, is returning 2.27 s/ts and calculating around 500 hours for completion. I wonder whether Amphlaon's total estimate has come down now, as BOINC needs time to give a reliable one.

As Amphlaon is running more than one project he should make sure that his preferences are set to keep the program in memory when suspended.

---

It got down to 770hrs to 840hrs after a few hours, looks like desk tops have the edge.

Tom

---

http://www.cray.com/products/x1e/index.html its a supercomputer. its predecessor has a price tag of 100,000 USD. ran lhc units in 2.5 seconds. dont know how fast this one is an average but the highest run was 18 giga flops. its a bit out of my league finacally

---

boinc beta, predictor alpha/beta, lhc alpha/beta, cpdn alpha/beta

Sigh!!! :(