Downloaded a WU from CPDN on April 27, 2006. Granted, my pc is not on 24/7, due to my efforts on saving ss much energy as possible (which is why I turn of my computer most of the day). I also have an account at Seti@home and Einstein@home. When I receive WU\'s from those accounts they get finished in less than a month. Now I know the CPDN WU\'s are much larger than the S@H and E@H-WU\'s. But with a used processor time of more than 240 hours and a finished percent of only 8.529 %, I\'m wondering if I did not make a mistake when subscribing to CPDN. It would seem the way I use my computer a CPDN WU won\'t get finished in a long time.

---

All of the short models have been completed, and the current models are 160 years long.
On a 3.2 GHZ P4, running 24/7 with no other projects, and little other use, one of these models will take 90-100 days to complete.
Slower computers, and/or running less than 24/7, and/or running cpdn for less than 100% of the BOINC time will take a LOT longer.

Some things to speed up the processing:
1) DON\'T run the screen saver. Slow computers / low memory / no or very low end graphics card, can cause the screen saver to take up to 50% of the processor time.
2) Set the General prefs option: \"Leave in memory when pre-empted\" to yes. This will save a LOT of time when the model is either pre-empted by another project, or Suspended to do other work.
3) Check with the \"Show graphics\" button to find the best time to exit from the model and program. This is after the count down timer has returned from zero back to 432, which is just after the data has been saved.
To see this timer, press Z, then 8.

Addendum:
Unlike the earlier types of models, with the TCMs, small amounts of data are now returned with each trickle. There is also a larger (approx 5 Megs) amount of data returned every ten years in the form of a zip file, and a bigger still file every 40 years, which contains \"restart data\", which will eventually allow an, as yet unwritten program, to start a model from any of the 40 year positions.

Data up 2000 is Hindcast, and from there on, Forecast, so just getting to the year 2000 provides a lot of info. Every 10 years after that is good, too.

Slower computers, and/or running less than 24/7, and/or running cpdn for less than 100% of the BOINC time will take a LOT longer.

Even though it\'s not top of the line, my cpu is not THAT slow, considering it\'s a 64-bit AMD Athlon 3200+, running ona 64bit Linux OS. However, whether my BOINC-setup running CPDN can take advantage of that is another matter. The CPDN WU is the only WU I have at the moment, so when BOINC is running, it\'s running that.

Some things to speed up the processing:
1) DON\'T run the screen saver. Slow computers / low memory / no or very low end graphics card, can cause the screen saver to take up to 50% of the processor time.
2) Set the General prefs option: \"Leave in memory when pre-empted\" to yes. This will save a LOT of time when the model is either pre-empted by another project, or Suspended to do other work.
3) Check with the \"Show graphics\" button to find the best time to exit from the model and program. This is after the count down timer has returned from zero back to 432, which is just after the data has been saved.
To see this timer, press Z, then 8.

1) Is there a way I can see how much cpu the screen saver uses? Since usually, when I touch a key the screen saver disappears. The computer isn\'t slow (look above), the graphics card isn\'t either (look below) and the memory is 512 Mb.
2) Since when shutting down my PC CPDN shuts down anyway (and the memory is detached), and the CPDN-WU is the only WU, at the moment this is irrelevant but that setting was on anyway.
3) I think when I do \'show graphics\' the computing time is slowed down considerably. My graphics settings are rubbish here, because everytime when I try to update the driver with a friggin\'-ATI driver my complete X-Windows screen settings crash and it won\'t start up again. SuSe/Novell and ATI are apparently NOT good buddies since SuSe had to make an ati-driver for its linux dis all by itself.

Data up 2000 is Hindcast, and from there on, Forecast, so just getting to the year 2000 provides a lot of info. Every 10 years after that is good, too.

I\'m just up to 1934 staring from 1920 (have I mentioned it takes A LOT of time yet? :-P) at the moment, so even 2000 is still a long way. Anyway, if I abort the WU when it reaches 2000 (January 1 or December 31?) CPDN will still be able to benefit from all my (pc\'s) hard work? After that, I think it\'ll be a long time before another CPDN WU shows up again, even if I didn\'t instruct BOINC to NOT get a new one (which I have, because even though BOINC complained it was overcommitted it managed to sneak in another CPDN WU nonetheless. Of course, with that new CPDN WU competing with the other one, neither would get finished by a long time, so that\'s why I set that).

---

1. Use \"Show graphics\", full screen. Check with top. You\'ll also notice a difference between displays, such as 3-D clouds and temperature.

2. The reason we shut CPDN down first is that there are so many files open that boinc doesn\'t always get the job done in time. (This is usually more reliable in Linux than in Windoze, in my experience, but not foolproof.)

Uploads occur ~04 December of each decade -- 1930, 1940, 1950...

I started using Linux nine or ten years ago. Graphics drivers have always been a pain in the backside (and not just graphics drivers).

---

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

I just checked, and my P4 is getting 2.75 s/TS on one half, and 2.78 s/TS on the other.
The 1st is checkpointing at 26 minutes, and the 2nd at 29 minutes.
These are more realistic \'benchmarks\' to use than the processor type or speed.

The \'problem\' with a 64 bit processor, is that these models use lots of floating point calculations, and the FPU (floating point unit) didn\'t change much between 32 bit and 64 bit processors. So there\'s not a lot of advantage to them, apart from perhaps getting through \'normal\' useage more quickly, and so not taking up as much modelling time as a 32 bit processor.

The models will take as many hours to run when the computer is only on for a couple of hours a day, as they will when the computer is on all of the time.
So you won\'t save any more on electricity; you might as well run 24/7 and get it over with.
And during the colder parts of the year, a well ventilated computer will heat up a room to a certain extent, as several people on the BBC part of the project have found out.

@astroWX

1. Use \"Show graphics\", full screen. Check with top. You\'ll also notice a difference between displays, such as 3-D clouds and temperature.

The \"Show graphics\" display in normal (not full) screen takes away half of the available cpu-usage, on full screen that\'s reduced to taking up one fourth/third.

2. The reason we shut CPDN down first is that there are so many files open that boinc doesn\'t always get the job done in time. (This is usually more reliable in Linux than in Windoze, in my experience, but not foolproof.)

? Shut CPDN down first? BOINC doesn\'t always get the job done in time? I don\'t get it. BTW, I pause BOINC I shut down the pc. Is that an unnecessary precaution?

@Les Beyliss

I just checked, and my P4 is getting 2.75 s/TS on one half, and 2.78 s/TS on the other.
The 1st is checkpointing at 26 minutes, and the 2nd at 29 minutes.
These are more realistic \'benchmarks\' to use than the processor type or speed.

2.44 s/TS.

The models will take as many hours to run when the computer is only on for a couple of hours a day, as they will when the computer is on all of the time.
So you won\'t save any more on electricity; you might as well run 24/7 and get it over with.

In the past I\'ve let the pc keep on running during my night sleep, for other reasons than BOINC, and when my log still made a line showing the time each hour, I noticed that during the last few hours the calculating speed started dropping dramatically, even to half of normal operation. And whilst running it 24/7 will not save me electricity, it will save me money because during the day electricity is much more expensive than during the night.
Thanks for the help on the time when progress is saved (432-timer).

---

In the past I\'ve let the pc keep on running during my night sleep, for other reasons than BOINC, and when my log still made a line showing the time each hour, I noticed that during the last few hours the calculating speed started dropping dramatically, even to half of normal operation. And whilst running it 24/7 will not save me electricity, it will save me money because during the day electricity is much more expensive than during the night.
Thanks for the help on the time when progress is saved (432-timer).

Do you have enabled a GNOME/KDE/what_ever_desktop_you_use screensaver too? You should disable it or set it to blank_screen or black_screen only.

---

Linux Users Everywhere @ BOINC

Do you have enabled a GNOME/KDE/what_ever_desktop_you_use screensaver too? You should disable it or set it to blank_screen or black_screen only.

It\'s KDE. Can\'t see an option to disable screen saver, so I\'ll just set it to black screen. Anyway, in the past, the screensaver kicked in only seven minutes after the pc was left inactive, and the CPDN WU only started slacking many hours later.

---