I just received work unit 7648484. I have now put in about 5 hours of computing time on it. It is running at high priority and, if experience is any guide, it will run in high priority until it completes. The current deadline is 24 Jan 2012, and according to the BOINC manager, it has about 2852 hours to go. It seems to me BOINC knows how fast my machine usually runs, since it runs a benchmerk every once in a while. If I calculated correctly my machine will take about 34 weeks to finish this, so it will be well past the deadline. I know late answers are accepted, but it puzzles me that they can be so far off in assigning jobs to my machine. My recent jobs tend to complete several weeks after the deadline. Is the benchmark timing used for anything? Also, is feedback built into the dispatching algorithm to determine what tasks I am sent so their predicted completion time would be withing the deadline?

---

The hadcm3 models currently being sent are for the RAPIT project (investigator's thread here, experiment summary here). The pre-release testing of these models took longer than expected, and the results are needed ASAP, so the artificial "deadline" is short.
They will run in high priority for a week-10 days or so, and by then BOINC should have figured out it's new work load/schedule. Provided that your other project's work doesn't differ too much from what it is now.
The "time remaining" should drop quickly, as BOINC works things out.

I don't think that the built in bench marks are used by many projects these days, but some still do.

There is code available to match work with computers, but it's not being used yet for various reasons. (I think that one is the extensive customising needed with this project.)

PS
Your computer is crashing all hadam3p (regional) models, because it's lacking a needed file.

../../projects/climateprediction.net/hadam3p_eu_6.09_i686-pc-linux-gnu: /usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.9' not found (required by ../../projects/climateprediction.net/hadam3p_eu_6.09_i686-pc-linux-gnu)

---

Backups: Here

My hadcm3n is still running in high priority mode after 108 hours and 9.9% done. This does not forbid other BOINC projects to run on my second core (I have 2 on my Opteron 1210 running Linux) but my BOINC_VM Virtual Machine, which uses 2 cores in its latest version of Test4Theory@home, won't start if hadcm3n is running. So I have to suspend it, risking to miss its deadline of February 19.
Tullio

---

Don't worry if it misses its deadline as the CPDN server always accepts overdue trickles and files.

It does seem crazy that Hadam3P which is relatively short has a deadline some 50 weeks away while the much longer Hadcm3N is given 3 months. It's because the oceanographers at the RAPIT project need the Hadcm3N results more urgently.

---

Cpdn news

I am a little confused about the deadline and hours to completion. This computer http://climateapps2.oerc.ox.ac.uk/cpdnboinc/show_host_detail.php?hostid=1060288 has been slowing down for the past 6 months to a very annoying slow pace. When I run a CPU benchmark the results vary each time. I noticed that when the results are a bigger number the time to completion jumps up dramatically. A full ocean just down loaded and the completion time is over 25,000 hours.

2/15/2012 5:37:09 PM Running CPU benchmarks
2/15/2012 5:37:09 PM Suspending computation - running CPU benchmarks
2/15/2012 5:37:41 PM Benchmark results:
2/15/2012 5:37:41 PM Number of CPUs: 4
2/15/2012 5:37:41 PM 322 floating point MIPS (Whetstone) per CPU
2/15/2012 5:37:41 PM 764 integer MIPS (Dhrystone) per CPU
2/15/2012 5:37:42 PM Resuming computation

The completion hours just jumped down to 5500 hours. If I remember correctly the numbers were 888 and 2300 now they are 322 and 764. Wouldn't it make sense that a higher floating point would yield better performance thus a shorter completion time.

The 16 to 25 seconds per time step is correct. It never was that way and I feel something is wrong somewhere.

Help me out here and I will clarify if you have any questions.

---

Dean,

The CPU on that machine is a Core i3 330M, which has two cores with hyperthreading giving four threads. These four threads are not independent, and as a result give only about a 20% speed increase (i.e. two threads = 1.2 cores): in other words, if you run four models instead of two models they will each take about 1.8 times as long. That's a net gain in throughput though each model takes longer. However, it's only a gain if the models run correctly, which is not the case for that machine - which has a large number of crashes.

So, quite separate from the problem of how long BOINC thinks each model is going to take, my advice is to back that machine off a bit. Try running two at a time instaed of four - and you (and the project) might then get a few completions.

Iain

Iain,
Thanks for the response. I suspected that backing off would be better so I have been running at 75%. That allows a EU, PNW and a seti to run concurently. I will keep it this way for a while to see if the results improve.

Would you comment on the benchmark test. Doesn't a higher number indicate a faster processor which should give a shorter run time. Maybe the BM application needs to be downgraded or re-installed. It's not like it used to be.

---

That is a processor for a laptop, so I'm assuming it is a laptop. Heat can be a problem with those, and if it is overheating, the CPU will throttle its speed down, and give poor performance in cpdn and the benchmark.

It's possible the CPU cooling may have accumulated dust over the time it's been running. Using compressed air to blow out the dust through the vents in the side/bottom may help. Also, a cooling pad for a laptop can help with keeping the laptop cooler. After cleaning it out and making sure it's well ventilated, if it's still running very slow, you should probably decrease the number of tasks being run at a time even more.

It's also a bit annoying that,once you pass the deadline, you keep getting red-highlighted messages telling you that you're x days past the deadline and may not get credit so should consider aborting. I bet quite a lot of people do abort! But I've just persevered and finished a long one - I think it was 25 days past the deadline in the end.

---

Visit the Scotland team

It's BOINC that's complaining, not this project.

---

Backups: Here

It's BOINC that's complaining, not this project.

Thanks, Les, I know that, and I also know to ignore such messages for CPDN but I wouldn't be surprised if some people do abort.

---

Visit the Scotland team

also TThrottle is a good program to help keep your cpu from melting. Also if it is plugged in all the time remove the battery.