My question is how many total timesteps are in a hadcm3n model? One of my computers is doing about 25,000 timesteps per day, or 3.5 TS/s. How long would it take to do the entire model?

Will it be able to do it by the deadline in august? Oh and it's running 24/7.

Here's a link to the UW: resultid=12885657.

That model type trickles once per model year and it's a 40 year model. You are trickling once per day so it's going to take 40 days to finnish it.

Timesteps = 1,039,392
Mine are looking like 550 hours on a Q6600 24/7

Deadline? Hello ..... What deadline? This has been posted about ad nauseam over the years. THERE IS NO DEADLINE!!!!!!!!! Just a BOINC artifact.

Admittedly, only those models fully returned in a timely manner will be used to seed the next stage, but the rest will still get used. As is the case with ALL of the model types run over the years.

Thread by the people running this model is here.

---

Backups: Here

I wanted it to be done by the research paper deadline or whatever the scientists were talking about.

40 years * 25920 credits/year * 3.5 TS/s / 3600 s/hr = 1000 hrs.

There's been so many people lately quoting deadlines it's getting past annoying.
However, in your case, I can see now where you're coming from. Sorry.

That computer is running at about half the speed of mine (3.5 s/TS as against 1.84 s/TS), so yes, about 1000 hours / 40+ days.
Call it 6 weeks total, so the end of June.
It could be sped up by not running your other projects at the same time, but that's up to you.

It may depend on how excited the RAPID group gets when a lot of the fast crunchers start finishing in the next week or so. 3 of mine finish in about 3 days, and another 3 a day later. Other people have faster machines.
That'll be about the weekend. Still, maybe they'll have a quick test run of a stats program.

---

Backups: Here

OK, how about the HadCM3L model? How many timesteps are in this model? I've been running two since March. These models are monsters; they must be 80-year or more models.

Here are the stats (both are similar):
Elapsed 39 days, 23 hours, 49 minutes
Speed 1.75 sec/TS
Timestep 1,918,080 (92%)
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=12717259

And where are the pretty graphs that we've been spoiled with other models? :)

The hadcm3igeo WU’s are also 80 years long and they have 2073960 time steps. The hadcm3L should have the same number. If you think these are “monsters” you should have tried running the old 160 year hadcm3’s. They took about 8 months on my old 1.5 GHz machine.

---

Graphs software wasn't written before the former project people left. There might not be any.

"I'm late! I'm late!
For a very important date!
No time to say hello, goodbye!
I'm late! I'm late! I'm late!"

The White Rabbit
Alice's Adventures in Wonderland

---

Backups: Here

This is more about the "reliability" of the time calculation, but timestamps are also involved - if there's a better thread, just tell me ;-)

I'm running a hadcm3n and got my first trickle points after about 100 hrs - I know some things are not yet perfect in that area. Soon after that, I had an error window pop up: Error in MicroSoft Visual C++ Runtime Environment and, in good old windows expertise 8-}, solved that with a reboot.

Now, the "time passed" is still going normal, but my "%completed" dropped from a bit over 10% to 0% and the "time remaining" increased from <700 to >1100 hrs.

Question: Is it worth finishing that WU and risk losing 1200 hrs?
Or abandon here and lose 140 hrs for sure?

Dear Andy

The are 1039392 timesteps in the CM3n models. You can see this in Boinc Manager by clicking “show graphics”. It sounds like you have a problem. 100 hours to first trickle is very long. My 2.2 GHz machine trickles about every 30 hours. The sudden drop from 10% to 0% indicates that your model may be looping.

---

Thanks, Jim!

Actually, the "time remaining" has increased another 50 hrs in the meantime :(
Sure sounds like a loop ...

So that leaves me 1.900 Pts for 150 hrs!

Andy

The error messages for somw of your models shows lots of Suspend request from BOINC
This may mean that you've left one of the options at it's default value of 25%. So when your usage of the computer exceeds this value, BOINC (and the model(s) ), will stop running.
Which in turn means that you're running the model(s) very slowly.

It also makes the model(s) more susecptable to sudden failure, if the suspenion happens at a crucial moment.

I think that the option is something like Suspend computation if CPU usage exceeds
It's of more use to other projects. Setting the value to 0 (zero), will turn this off.

---

Backups: Here