I\'ve got an old 750MHz machine that is acting as a sort of glorified data logger and is on 24/7. Over the years it\'s done half a dozen slab models and is shortly going to finish a sulphur model (~192 days) - all without any problems.

I\'d like to leave it running CPDN, I prefer it to the other projects, but a \"back of an envelope\" calculation reckons it\'s going to take ~412 days to finish one of the new models.

I can\'t decide if it\'s worth it or not. I know CPDN will still accept the results after the deadline but will the result it generates be any use by then? Is there no way I can force it to get a new sulphur model?

Any comments to help me decide what to do...

Ian

---

Slab models were needed to find combinations of values that produced stable models for the next phase of the project. Sulphur models were needed to get values of sulphates for the next phase, along with the same data as slab.

The \'next phase\' hase now started, and both slab and sulphur are obsolete, so unless someone is completing one of these, they will just be wasting time and electricity starting a new one.
The Coupled Ocean model is now IT.

As for how long this part of the experiment lasts, you\'ll have to ask again in a year or so. So, unfortunately, you\'ll just have to take your chances.

Thanks Les.

Looks like CPDN has decided for me anyway :-) I tried to download a coupled model and it wouldn\'t let me - told me that it \"wouldn\'t finish in time\".

I\'ll wait until the sulphur is about to finish then try again, but it looks like that machine will have to switch to a different project.

Ian

---

.....and both slab and sulphur are obsolete, so unless someone is completing one of these, they will just be wasting time and electricity starting a new one.

I have a sulphur model that\'s only 8.3% into phase 1 which I suspended when I joined the Combined beta. My intention was to complete one of the BBC models then go back to the sulphur one. But if these are now no longer required, and being replaced by the Coupled model, would it still be useful to keep it going? I let it get to a trickle point, which would seem as good a place to stop as any, but I\'m reluctant to kill it if it may still be useful, but also don\'t want to let it run further if this would be wasted effort.

---

Hi Ray

I had 16 trickles to go with my one and only sulphur, which finally finished 5 hours ago.
Any one this far into a sulphur may as well finish it, but Carl has said that starting a new sulphur is a waste of electricty.

Best I can think of is to let it run to just past the end of phase 1, which contains extra data they want/used to want. If it\'s your only sulphur model, you may want to let it finish just for the hell of it.

Just had a look at your models. Is the one in question sulphur_j1u7_000888847_0? Because if it is, it may be one of the faulty batch that fail just after the end of phase 1.
Which would decide things for you :)

Thanks Les,
At 188 hours the sulphur model is at trickle 10 of 24 of phase 1. It is that WU you highlight so if it commits suicide anyway I don\'t have to worry. I\'ve got it suspended just now anyway as I want the BBC model to return as much data as poss before their \"Results\" program although I can\'t see them having many complete runs to report on. It\'s currently approaching the end of 1940 at 12.5%. I foolishly started a Seasonal model as well before reading that I should finish the other models first. It\'s in July at 28.4%. I\'m going to let these 2 finish before making a decision on the sulphur one so I\'ll be back in a month or so to see if there is any update on what the guys would like me and others in the same situation to do.

---

One other thing to note...

Looking at the history from slab, to sulpher, to coupled... I almost would gaurentee that the next phase will not be shorter then coupled. If you notice the trend, the WUs are getting larger. So if they go to another phase in a year I\'d expect the current trend to continue with WU sizes... That comp might be done with CPDN if it won\'t let you d/l...

Now, this leaves 2 matters. For now, faster computers can pick up much of the slack and still let people complete it in deadline with time to spare for other projects. My Athlon 64 in fact has no seeming deadline trouble project sharing between CPDN with a coupled, seasonal attribution, and some other projects...

In time however, as AMD and Intel go multi-core as a means to increase computing power; in part because they can\'t using current manufacturing technology just continue to ramp up clock rates as of old. Conroe, will based on some preliminary benchies give a mighty boost to performance, but like the Athlon 64 it\'s going to be a 64-bit x86 CPU, and will also move in a new direction from Intel, become much more efficient as they also drop the net-burst arch of the P4 altogether. Conroe will also be multi-core like the X2, based on the preview they demonstrated comparing it to an OCed FX60!...

Part of the problem is heat dissipation on these newer procs, and part of it is the laws of physics facing companies such as Intel and AMD... One of the main methods of dealing with heat dissipation was a die shrink. However the silicon atom has size, and eventually one just can\'t shrink any further without hitting the sub-atomic, or going to a smaller atom such as carbon (which shares many similar properties to Si on the periodic table).

The other thing is that die shrinks had posed a bit problematic, difficult to get in working order. There\'s most definitely a reason that Intel hadn\'t continued ramping clock rates to a doubling every 18 months or so for awhile.

Long term, places such as Sandia National Labs (which is local to Albuq. NM here) are looking for replacements to the silicon based semi-conductor, which won\'t share the same technological limits; however such research takes time. In the mean time, multi-core and other such introductions does seem to be the way these companies are going on to compete, as they\'re comming closer to certain limits. Of course 2 cores can do more, however it\'s the same deal as with SMP...

---

However the silicon atom has size, and eventually one just can\'t shrink any further without hitting the sub-atomic...

I\'m not sure if I follow you fully (on your whole note), but are you saying that because I have a P4 hyper threading I should be running \'this\' type of model, or if a 64-bit x86 CPU I should be running \'that\' type of model, and that some models have WU that are not appropriate for some processors?

Surely the faster processor you have the quicker it completes, and what it completes, no matter how quickly, has the same result? And is there a deadline for some models that if they are not completed by a certain time they are no good to the scientists and therefore I could be wasting my time even though I am leaving my PC on 24/7 thinking this is all worthwhile when \'they\' can see I will not complete in time for whatever reason and don\'t want to tell me in case I feel bad about a model I complete but is no good because it took too long?

Sorry about the length of that sentence, but sometimes it\'s best to let it flow.

---

Re: sentence length. Yeah. Know what you mean. :) Sometimes I have to stop and try and work out what I started to say so that I can finish it.

The \"deadline\" exists because something had to be put in there. It\'s intended for the other projects, all of which complete in hours.

However, if your model takes too long to complete, the info may be obsolete.
e.g. One person had problems uploading a THC model (Thermohaline). But by that time, that \'special\' project had long been finished, and the research paper written. Similarly with sulphur models. These were to generate sulphate data to start the Coupled Ocean models.
Some sulphur models are still being completed, but if one is going to take another year to finish it, then it\'s a waste of time.

So, there is a \'sort\' of deadline; the usefullness of the data it\'s creating.
The current Coupled Model is not a problem at present, as this part of the project will run for at least a year.

But by that time, that \'special\' project had long been finished, and the research paper written.

Thanks Les. I\'d hate to hold up a research paper because I forget to leave my PC running when off on an extended holiday (I wish). But now I know I\'m not that important and no one is waiting on me I can rest easy :)

---