Need POWA for ClimatePrediction!

[Rattledagger]

The CPDN-stats is probably just like in seti, there someone that haven't done anything doesn't show up.

They've also got the stats sorted now. :beer:

 

[Silverthorne]

This is also strange since our total credits seem off by 1436.66.

 

[Rattledagger]

In BOINC the credit earned by a user while on a team will increase the "team-credit" the same amount. This credit will stay even if the user moves to another team. If a user is not on a team of course no team will get this credit. The user will always have his earned credit regardless of team or not.

In CPDN you'll get credit for every trickle, and most likely the "team-credit" increase at the same time. If a user had trickled before joining the team, this credit will therefore not show up under "team-credit". 1436.66 is the credit for 19 trickles.

 

[Ninjazx]

its me :X

but i cant seem to get much processing done, every once in a while it says that theres no work to be done :-\
In the messages box "requesting work from server: no work recieved (daily quota exceeded)

 

[Silverthorne]

Originally posted by: Ninjazx
its me :X

but i cant seem to get much processing done, every once in a while it says that theres no work to be done :-\
In the messages box "requesting work from server: no work recieved (daily quota exceeded)

I may be wrong about this but I think you need to log into your account and change the settings under "preferences" and then "general".

 

[Rattledagger]

You shouldn't get "daily quota exceeded" if you haven't got assigned some work. If you get this in CPDN you're most likely constantly crashing the wu you've got assigned...

You can try completely wiping out the install to make sure there's no problems, make sure you're not overclocking, also note there's problems with some OS in the beta...

 

[Ninjazx]

its all working now, it seems. the GUI doesnt want to download any "work" but the non-gui works fine. It makes no difference to me, but now im making progress!

 

[Ninjazx]

i dont seem to be getting any credit now though

 

[Rattledagger]

You only get credit then you've uploaded a trickle... According to the stats, you've trickled 2 times, the last roughly 1 hour ago.

Also, one of the machines has had many computational errors... Of course it can be due to the parameters or a client-bug, but overclocking or some hardware-problems apparently often leaves to crashes in CPDN.

 

[Ninjazx]

the successfully running machine is a laptop sitting next to me, but this box IS overclocked, and is giving me client errors out the end. I'm not sure why, as I ran prime95 for over 12 hours, installed and played through, and beat doom3, installed, played through, and beat painkiller, all without restarting (total uptime ended up being 21 days before i finally gave in and rebooted). but this is the only software that doesnt seem to want to run.

Ill re-clock tonight and give it a shot that way, just to make sure though.

 

[Assimilator1]

RD
Overclocking isn't a problem ,an unstable overclock is though ,I'm sure that's what you meant but you might want to specifie next time
Still ,it is worth downclocking as a diagnosis

 

[Ninjazx]

indeed.

 

[Rattledagger]

Well, from that I've heard many overclocks till it locks up, backs off a little and runs prime95 and a memory-test and calls this "stable".

But CPDN with its very long wu and high system-requirements very often crashes on these "stable" overclocked systems, so atleast in CPDN overclocking very often is the reason for a crash...


Since overclocking from it's name always is to run something over it's spec, and therefore bigger probability of crashing, all overclocks is more unstable than before so...

 

[networkman]

The old saying that "There's no such thing as a free lunch." still plays well here. Sure, you may be able to "overclock" and get a performance enhancement, but at the expense of what? Stability? Reliable data? Chip life? Additional heat? Something has to give.

With the workunits taking such an inordinately long time to complete(as compared to other DC projects), why run the risk of an incomplete or failed cycle? Incomplete workunits really don't do alot of good.

 

[Ninjazx]

Seems to be ok now, aparantly I had some kind of software conflict with something else I had been running. Since rebooting a few times and making sure nothing else was running, it now runs perfectly, and still at 2.4ghz (a-xp). I suppose I'll find out in the morning if it really is working.

 

[Assimilator1]

Originally posted by: networkman
The old saying that "There's no such thing as a free lunch." still plays well here. Sure, you may be able to "overclock" and get a performance enhancement, but at the expense of what? Stability? Reliable data? Chip life? Additional heat? Something has to give.

With the workunits taking such an inordinately long time to complete(as compared to other DC projects), why run the risk of an incomplete or failed cycle? Incomplete workunits really don't do alot of good.

Reliability is only lost if insufficient testing is done ,assuming the user doesn't run it to the ragged edge
As for a little less life & a little more heat ,who cares?
To answer your last question ,make sure its stable before running CPDN
(Btw I didn't realise you were an 'anti' overclockerlol)

RD
lol ,a stable system is a stable sytem if tested properly (regardless whether its overclocked or not).I take your point though ,many people don't

 

[Rattledagger]

Originally posted by: Assimilator1
To answer your last question ,make sure its stable before running CPDN

But that's just the problem, the only way to know your system is stable running CPDN is to run CPDN...

Of course a wrong calculation in prime95 or error in memory-test or similar can very fast prove a system is unstable, but even passing these tests with no errors isn't a guarantee CPDN will not crash after a couple of hours/days/weeks due to a hardware-problem.

In seti@home there an average machine now crunches a wu in... 6h50m and every wu is crunched by atleast 8 different users, someone returning an occassional wrong result due to unstable overclock or something isn't really a problem. But on a wu that on average takes 3-6 weeks, uses 50 MB memory, and 650 MB hd-space...

 

[networkman]

I'm not an "anti-overclocker" so much as I am "pro-reality". The reality is that prices have dropped so much that the amount of power needed to accomplish X task in a given period of time is cheaper than it ever has been, to the point that there isn't much in the way of need for over-clocking. Sure, it's a great hobby for some, but for the rest of us, it's more economical and safer to put together another system, especially when speed AND stability are of paramount importance. CPDN is much more sensitve to a stable process than other DC projects.

 

[Assimilator1]

RD
Ok ,you have a point if the WU takes 3-6wks!:Q ,I was under the impression it would take a few days ,which would be no great loss IF CPDN did prove a previously thought stable overclock to be unstable and trashed the WU.

NWM
Your implying that I'm 'anti reality' ,I don't think so

Re cost of CPUs etc ,well I guess that depends how well off you are.If forking out £60-£80+ for a cpu is no problem then I'd agree with you.However I know plenty of people who would not want to spend more than £40 ,or even less ,in which case if that person needs 'every ounce' of speed they can get overclocking is a very viable way of getting it (assuming no silly priced HSFs).
And unless you put high amounts of voltage through the cpu ,overclocking is safe,its economical too unless silly amounts are paid for the cooling.
Btw I'm not saying you should (or shouldn't overclock),I'm just saying that is a safe & viable option for many people & can be made stable.
For myself,as far as CPDN goes ,I'd suck it & see ,if it trips up I'd downclock a little.
(anyway ,going back on topic)

If a WU on CPDN typically (?) takes 3-6weeks ,what happens if your machine crashes? (for none overclocking reasons).Do you have to start the WU all over again then?

 

[Idlorj]

It depends on what crashes on the system. The project saves to disk every 144 timesteps.
In a three-phase model there are 777,744 timesteps total.

 

[Assimilator1]

How long does 144 timesteps take on what speed cpu?

Would I be right in saying that as long as no data is corrupted in CPDN itself that a crash wouldn't effect the WU? (aside from the last 144timestep 'block')

 

[Rattledagger]

Well, you can take a look on the top-computers-list...
From this a 3.4 GHz p4-HT uses 1.6s/timestep.

If the model has become "unstable", it will normally try to re-wind last day/month/year and recalculate, so depending on the reason for "unstable" it maybe will continue to crunch...

 

[Idlorj]

The BOINC climate prediction is offically live.

 

[Ninjazx]

well, for whatever reason, i cant run the boinc client on this pc.

non oc'd, overclocked, underclocked. makes no difference.

When it runs, it now runs continously, but it doesnt upload anything. period.

I've tried everything. . . I give up. I'll just run it on my laptop, and probably go back to the old CP on here

 

[Assimilator1]

Maybe the others could still help you?

RD
Btw S@H1 WUs are crunched 4-5 times & not 8 times AFAIK

Going by that P4 3.4GHz you quoted then it saves to disk a little under every 4mins,that's ok.
Does any particular cpu have an advantage over others ? ,e.g. Ath XP vs P4 vs Ath64 ?
Does L2 cache size make much odds?