And, just think, not THAT long ago we were wasting all those GPU cycles completely 24/7 ... now we are wasting them only SOME of the time ... in other words, the cup is really half full ...

Look @ it this way Paul, for some people the Cup is never full or even half full, if they had full Caches they would just find something else to Whine about ... :)

The pentagram goes on the ceiling, a ring of salt on the floor:D

Oh man, no wonder that one didn't work!

You say "set to DL 3 days of work". Do you mean the cache is set for 3 days? GPUGRID has short turn around times and typically will want work back within 3-4 days so that might be the reason why you can't get any (as well as BOINC 6.4.5 not working properly). Try setting the cache a bit lower to say 2 days, that way BOINC should think it can complete the work in time.

I tried both longer and shorter caching, neither worked at the time. It was a server problem yesterday AM. Those who had work in their caches probably didn't notice it but many had the no work problem. I don't use v6.4.5, it's highly flawed (they need to take the message saying to use v6.4.5 off the front page). At the time I tried v6.3.21, v6.4.1 and v6.4.2. None of them worked during the server problem. Things seem to be working again (at least here) and I've received 3 WUs since the server started working more or less properly again.

I use v6.4.5 on Ubuntu and a 1.5 day cache. The system has worked as it is supposed to for the second day in a row. When work is completed, it gets one more unit. No intervention on my part.

Of course, now that I have said that....everything will fall apart.:D

Need more salt.

mike

- I'm sure the devs know the message "no cell blabla" doesn't fit.. but since the real reason is not known yet they didn't make it more specific. Might be good to at least include one for "really no gpu work available", though!

- yesterday was a pure server problem, the server logs have been sent to the BOINC devs.. let's keep our fingers crossed they find the bug

- for everyone who accidently painted the pentagram on the floor: you can still use it, no need to rip off your parquet just yet! Do the dance upside down and mount the salt to the ceiling, just be careful not to choke yourself while blowing the horn.

MrS

---

Scanning for our furry friends since Jan 2002

And, just think, not THAT long ago we were wasting all those GPU cycles completely 24/7 ... now we are wasting them only SOME of the time ... in other words, the cup is really half full ...

Look @ it this way Paul, for some people the Cup is never full or even half full, if they had full Caches they would just find something else to Whine about ... :)

Like me complaining at Rosetta that 4 tasks died with a File lock problem ... I have since, now, run one task to completion and have another with an hour to go ...

Ah well, if not one thing it is another ...

My last manual update now dates back to two days ago when I managed to get 4 WUs. Since then work fetch works again automatically, my cache being kept at 3 WUs almost continuously with one running. Great! And just in time for me since I am leaving in an hour for a week to visit my parents.

Just one remark though. I noticed in BOINCVIEW that my waiting WUs have a completion time of 11 days and almost 16 hours. As soon as a WU starts it immediately jumps to a completion time of 7 hr 50 min, which is BTW pretty accurate for my GTX260.

Kind regards and happy crunching.

Alain

I got the now work message from a call 12/29/2008 4:59:50 (California time) and I think the situation was caused THIS TIME by having one task in my queue that was rated at 30+ hours. Now, that was from having that 2 plus day task the other day and still thinking that the computer has a 9800 GT vs the new GTX 280 ... I am 36 minutes in, 9.8% done and the time is dropping rapidly from 30 hours and is currently at 26:06 and falling.

*SO*, what that tells me is that the "generic" message we are getting for all sorts of cases is probably related to the scheduler thinking that I have plenty to do and there is no need to give me more work ... my request asked for 56,124 seconds of work, my queue is set to 0.50 days extra work, I have 6 projects on the computer with two out of work (Pirates and LHC) with work from Cosmology and WCG, I don't know why Malaria is dry ...

So, I have 9 tasks in progress, one GPU Grid, 4 WCG, and 4 Cosmology... I have 7 cosmology queued and 4 additional WCG ... runtimes span 2 hours remaining to 20 hours to complete from cold start ...

Thumbnail look seems like I only have about half a days work in hand ... though I may be just on the other side at 0.6 days in hand...

I need a nap ... it is likely I will be back up in a couple hours and I will report if anything changes ...

Oh, when I looked there were 700 plus tasks in the feeder queue ...

That's similar to the situation here. To get new work I have to stop the other projects. I think it's a problem with the handling of the scheduler and the high prio mode for the core to feed the GPU in the background, hidden for the user. Only if all other tasks are stopped and the system makes a call for more than 250,000, I get new work. Here are running: CPDN (over 800 hours), PrimeGrid and MilkyWay with work for the workcache of 0.5 days, LHC and Pirates calling for work without success.

---

Ok, up from my nap, one task nearing completion I asked for more work and got two tasks.

Yes there are issues with the work fetch policy in that it has not been generalized. Worse ... well, I posted another thread where you can see the discussion as it stands at the moment...

But, yes, there are issues and the current policy as coded in the BOINC Manager may let the CUDA resource go idle ... oops ... :)

I always had these messages :
Full-atom molecular dynamics for Cell processor is not available for your type of computer.
Full-atom molecular dynamics on Cell processor is not available for your type of computer.

They irrate me a lot now. I tried reattaching project and all of the options available but nothing solve the problem.
These messages happened on host 6362, 5716 and 15576.
I am really tired of this situation.
Fix it rapidly please.

This is really a problem for us as well. There was a change of scheduler policy a couple of weeks ago in the BOINC server which is buggy. Please post on the BOINC dev forums as well. The holiday period is slowing them down in fixing it.


gdf

I had the problem for a couple hours this morning but just got two more (whew!) ...

Are you feeding PS3 and Nvidia systems off of the same feeder?

I wonder if the pull rate is such that the feeder gets "clogged" with tasks for the other platform ...

Other projects have had issues like this ... in the dark past ...

Looking at the server status it seems there's only one feeder for all platforms...

---

pixelicious.at - my little photoblog

No work again. Second day in a row. (As my 280's sit idly by contemplating their existance!!)

Pat

When is this "Atom Not available " Problem Going to be fixed? It Has been going on for about 2 weeks now.... Is Any one addressing this problem? I have tried 6.4.2, 6.5.0 versions, Reset project several times , Reloaded drivers Etc..and nothing works... Please help... Thanks.

---

"lllvette" wrote:

When is this "Atom Not available " Problem Going to be fixed? It Has been going on for about 2 weeks now.... Is Any one addressing this problem? I have tried 6.4.2, 6.5.0 versions, Reset project several times , Reloaded drivers Etc..and nothing works... Please help... Thanks.

Since you are posting in the right thread, could you please look at the last posts? ;) Especially the one 4 posts before yours...

The project admin is aware of the problem but he can not fix it because there's something wrong with the server software and they are also waiting for a fix from Berkeley...


"GDF" wrote:

This is really a problem for us as well. There was a change of scheduler policy a couple of weeks ago in the BOINC server which is buggy. Please post on the BOINC dev forums as well. The holiday period is slowing them down in fixing it.


gdf

---

pixelicious.at - my little photoblog

Not able to get any new WU's even with manual priming...5 machines... old "Atom Not Available" message...server problems???

Same here, got the last WU at 4:44 UTC, now 3 machines will running dry in the next 4 to 8 hours.

---

When is this "Atom Not available " Problem Going to be fixed? It Has been going on for about 2 weeks now.... Is Any one addressing this problem? I have tried 6.4.2, 6.5.0 versions, Reset project several times , Reloaded drivers Etc..and nothing works... Please help... Thanks.

Nothing you can do about it lllvette, some of my Box's are out too and I've switched them to 4&1 from 3&1 so 1 Core doesn't sit Idle waiting for another GPU Wu.

It wouldn't be so bad though if the Project would change that 24 Hour Wait Period before it trys to contact the Server again so we didn't have to Manually Contact every so often to try & pick up a Wu ...