I have seen that when there I get the "No work available for your type of computer" message BOINC backs off 24hrs. (I think this is a project setting). This means that I run out of work and my GPU sits idle. Any chance of reducing this time?

Many thanks...

I have seen that when there I get the "No work available for your type of computer" message BOINC backs off 24hrs. (I think this is a project setting). This means that I run out of work and my GPU sits idle. Any chance of reducing this time?

Many thanks...

Exponential back-off, lord I wish they would use that instead ... especially since there is a "cap" of 4 tasks in the queue ... which means I run dry in 6 hours or so ... tops ... on my most productive system ...

I know they want to limit the loss of tasks and tasks completed are used to build the next series ... and the project is pretty reliable and on-line time is one of the best ... but ... jeeze guys ... look at my i7, 4 cores, few errors ...

I have seen that when there I get the "No work available for your type of computer" message BOINC backs off 24hrs. (I think this is a project setting). This means that I run out of work and my GPU sits idle. Any chance of reducing this time?

Many thanks...

I got this yesterday after getting two tasks.

If you turn on sched_op_debug you will see that the project is indeed requesting the 24hr backoff. As Paul alluded to it isn't exponential. I guess they don't have enough work for all of us.

I would have thought a smaller backoff like 1 hour (or 2) would be sufficent. Maybe GDF would care to comment.

---

BOINC blog

Exponential back-off, lord I wish they would use that instead ... especially since there is a "cap" of 4 tasks in the queue ...

The cap of 8 tasks total regardless of the number of GPUs or CPUs isn't very helpful. When will they implement a cap based upon a multiple of the GPU count?

The 24-hour back off seems especially silly when the central server only goes down for 10 or 15 minutes.

We have not changed anything.

gdf

Any chance you could change the 24h back-off to something smaller?

MrS

---

Scanning for our furry friends since Jan 2002

We have not set this anywhere. It must be something of BOINC.
There is no reason for it to be like that though.
Does anyone know how to change this parameter?

gdf

I don't think you can change it from this end. I think it is tied the the whole Work Fetch (Scheduler, GPU and LTD) minefield...and must be resolved by the Boinc Team.

---

Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

I’m experiencing the same problem since I installed a 2nd 9600GT.
Boinc is running fine (E4600 Due + 2 x 9600GT) 2 x CPU & 2 x GPU tasks.
Tried Boinc 6.69 but I get the same 24 hours back off after a task completes.
Only get 1 new task after a manual update even with cache set at 10 days
Need to babysit project to get it to run 24/7

Just installed Boinc 6.6.10 - Will wait & see what happens when a task completes

---

Did you try 6.5.0? That one should request enough work for both GPUs.

MrS

---

Scanning for our furry friends since Jan 2002

Did you try 6.5.0? That one should request enough work for both GPUs.

MrS

Not yet. Will try it and report back

Did you try 6.5.0? That one should request enough work for both GPUs.

MrS

Not yet. Will try it and report back

Thanks! Looks like 6.5.0 fixed the 24 hr delay

Thanks! Looks like 6.5.0 fixed the 24 hr delay

No, that is a server function. But 6.5.0 is better behaved. If you get the right kind of "miss" from the server, you will still see the 24 hour Back off ...

IN GENERAL, though, 6.5.0 is less likely to get hits on that ... but it is not guaranteed ...

I sent some notes to the project in a PM, not sure if they will help ... I hope they will ... with luck they will ...

Thanks to Paul's suggestion, now the server should force a WU request every 3 hours maximum.
Let see if this improves it.

gdf

Thanks to Paul's suggestion, now the server should force a WU request every 3 hours maximum.
Let see if this improve it.

Yea! Thank you GDF ... we love you ...

Thanks to all for putting the effort in. My gpu should be happier now ;)

Sounds like a good fix..except...I guess it depends on which client you are running (stable Linux-64....6.4.5)...

Mon 23 Feb 2009 04:54:04 PM EST|GPUGRID|Sending scheduler request: To fetch work. Requesting 53678 seconds of work, reporting 0 completed tasks
Mon 23 Feb 2009 04:54:14 PM EST|GPUGRID|Scheduler request completed: got 0 new tasks
Mon 23 Feb 2009 04:54:14 PM EST|GPUGRID|Message from server: No work sent
Mon 23 Feb 2009 04:54:14 PM EST|GPUGRID|Message from server: (reached per-CPU limit of 1 tasks)
Mon 23 Feb 2009 04:54:14 PM EST|GPUGRID|Message from server: (Project has no jobs available)

Mon 23 Feb 2009 04:58:17 PM EST|GPUGRID|Sending scheduler request: To fetch work. Requesting 56520 seconds of work, reporting 0 completed tasks
Mon 23 Feb 2009 04:58:22 PM EST|GPUGRID|Scheduler request completed: got 0 new tasks
Mon 23 Feb 2009 04:58:22 PM EST|GPUGRID|Message from server: No work sent
Mon 23 Feb 2009 04:58:22 PM EST|GPUGRID|Message from server: (reached per-CPU limit of 1 tasks)
Mon 23 Feb 2009 04:58:22 PM EST|GPUGRID|Message from server: (Project has no jobs available)

This is happening every 4 minutes...

EDIT: the timing of this is very sporatic. was just every 4 minutes. earlier it was about 14 minutes. It goes for a while, stops, then starts up with another time interval...very funny.

---

Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

Not sure what you want to say, really. I'm seeing that you reached the maximum allowed queue size, which has nothing to do with the 24h-back-off-fix.

MrS

---

Scanning for our furry friends since Jan 2002