The server should be now ignoring dcf for the time being.
I will reuse it in a couple of weeks when things are stabilized.
Let me know if the situation improves.

gdf

OK

It has started to drop from the 100.

Last check shows DCF at 99.011805. I will leave my host 6133 untouched, so as it can be monitored to see if it corrects back to near 1. Although it will be slow as it takes 16 hours per task. I will check again before I leave on vacation, in about 5 days, then it has 10 days to run untouched.

Followup. DCF still dropping.

Went to 98 and now down to:
Task duration correction factor 85.950111

To completion dropping too, it was some 785 hours maybe when DCF was 98 to 100,

Last task started with 128 hours to completion estimate.

I still do not get another task when one finishes, only way is to suspend other running projects which will force the schedule to request work, then I can resume the other projects. I assume once the DCF drops lower this behavior will stop.

---

Alpha Tester ~~ BOINCin since 10-Apr-2004 (2.28) ~~~ Join team USA

I still do not get another task when one finishes, only way is to suspend other running projects which will force the schedule to request work, then I can resume the other projects. I assume once the DCF drops lower this behavior will stop.

I manually reset the project after I ran out of work recently. This changed the DCF back to 1, but I still had one or two "High Priority" tasks and had to suspend other project to get work. After those two finished, I now have normal priority GPUGrid WU's processing and it keeps 4 WU's in my queue (2 processing on my 2 8800GT's and 2 waiting). So, I would expect it would do the same for you, eventually.

The DCF is again goof up on my PC. After this WU 161943 the next WU 162890 have again to run with a destroyed DCF of 60.830372 and a estimated runtime of over 80 hours. I'm really one of the hardest while testing software, but the actual situation is annoying. Looks like it's the wrong way to play with the DCF to control the behavior of he GPU work priority.

---

Anderson says that resetting project will reset the dcf.

In the meanwhile, I have returned to previous values, with factor two corrections applied. I will slowly move the server to the correct values.

gdf

The DCF is again goof up on my PC.

My DCF is ~0.7 after going back to 1.00 after a manual reset. However, my estimate times for GPUGrid WU's are back in the 100's of hours (starting at about 200 hours). This is after they went back down to approx. normal.

I have several other non-GPU WU's crunching, with High Priority as well. Not sure if the DCF is completely to blame. I now have 6 tasks running (5 in High Priority, 1 of those in GPUGrid, with another GPUGrid not running High Priority).

I know I could set the processor usage to 99% to get down to at least 5 active WU's (both GPUGrid and not), but don't want to have to apply a fix that will need to be changes when everything returns to normal.

It appears that I always have 1 GPUGrid tasks running High Priority and the other not. It will not fetch new work. All this starting after I "upgraded" to client 6.4.5. I'll let it run, only manually suspending other tasks as necessary to get new work for GPUGrid. I'd prefer not to babysit Boinc, though.

My DCF on this host has jumped back up to 100.

I reset several Box's this morning, when the new Wu's Downloaded 2 of them showed 22:00:00 Hours to completion (which is still about 4 times what it should be) & 2 showed 279:00:00 Hours to completion. I'm about 1/2 thru the first Wu & when I checked my DCF in the client_state.xml file it showed only 1.000000 for a DCF Value.

So I take it the DCF Value in the .xml file won't change until after the first Wu is finished, also if the Value was reset to 1.000000 why do 2 of the Wu's still show 279 Hours to completion when they Downloaded ... ???

hmm - I have been seeing this message quote:
"Message from server:(wont finish in time) BOINC runs 97.4% of time, computation enabled 100.0% of that" and cant get any WU's.

I went and reset the project and immediately gpugrid downloaded two tasks. Time to complete is just under 14 hours, about right.

Not sure that is going on and that error message does not make a lot of sense. I always have computation enabled 100% and BOINC runs all the time.

With the current update of the server everything should more or less work.
However, the server still cannot discriminate between an 8800 or GTX280, so the WU estimates are tuned on a 8800. Soon, the server will improve is estimate on the speed of your GPU.

gdf

hmm - I have been seeing this message quote:
"Message from server:(wont finish in time) BOINC runs 97.4% of time, computation enabled 100.0% of that" and cant get any WU's.

I went and reset the project and immediately gpugrid downloaded two tasks. Time to complete is just under 14 hours, about right.

Not sure that is going on and that error message does not make a lot of sense. I always have computation enabled 100% and BOINC runs all the time.

I have several Box's dead in the water at the moment because the Server won'r give them work, but it'll try in 24 hours again "rolls eyes" ... ;)

.......... I'm seeing some improvement - last couple of WUs came down with a 30hour ETA.
GDF - it would be nice if you could summarise the project's current position & advice to crunchers. In one place it would be good to have your thoughts on downgrading the client - or not, manually adjusting the DCF, hanging in there....etc.

P.

.......... I'm seeing some improvement - last couple of WUs came down with a 30hour ETA.
GDF - it would be nice if you could summarise the project's current position & advice to crunchers. In one place it would be good to have your thoughts on downgrading the client - or not, manually adjusting the DCF, hanging in there....etc.

P.

It doesn't seem to matter if you Downgrade the Client which I did on all my Box's last night, or if you manually adjust the DCF, or if you Re-Set the project, the Wu's still come with hap-hazardly To Completion Times.

I can Reset the Project & get 4 new Wu's & 1 will be at 5 hours 1 at 22 hours and maybe 2 at 279 hours. I'm still getting some as high as 27,000 hours even after a Reset of the Project.

BOINCView has been even messing up the last few days on me & I think it may be some something to do with the high To Completion Times, could be wrong on that though ...

I'm seeing some improvement - last couple of WUs came down with a 30hour ETA. - Burdett

None of my WU's (GPUGRid or otherwise) are running High Priority now. It's been a little see-saw over the past 48-72 hours, but everything appears normal again now. I have two WU's crunching and two pending, 1 with a 5 hr estimate and 1 with a 21 hour estimate. Now, I have 8800GT's and those estimates are both low and high, but certainly much better than the 1000's of hours I was seeing before.

It doesn't seem to matter if you Downgrade the Client which I did on all my Box's last night, or if you manually adjust the DCF, or if you Re-Set the project, the Wu's still come with hap-hazardly To Completion Times. - Poorboy

I agree. After my manual reset, I did see an improvement but it eventually went back to high estimated times. I didn't do anything at that point (besides suspending other tasks to get GPUGrid work) and it appears that everything is OK for the moment.

about time for a new highscore-thread?

latest one started with 1190 hours on an 260..

next one: 18300 hours - rising like mad..

My latest completed WU was back to the shorter ones: Last WU.

2435 credit vs. 3232 credit

Is there any correlation between this issue with high estimated times / high priority tasks and the smaller WU's? Once it reverted back to the "3232" WU's 2 days ago, my rig appear to stabilize and work correctly. That is until it started process the "2435" WU's again.

Which workunit name?

gdf

... I'm back to 'work not available for your type of computer' message - followed by the 24hour 'sod off' back off.
As it now appears we have to manage GPUGrid manually - could we remove (or double) the 1 WU per CPU limit - I can get through 8 WUs in 24 hours or so on one of my hosts ... if I could load up, the 24hours backoff could pass without idle time. Or just reduce the 24 backoff to something more useful - 6 hours?



P.

Which workunit name?

gdf

WU:124466 currently @ 39% showing 6725 hours remaining.

host-id is 16930

OK, Mine dropped again from 100 to 'Task duration correction factor 91.193244'

This will be my last post. I have some more last minute preparaions for vacation tonight, then I'm off tomorrow for 12 days. MErry Christmas to all, see you all later, much later.