I got 4 new WUs sitting on my puter, all with quite wrong duration estimates (far too short) and a very short deadline: 29 Jul 2008 18:36:19 UTC +/- an hour. They will probably take quite some time into august to complete, should I rather kill them prematurely or wait how far I get?

One more obstacle is this strange messages I receive from your server:

Sa 26 Jul 2008 14:18:32 CEST|PS3GRID|Sending scheduler request: To fetch work. Requesting 1470 seconds of work, reporting 0 completed tasks
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Scheduler request completed: got 0 new tasks
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Message from server: No work sent
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Message from server: (reached per-CPU limit of 1 tasks)

I got 4 new WUs sitting on my puter, all with quite wrong duration estimates (far too short) and a very short deadline: 29 Jul 2008 18:36:19 UTC +/- an hour. They will probably take quite some time into august to complete, should I rather kill them prematurely or wait how far I get?

One more obstacle is this strange messages I receive from your server:

Sa 26 Jul 2008 14:18:32 CEST|PS3GRID|Sending scheduler request: To fetch work. Requesting 1470 seconds of work, reporting 0 completed tasks
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Scheduler request completed: got 0 new tasks
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Message from server: No work sent
Sa 26 Jul 2008 14:18:37 CEST|PS3GRID|Message from server: (reached per-CPU limit of 1 tasks)

Can anyone enlighten me what the number of CPUs has to do with GPU crunching? I need one per GPU, and as I only have one those four will never run in parallel (or at least I think so).

I've set the project to NNW and will manage it by hand for the time being, it's alpha, so such things are expected and manual handling is one of the fun things in alpha ;)

You get 4 days to do each workunit.
The EST time is based on your CPU speed not on your GPU speed, guess that's a Boinc limitation at the moment.
And the project gives you 1 task per cpu/core so on a quad core you can end up with 4 at a time :(
That's ok if you have a card that can do 4 in 4 days, but if not, best to abort the ones you won't do.

What I done when I was running it on my quad (it took about 12 hours 40 minutes per workunit), was let it 2 at most and then set it to no new work. wait till it was near finishing the 2nd of the 2 workunits, and let it get more work.

If you have any workunits with the name FASTTEST in them, you can leave them, as they run fast, for me they took about 47 seconds to complete.

Yeah I agree the 1 task per host would be a better option, since there is no multi GPU app at the moment, and probably even fewer users who have multi GPU systems. As when I was running it on the quad if I got 4 workunits, Boinc would start to painc, with it thinking each one would take 38 hours :(

Surprised that boinc estimated it at 5 hours for the workunit.
For me it est them at 38 hours for the normal ones and 13 hours for the "shorties"

With the 2 workunits on my 8800GT the AVG time for the 2 of them are 11 hours 44 minutes 50 seconds, for the "normal" length ones, and 42 Seconds for the "shorties"
Of course depending on what type of Nvidia card you have your results are going to be different.

I don't think the shorties are of much or any importance project wise, I think they were just added to try to find out why a lot of multicore (mostly quads) were having problems with an error on a workunit just after 1 finnished.
____________

Down with the Kredit Kops!!!

Yeah I agree the 1 task per host would be a better option, since there is no multi GPU app at the moment, and probably even fewer users who have multi GPU systems. As when I was running it on the quad if I got 4 workunits, Boinc would start to painc, with it thinking each one would take 38 hours :(

Surprised that boinc estimated it at 5 hours for the workunit.
For me it est them at 38 hours for the normal ones and 13 hours for the "shorties"

With the 2 workunits on my 8800GT the AVG time for the 2 of them are 11 hours 44 minutes 50 seconds, for the "normal" length ones, and 42 Seconds for the "shorties"
Of course depending on what type of Nvidia card you have your results are going to be different.

I don't think the shorties are of much or any importance project wise, I think they were just added to try to find out why a lot of multicore (mostly quads) were having problems with an error on a workunit just after 1 finnished.

The first one took 50h, the current one is planning to take 50h, it's not that big sample, but it's coherent ;)