I just started running the project a couple of days ago and managed to process the original 4 units without problems Tonight I got new units downloaded and the GPU started a unit. About 2 minutes later it loads a 2nd wu and begins crunching on it. The 1st unit went to "Waiting to run".

I aborted the one that was waiting and then another unit was loaded into the GPU and began running forcing the one that had been running to go to "Waiting to run".

Any ideas?

You could try BOINC 6.6.36. I know 6.5.0 isn't as switch-happy, you'd just have to set GPU-Grid ressource share and cache low enough so that it does not always download the maximum of 8 WUs.

MrS

---

Scanning for our furry friends since Jan 2002

I just started running the project a couple of days ago and managed to process the original 4 units without problems Tonight I got new units downloaded and the GPU started a unit. About 2 minutes later it loads a 2nd wu and begins crunching on it. The 1st unit went to "Waiting to run".

I aborted the one that was waiting and then another unit was loaded into the GPU and began running forcing the one that had been running to go to "Waiting to run".

Any ideas?

You don't need to abort it. If it swaps one out it will eventually swap it back in and continue crunching. Normally under 6.6.36 cuda tasks will run in the date/time downloaded order unless it doesn't think it has enough time, in which case it may run another task with an earlier deadline. This is commonly called EDF (earliest deadline first). It will also swap tasks out if you have more than one project using cuda, in order to honour resource shares.

I'd just leave it to crunch and let it work things out. What sort of graphics card do you have?

---

BOINC blog

I'm running a GTS 250 for the GPU. My BOINC version is 6.6.31 and it has always performed as you stated. FIFO unless it panics and goes into "high priority" mode.

That is what surprised me, the GPU workunits all had the same deadline since they had just been downloaded.