Under advanced/preferences set the additional work buffer to 0(zero). Thsi should only allow one task per time. When the working task is almost finished it will download another. BTW what kind of machine are you using, what processor, what OS, which project and how long do you leave the machine on?
Cheers

Thanks, it seems to be working now. I have an Athlon 64 X2 5200+ running Windows XP. The GPU is onborad: Nvidia GeForce 6150SE (nForce 430), and it says "No usable GPU found" in the messages log.

The problem is I'm not leaving the PC on as I used to. That leaves me with the choice of either a) slowly and continuously compute even while in use (something I avoided); or b) full-throttle compute during the minutes I'm away from keyboard.

I'm trying strategy b and so far it seems to work. It would be good to be able to use both: while in use compute at 5%; when idle use 90% CPU.

I can't see my projects right now, but I can enumerate them later if you think it could be relevant.

Thanks!

Conectar cada = 0.0

The 0.3 days makes BOINC fetch 0.3 days of work into the buffer and then connect to WCG in intervals of 0.3 days. If there is though a permanent internet connection, this is ignored. If work is needed, the client will fetch more and then clear any completed work reports.

Anyway, your screenshot suggests that jobs are running fine. Hit Update button in the project tab, and the HPF2 should clear. The client scheduler would clear this anyhow on the next workfetch.

Ohhh now I get it... when it goes working it does both tasks at parallel. I'll try switching to 1 core only and I'll compare the results.

Thanks everyone for your help!