Originally Posted by
vaughan
Aqua and QuantumFire are both rewarding with big points for CPU users.
Just a note on the Points o' Credit issue with Aqua; the Fokker_Planck wu's are much more reasonable for dual-core machines:
1) They pay well (points per compute minute)
Depending on your processor clock rate and whether you are running a 64-bit OS, the points per compute minute can be really excellent. As an example, my main machine is running Win XP on a 5400+ Black clocked @ 3.3Ghz and is NOT a totally dedicated tasker; it manages to score almost 7 points per compute minute with FP wu's. I'm lucky to get 2.6 ppcm out of it with the 32-bit Aqua MT wu's. My 'new' recycled 5600+ 2.8Ghz dedicated cruncher running 64-bit Kubuntu WALKS ALL OVER my main machine with almost 11 ppcm; even though it is running at a lower clock rate (I believe that 64-bit vs. 32-bit is the main component difference, although the 5600+ also has larger L2 cache @ 1MB/core)
2) They execute relatively quickly
Again on my main machine, an FP wu will usually take about 40 minutes to complete; on the cruncher that comes down to about 26 minutes. On the 2Ghz Celery [32-bit Single Core], run time goes up to a 'whopping' 190 minutes. As a reference point, an Aqua MT wu on my main machine can take something in the order of 4,800 minutes (80 Hours) to complete. The longer it takes, the more things can happen to disrupt the work (or worse, cause a complete scrub, with the work lost).
3) They don't tie up the machine
Aqua@home has had issues with checkpointing, suspension and restarting of wu's not behaving very well. Ditto with not playing well with others wrt sharing cores and releasing cores at switch times to other BOINC tasks. This pretty much means that to be successful running Aqua wu's, you should not be running other projects simultaneously. It also means, you should run Aqua wu's from start to completion uninterrupted. That said, the FP wu's don't tie up a machine for days at a time, which is usually a GOOD thing.
That's why I love 'em...