Erm, political sensitivities aside, can I post here?

Don't necessarily agree with you there David. Since all of the jobs here are CPU hogs, CPU should be the primary metric. the Boinc scoring mechanism appears fairer.

(Lights blue touch paper and stands aside )

I'm with you as well JM and Arnold - Ooops, I mean Graham - Boinc fairly benchmarks your PC according to performance, whereas with UD there is a limit as to how far you can improve your benchmarks regardless of how big the CPU or how much RAM you install.

I think this arguement is neverending on both sides - the only answer to the fairness that would satisfy all users is to have only one agent available.

Personally I would go for Boinc, as its more flexible in allowing other projects.

Well thats my 2 cents worth anyway!

I'm not even getting involved with this one

We just need some correction factor to hopefully bring a bit of parity (like it really matters )

I would imagine it will go the boinc way with multicored processors becoming the norm.

UD has served us well I've had 8 years and 233 days worth of crunching out of it since the 7th April 2001 without flaw in the agent (ligandfit/think/rosetta/faah) not bad, not bad at all

Dave

Just for Clarity to make sure there is no misunderstanding right at the off

I was composing my post while Bill snuck his in there

Well, as I understand it, BOINC is multi-threaded, UD is single-threaded so that allows BOINC to process the same amount of work more efficiently on a device. That difference will increase I believe on your higher end machines with the dual cores and such. UD has upper limits on its benchmarks and BOINC uses a different set of benchmarks that do not have upper limits so UD could in fact be giving less points than deserved to those machine that hit any of its benchmarks' upper limits.

The WCG admins determined that BOINC credits are converted to WCG points by multiplying by 7. I haven't seen anything that I can specifically point to that says that is the wrong factor. I've also seen nothing that would indicate, much more importantly, that the WCG admins determined this conversion factor incorrectly. I am presuming that they know what they are doing due to the lack of evidence to the contrary

The hardest part is that, no matter how hard we try, we just can't make an apples-to-apples comparasion of UD and BOINC. They are just too different for that. You have to weigh the pluses and minuses of both and make your own decision. I considered using the number of WU's returned as a measure for comparasion but that can easily be biased by the level of complexity of the particular mix of WUs that you happen to get. I've settled on average points per hour, given my acceptance of the conversion factor above, and thus my individual preference for BOINC. The average points per hour is not driven by wanting to max individual points but as a way to measure that, for a given amount of real time, I am maximizing the amount of crunching that I do. With the latter, WCG benefits, not just me.