I don't now if there are technical ways to put it in place, but it is a creative idea to pursue in WCG and BOINC too.

It is a great question bring by Shinobi. I don't have sure but read in someplace that when we are running Vista (for example) we can determine how much % from each processor will be used for each processing task.

This kind of feature in WCG, if driven in right way and with clear criterias, for sure could help a lot WCG in Beta tests urgencies returns results and from other side will help a lot us crunch monsters WUs that many times we receive... Think about and share it with techs for yours next breakthrough sessions of new ideas and new features or something like that you use to present news on the table...

Completely right Lawrence. That is why Distributed Computing projects like WCG taken the other approach to parallel processing: Breaking the work up in small work units which can be processed in parallel independently of eachother.

Note that there is no real benefit for distributed projects to implement such SMP solution, as the goal of the project is do get all WU's done as soon as possible. From that perspective doing one work unit in 2 hours on a dual core computer is not better than doing two work units in 4 hours.

Using it only for beta would be quite useless aswell, as the whole purpose of beta testing is to reproduce the real situation as closely as possible. Using a different calculation path for beta testing would be a bit silly then eh?

Thanks for the input Lawrence but I don't think in SMP only (that is old, from early mainframe era, well knew and limited way to do it...by itself),

I think that we need explore new ways beyond GPU's like and combining it with some SMP/Grid techniques to explore really new kinds of virtualuization with softwares to get something different...

CUDA approaches is one good way to work in near future and that use SMP techniques (but need a lot of improvements).

Another new approach that are under studies for a select scientits club is change the rules of computation from binary to another ones. There are many simulations over codes based on hexadecimals (well knew for many people), over exponencial logs and over some kind of fractals equation based on fixed basis, and so on... (thousands of ideas to try insert quatum computation as quickly as will be possible due the high lebvel of output that this kind of solutuion will provide).

But for pratical uses and based on original thread post here, I guess that a simple programming change in BOINC rules could driven new features that allow do something like that in Linux approaches and if the techs get change Microsoft opinion about it bad windows code, in Windows versions too...use the feature existing in Windows...

It is my guess...

Do you know something I don't, GIBA?