Taeric

It's pretty common knowledge that one instance of F@H can be run on each CPU core, either physical (dual core) or virtual (Hyperthreading). There are FAQs on how to accomplish this by Stainless and intelc4004.

What may not be as commonly known is how this affects the completion of work units and affects the project as a whole.

Hyperthreading CPUs

Running two instances of F@H will cause both WUs will be completed more slowly than running them one at a time, but this will generate approximately 15-30% more points than running a single instance. On the surface, this sounds great, but this is not as beneficial to the overall folding effort as it sounds on the surface. Projects are released in generations, and all of the WUs in one generation must be completed prior to the start of the next generation. By slowing down individual work units, it slows the completion of one generation and progression on to the next generation. Generations can number in the hundreds, so a delay of even a few hours can cause significant delays in the completion of a project. So, even though you're cranking out more points, you may actually be slowing things down.

Dual Core CPUs

Since each core is a physical entity, no slowdown is experienced by running multiple F@H instances.

Conclusion

Stanford's philosophy is that they are perfectly happy accepting any and all contributions to the overall effort. On hyperthreading CPUs, running either one instance or two are much more helpful to the cause than not folding at all, however, running only one instance will make the projects progress more quickly even if your points don't accumulate as quickly as running two instances.

For more information, please read this information on the Folding@Home Support Forum. Additional information was obtained through a PM exchange with the author (who enthusiastically supported sharing the information here).

13 3 @ 7 l 3 13 0 y

good to know

__________________

11,009 in 3DMark06! on a single 8800GTS320 @ 660/2000
Can i search a specific site in google? Is there a title associated with my reps? What are fragmentations/how do they occur?

thehybridpyro

good to know man
what about HT dual cores?

Taeric

The same principles apply - if you run 1 per physical core, everything is great; if you run 1 per logical core (2 per physical core), it can slow the research down. So, 2 instances on a dual core CPU is optimal and 4 will produce about 30% more points but at a potential cost to the project as a whole.