Folding Forum

I have dual CPU's, each CPU having six cores. I'm wondering if it will work faster with a single slot, using 12 cores, or if I should run two slots, each of them getting six cores.
I've not yet experimented, but on alternate days, I convince myself that I could get more points with two slots running, then convince myself that I should leave it with a single slot using all the cores. Several searches have failed to find anything I might consider "authoritative". I'm sure there are other users who have experimented - I'd love to hear their results!

In general you will get more points by running a single slot using all 12 cores. The Quick Return Bonus is exponential, so completing a WU in about half the time results in more than twice the points.

Also it is better for the science, earlier return of a WU allows generation and assignment of the next WU to be done sooner rather than later. Having two WU's out that take twice as long introduces delays.

Thank you, Joe_H - I'll refrain from fooling with configuration then.

The number of CPUs that can process each project must fall in a reasonable range. The majority of projects will run with 12 CPUs so I don't expect any problems, but when unusual things happen, it's possible that you won't always get a new assignment. If that ever happens, you can check with us and we MAY recommend that you temporarily reconfigure for two slots. (It's better to reconfigure so you can keep folding than to let your machine be idle.)

I don't expect that to happen, though.

You do not mention if you have a GPU. If you do, FAH may reserve 1 CPU to 'feed' the GPU. For your rig, that would leave 11 CPUs to fold. FAH dislikes 'large' Prime numbers so it may shift down to 10 CPUs. If it does so you, could set a slot with 10 and a slot with 1 CPU, avoiding a prime.

If you do not fold on a GPU, this is not an issue. Even if it is an issue, you may see no real gain in points with a 1 CPU slot.

JimboPalmer, I'm not running a GPU at present. I do have an Nvidia 630 that I plan on installing, but haven't done so yet. Reading about GPU's, it seems that Nvidia doesn't need or want a CPU to "govern" them. There are no ATI video cards here.

That information on CPU usage with GPU folding on nVidia cards is out of date wherever you read it. With the current GPU folding cores, Core_17, _18 and _21, one CPU core is used continuously with nVidia cards due to the way the nvidia drivers process OpenCL. The low usage of CPU by nVidia cards during folding applied with the older GPU folding cores such as Core_15.

Depending on which variant of the GT 630 you have, it should be able to fold WU's from the projects that use the newer OpenCL cores and complete them within the deadlines. The 630 is about the minimum to do so. Otherwise the folding slot can be configured to request the older Core_15 WU's to process. They are approaching end of life, but still are assigning work to finish up the projects.

Thank you, Joe_H - that is certainly useful information!

I believe that I'll get that GPU cranking this weekend. Sometimes, it's a pain to install video cards on Linux, sometimes not so much. Maybe I'll get lucky, and have it done in a quarter hour.

Well running on Linux will preclude configuring for Core_15 projects, so you will be only able to ask for assignments from the newer projects. Core_17 and later are supported on Linux.

Update:

The GPU was idle until two days ago. In my quest to enable the GPU, I screwed things up quite badly. Eventually, I gave up, and gave a thought to Gentoo. Sabayon, a derivative of Gentoo, advertises itself as having "native support for proprietary drivers". I installed, reinstalled the FAH application, and I'm cranking out both CPU and GPU work units. This morning, I passed up my team mate, whom I've been chasing for weeks.

Linux and video drivers are truly frustrating sometimes.

Linux and video drivers are truly frustrating sometimes.

That's an understatement.