Cranking comes at an efficiency price [the more CEP2 running concurrently on one device, the higher the difference becomes between wallclock time and actual CPU cycles used, an issue to many] in addition to a temp issue, not here yet. This efficiency deteriorates exponentially [mostly HD i/o driven], but if that is not an issue for you, go for it. Quite long discussions have gone before on the forum what works... my setup now running 96%+ efficient, but that's only allowing for 2 to run CEP2 out of 4 cores.

--//--

Concerning the custom project settings for CEP2, here's my take on it:

Using a quad core for work @ home with lots of Photoshop and Dreamweaver plus some surfing/gaming in the evening, I've set the machine as follows:

Max number of cores in use to 75% (3 cores), run when computer is in use and while processor use is less than 20%.

That way I get a zippy computer because of the unused core and if I play games, the other three cores become available. It pays more to do it this way than shutting down all 4 cores when in use simply because the computer is in use too often.

My 2 cents

Question - If I configure my home profile online for unlimited CEP2 tasks, stock up on CEP2 tasks, say 3days worth, and then reconfigure my home profile to only crunch 3 CEP2 tasks at a time, will this ensure I am always running 3 CEP2 tasks?

Why? If I configure 3 tasks, my daily turnover is usually well under that (even with the difference between run time and CPU time considered), and if I set it to 4 tasks I find the time delta increases too much, and system performance can become sluggish at times.

My guess is that the number of tasks only controls what is sent, and not what is actually run, (WCG being treated by Boinc as one project) but I might have missed something. I don't want to download 3days of tasks and find I have to micromanage what tasks are running.

On exclusive WCG this works, very closely to objective:

- Set 10 CEP2
- Set anything else [I do Clean Water]
- Set cache to 0.9 days.

WCG generally pushes enough CEP2, prioritized, that the mix will lead to mostly CEP2 running on 3 threads and when 4 seen it's a brief manipulation of suspend/resume. This is on a mean completion time that my quad is doing for CEP2... 6.51 hours. Still getting a 99% efficiency on that with CEP2 and 99.5+% for Clean Water

Experimenting is the name of the game to find the best balance. With the above mix, the mean "Time Out" is about 28.3 hours.

--//--

Just as an "FYI" for new people contemplating dedicating crunching boxes:

What I did (for summertime use, where I am more concerned about processor Tj temps than anything else) to ensure I always left one "core" (that is, on hyperthreading CPUs one physical or virtual "core") free was to:

BOINC manager/Advanced Menu/Preferences/Processor Usage tab:

Since it is CEP2, "Use at most 100% of processor time"

Then to control # of work units running simultaneously:

"On multiprocessor systems, use at most (Ct - 1) / Ct) % of processors"

where Ct = total (physical + virtual) cores

So on a hyperthreading i7 950 with 4 physical cores + 4 virtual cores, I set % processor usage at ((8 - 1) / 8) x 100 = 87.5%.

On a hyperthreading i7 980x with 6 physical cores + 6 virtual cores, I set % processor usage at ((12 - 1) / 12) x 100 = 91.67%.

That has resulted in my PCs consistently running (Ct - 1) work units; e.g. 7 work units running simultaneously on an i7 950 and 11 work units running simultaneously on an i7 980x.

Bearing in mind that I have an "always on" internet connection throttled at 3 Mbps upload/15 Mbps download, on the Preferences/Network Usage tab, I set the "Additional Work Buffer" to 0.6 days for all dedicated crunching machines. This has consistently resulted in roughly (Ct + 1) work units "ready to run" in queue locally.

Total CPU loading does indeed track the percentage I specify (wtg, WCG/BOINC!!) - and, importantly (to me), in an ambient temperature of 23.3 degrees Celsius/74 degrees Fahrenheit, my Tj temps average 60 degrees Celsius/140 degrees Fahrenheit, well within the Intel specs.

Now during the winter I tend to run all cores flat out because a) the ambient temperature can be reduced without costing me any money and b) the idea of heating my office and beyond with computers that are searching for alternative energy sources tickles me.

I hope that information may prove to be of use to someone new....heck, to tens of thousand sof new crunchers each and every day (one can always hope).