Set to 2 and now have 2 in the buffer. I hope to keep one crunching per quad, for now.

Thank you techs/admin for going the extra mile and answering our pleas for help.

First, thanks for the quick fix to allow one to select the number of threads to run. This will help a lot of people run this project. Thanks also for putting up with all my questions.
@cleanenergy:
I have completed a lot of experimenting trying to efficiently run this project. I hope cleanenergy will give me credit (or 7 WCG dog year points ) for trying to run this project but it just does not work for the configuration of my machines. Some of these problems are due to the fact I am running 10-12 GPU’s at any given time.

1. Due to increases in the amount of CPU the GPU’s are using, I cannot complete the wu’s in 12 hours. Some GPUGrid wu’s now use 31% CPU/GPU; up from 16-18%. I watched one CEP2 job exit this morning with only 9 minutes remaining but none are completing before 12 hours. I’m not sure why a 12 hour cutoff is mandatory for this project but it is a deal breaker for me.
2. Clock time vs. CPU time is very bad. For example 11:11:38 (9:48:12). Very inefficient. This is not a GPU issue as my Dengue’s on the same machines are getting times like 8:34:12 (8:29:14)
3. Bandwidth collisions are still a big problem for me. I think you said you are not throttled on your end but, in round numbers, it takes just ~80% of the time for me to up load a 26 meg file to GPUGrid as it does to upload a file of the same size to Harvard. I tracked 6 pairs of these to validate this.
4. Checkpointing is insane and virtually useless. I am looking at some now; 9 checkpoints in 4 hours, 3 checkpoints in 7 hours and one with 1 checkpoint in 11 hours and that single checkpoint only happened 1 ½ hours ago. I lost a lot of time Wed due to the required Windows Updates restarts. Another deal breaker.

As I posted before, I think the science is good and certainly worth pursuing but I will leave it up to other crunchers to do this project.

Thanks, I will take a look at that when I get through my task/social engagement list. As most men know, there are some times when the only acceptable response is, "Yes Dear I would love to do that for you right now" or she gives you "The Look". Most men fear "The Look".
Thanks again.

Thanks to whoever implemented user selectable for the number of wu's for this project to crunch. I think the best thing is to run 1 at a time. I was getting around 20 minutes difference when run with other projects. When I switched to all cores running cep2, it increased the time to over 40 minutes. This is on a raptor drive. Will check differences when my ssd's arrive.
Sek, thanks for the tip. I think this will make a big difference. As for the upload issues, unless your on a fibre optic, were stuck dealing with network clog. Best solution I came up with is restrict uploads to while your sleeping only.