I was running CEP2 on Ubuntu 10.4 LTS and for a short time 10.10 RC, both 32 bit. This project was run on I7 with and without hyper threading, and on an Intel Q9400 (775 socket 2.66 quad) chip. I ran CEP2 by itself, and with other project WU's (for i7, hyper threading was both on and off) with no change in results. The OS was running on a hard drive solely dedicated for Linux and was not physically attached to the Windows 7 hard drive. I tried everything to make this work, including upgrading the OS to 10.10 and BOINC 6.10.58, installing a fresh clean hard drive, installing 10.4 64bit version, and I upgraded the processor from Q9400 to a more powerful CPU. As for RAM, 4 gigs is the norm for each box, and these were originally crunched with CEP2 WU's LAM setting off and later on, and no change in results occurred. Nothing has worked. Wall clock times are between 6-11 hours with CPU time between 1.7- under 3 hours. I have already posted these huge gaps and provide other examples, my favorite being CPU time greater than wall clock time. That occurred after the OS crashed, but please note that this OS crash did not cause the poor crunching results. The poor crunching results were already in effect before the crash.

The one common denominator the computers usually (except for one computer) did share before the WU's became problematic is the OS crashed. The results were never the same for CEP2, but for any other WU project, no problem. That is why I used another clean hard drive to see if the OS or BOINC client were corrupted. Did not work.

If anyone has other other ideas or wishes to try an experiment with CEP2 work units, I am open as I still have an extra hard drive. What I do find interesting is initially the CEP2 crunching works fine for about one hour, then at some point thereafter the CPU and wall clock times diverge.

Done. I will download and install Kubuntu 32bit version tomorrow night (very busy tonight). I will put this on a non-hyper threading computer and run CEP2. I will report back as the results occur.

Yup, that is the known point where this seems to happen. Wish we knew why. Is it because some BOINC are not quick enough to register the cumulative CPU time and pass that to the next job in a task. It's almost as if only the job setup time is being recorded.

Different versions or you mean different distribution

Every linux kernel has different effect on how it handle things - cpu and i/o scheduling.

Right now I have 3 test kernels on 3 rigs to determine whc one is better.
1) ubuntu 10.10 server, using 2.6.35.7 kernel (customize kernel - removed amd cpu features) and no x-window
2) ubuntu 10.10 server, using 2.6.35.7 kernel (stripped down kernel (9MB using localmodconfig) and no x-window
3) ubuntu 10.04.1 desktop, using default kernel and with x-window.

Now its waiting time to gather results whc one is better.

Update:

I successfully installed Kubuntu 10.10 32bit version and I am now running the following WU's:

E200454_ 062_ A.24.C20H13NS3.189.0.set1d06_ 1-- Catseye In Progress 10/15/10 01:43:37 10/25/10 01:43:37 0.00 0.0 / 0.0
E200454_ 101_ A.24.C18H13N3S2Si.178.1.set1d06_ 0-- Catseye In Progress 10/15/10 01:43:37 10/25/10 01:43:37 0.00 0.0 / 0.0
E200454_ 003_ A.24.C19H13NOS2Si.189.4.set1d06_ 0-- Catseye In Progress 10/15/10 01:23:00 10/25/10 01:23:00 0.00 0.0 / 0.0
HFCC_ n1_ 02420243_ n1_ 0001_ 0-- Catseye In Progress 10/15/10 01:43:54 10/25/10 01:43:54 0.00 0.0 / 0.0

I will report the results of this experiment tomorrow.

Edit: CPU time: 27:05 min, elapsed time 1:33:21. The WU was not far behind then the WU apparently reached checkpoint. I will run all CEP2 downloaded WU's, but this is not encouraging.
Please note the HFCC WU: CPU time: 1:36:02, elapsed time: 1:37:17.
All WU's running concurrently.