About 2.5 million of these jobs have been completed and seemingly a few of Kaspersky running hosts suffer from disk trashing on top, but first let us hear from l_mckeon if the Run Always setting makes a change. As I wrote in me previous post, all run perfectly fine, near 100% efficiency, and monitoring with Process Explorer rather low disk IO, kernel time a few seconds on a total run time of 2.5 to 6 hours depending on the device I'm looking at.

A few words from someone (myself) who is using extensively Kasperski Security kit 2010 and 2011. It is now a few years I ended up standardizing on Kasperski after having tested in the past Norton, Panda and a few others but not G Data. All have issues and from one year to another things may change.
Anyway to make it short I have Kasperski installed on my 15 machines.
All the machines run Windows 7 64 bit with Intel i7 processors. All machines that are dedicated to WCG only, have a minimum of 6GB RAM and fast disks up to 300 GB. Other machines have 12 GB RAM and up to 1TB fast disks. I do not use SSD disks yet. They are too expensive and not yet mature enough as a product yet in my opinion. Their limited write cycle time is also a concern.
All WCG projects on my machines had no issues except C4CW and HPF2. With C4CW I got on three machines only once an alert considering the WU as a dangerous program and asking me to decide if to accept it or not. Once accepted it was ok. On other machines with C4CW no alert from Kasperski at all.
I have now returned about 7'700 valid results. Frankly three alerts are really nothing to fuss about.

But there is another issue more puzzling that I have. Some of my machines when set to receive C4CW will generate only errors with zero cpu time. And there is no alert or whatsoever from Kasperski. And to be sure Kasperski has nothing to do with it I have put the whole BOINC directory off limits to Kasperski and still that makes no difference. And the most amazing is that we talk about completely identical machines.
So the problem does not lie with Kasperski and remains a mystery to me. What I did is that I tested each machine for C4CW and those generating no errors were set to run exclusively C4CW and the others were set for other projects. I did exactely the same for HPF2. Except that for HPF2 some machines made no errors at all and others about 40% of errors and 60% valid. In the case of C4CW it is all white or black. I mean all error or all valid.

God works in mysterious ways

Someone commented recently about silly diskfile formats giving silly performance v.v. CEP2 on Linux having 25-40% efficiency, so what is your disk formatted at... NTFS / FAT32 / FAT16? (Not a probable cause since I understand that other sciences have no issue)

I'm running W7-32 on this 4 year old laptop duo, VM set to fixed 1.5GB. Can't understand the rumbling unless there is massive swapping between RAM and disk, the RAM use setting of BOINC also a possible culprit. Mine is set to use 85% of 2GB ram when in use and 95% when idle (in screensave mode).

Here's a screen capture after 75% complete, 4:45 hours, all jobs taking about 6:15 for the present target02. When looking at this, it just is compared to some of the other WCG sciences highly benign in all departments... really do not notice it running even when set at 100% CPU time and USING the system.

Suggest you really check the exceptions in Kaspersky (do not check the data_dir), delete any there are and redo and report it with them. Maybe Kasperski is paranoia about the data_dir being inside the C:\Program Files\ structure. Certainly Vista and W7 are so, the 6.2.28 client putting the data in C:\ProgramData\BOINC.

Sekerob: My secondary posts were done with Kaspersky switched off. It was just C4CW running on a naked XP SP3 system.

I've changed back to HCMD2 as my main project. I may let C4CW run at night some times, when my AV can be switched off and C4CW's behaviour won't annoy me.

Thanks for your efforts.