Thank you for your answer, I only did announce this concern because around 12 WUs with an average computing time of 15 hours each are affected. In comparison, DDDT grantes in average 16 points per hour with a 10% variation between claimed and granted points.
In the case of HCC, the variation is between 200% and 450%. This situtation is not really fitting with other projects results.
Good night

Yes, it is a problem. On machines with more than 4 cores there is a definate issue. I've been looking at this closely for weeks now and there is a definate pattern that just effects these machines.
I have 2-8 core clovertown machines here and one Q6600.
The clovers run at over 3000mhz and the Q6600 runs at 3600.
The Q6600 always gets what it claims or close. The clovers do not.
One clover is on win2K3 64 bit running Boinc 5.10.20 64 bit and the other clover is on win2K3 32 bit running boinc 5.10.20 32 bit.
The Q6600 is on winXP pro 32 bit.
You have, or should have, the ability to look at my account.
See for yourself the claimed vs granted on these machines.
The clover on win2K3 64bit is in meshmesh's name but here at my home.
Look at the claimed vs granted on that machine since the HCC project started. It went down the toilet and fast.
That machine does absolutely nothing, I repeat, nothing but WCG and with 2 exceptions since last December runs 24/7.It is dead stable and the machines average should be pretty static but it is not.
There is an unidentified problem inside the HCC WU regarding these machines.

Hello,
the posted answers are interesting because they meet my own observation. The impacted host is a Dual Xeon Quadcore running W-XP Pro 64 with 4 GB RAM. The system is not overclocked.
Based on the last answers, I checked CPU load and memory usage.
- As expected, the CPU load is at 95%.
- Surprinsingly, the memory usage is only 800 MB with 8 crunching cores (I think that it is probably very low).
If, from time to time, I take a look on my crunching systems, it is not especially for some competition regarding granted points but for good practice regarding performance monitoring. I assume that the granted points reflect activity and performance of a specific system for a specific project.
I consider a variation less than 25% between claimed and granted points as acceptable (because the effective point allocation is not an "exact science"). Anyway, variations over 25% - or in my specific case until 450% - are, IMHO, the signal that something is going wrong. Only communication and dialog can help to identify problems and finally to solve them.
Based on the other made observations, I will activated HCC on single and dual core hosts and deactivated this specific project for the Xeon host and privileged DDDT, HPF2, and AfricanClimate.
If some particular issues should be closer investigated, please let me know in order for me to provide more accurate and consistent feedback.
Have a great week,

Hi,
I confirm my observation:
- with 8 cores crunching for HCC on W-XP Pro 64: around 800 MB RAM used
- with 3 cores crunching for HCC and 5 cores for others projects (DDDT, HPF2) on the same system: 1.55 GB RAM used.
I conclude that HCC should have a memory allocation failure. It will also probably explain why such WUs take so long for a poor result (only few granted points).
Within some days (since I selected HCC), the average earning drop down from around 2800 points daily to less than 2300 points, although this host is only crunching at this time.