My display was done in by the last batch of WU's starting with 39312xxxx. This was with the new 181.20 whql driver. Some statistics (34 wu's):

5 good or probably good WU's
16 "success" that were obviously errors
13 computation error

So out of 34 wu's (39312xxxx) processed on my gfx280, there are 5 good ones.

The first 6 "computaton error" I checked the tasks info and sure enough the angle was small, about 0.01

The 16 "success" ones that were obvious error were all less then 20 sec duration and the task info indicated there was no room for results. This should have been declared an error.

The good WU's seem to be scattered about the bad one and it is not obvious if a bad WU, and the resulting snow, affects subsequent work units. All the WU's are reported received about the same time (obviously) and I am going with the serial number of the task id to determine the order they were processed. It would be nice if the project reported the elapsed time and gave a time stamp of completion. That was one can easily determine if a bad WU affected the remaining WU's or not.


system info: vista 64, intel, gfx280 http://setiathome.berkeley.edu/results.php?hostid=4711247

---

Looks likes you have the same problem with the Nvidia drivers 181.20

---

Q6600 @ 2.40GHz
Vista Ultimate x64 SP1
Mem: 4GB
8800GT 181.22
BOINC 6.6.3 Win64

ACTIVE PROJECTS