I started GPUgrid yesterday on two systems (overview computers), but a third system without any CUDA card was reported to have a "client (compute) error" (239257).

First:
It´s a G33-Board, not able to run one second of CUDA code, I´m quite sure that I had not started GPUgrid on that computer (20725), in former times yes, it was used for GPUgrid.

Second:
The reported BOINC version is wrong, I installed days ago 6.6.20, not 6.4.5. A further indication there´s something going wrong in the report or assignment of the host IDs.

Shall I check out this computer ? I thought it is sufficient to stop projects simply.

Perhaps you should proof the rules of distribution for the workunits to avoid double/redundant downloads.

Or it´s a matter of that 6.6.20 version.

 - exit code 98 (0x62)
</message>
<stderr_txt>
# Using CUDA device 0
# Device 0: "GeForce GTX 260"
#  Clock rate: 1500000 kilohertz
#  Total amount of global memory:                 939524096 bytes
#  Number of multiprocessors:                     27
#  Number of cores:                               216
MDIO ERROR: read error for file "input.vel", byte number 4: number of atoms (1625495040) != (39910) expected
ERROR: mdioload.cu, line 146: Unable to read binvelfile

The result is from January, so don't worry. The succesful WUs are purged from the DB at some point, but the errors stay far longer.

MrS

---

Scanning for our furry friends since Jan 2002