Hi, I have a Windows 7 laptop with a GeForce GT 330M which is accepting and processing GPUGRID work, and earning work recognition. It takes about 60 hours to process a ACEMD2 molecular dynamics 6.13 task. I have seen GPUGRID recognise that my GPU does not support double precision math. Am I wasting my or GPUGRID's time, or should I keep crunching?

GeForce GT 330M according to my NVIDIA control panel
Driver 260.99
CUDA cores: 48
Graphics clock: 475 MHz
Processor clock: 1045MHz
Memory clock: 700 MHz (1400 MHz data rate)
Memory interface: 128-bit
Total available graphics memory: 2296 MB
Dedicated video memory: 1024 MB DDR3
Shared system memory: 1272 MB
Video BIOS version: 70.16.4C.00.0A
Bus: PCI Express x16

Thanks
Adrian

...and I should have added this from the BOINC messages:

NVIDIA GPU 0: GeForce GT 330M (driver version 26099, CUDA version 3020, compute capability 1.2, 987MB, 100 GFLOPS peak)

Cheers
Adrian

It’s really touch and go with that card, and that’s if you run it 24/7. As it is a mobile we generally do not recommend running 24/7, but it’s your card and your decision.

With only 48shaders it would be barely capable of completing a task within 2days, running 24/7. As the shader & GPU clocks are quite low, and because you are using Win7 the tasks are presently taking a bit over 48h:

It takes about 58h to finish a KASHIF_HIVPR task and 49h to finish a GIANNI_DHFR1000 task. I would be happy enough with the 49h for the GIANNI task, but not the 58h for the KASHIF task, and these are the two fastest tasks presently being run.

Three questions,
Are you using swan_sync=0?
Have you set your CPU usage to free up one core?
Are you reporting tasks immediately?

These are my recommended optimizations for your card.
You can read about them in the FAQ: Best configurations for GPUGRID thread.

Thanks for the feedback, I have changed cc_config.xml as recommended and I will set swan_sync to 0 and see if it improves my performance.

Cheers
Adrian