i just got a new laptop with a new graphic card and it has a CUDA card with the Geforce G102M GPU and 512 mb ram. i have an Asus K50IN-X4-BIL however on the website it's just a K50IN. I updated all the drivers from Nvidia.

i just got a new laptop with a new graphic card and it has a CUDA card with the Geforce G102M GPU and 512 mb ram. i have an Asus K50IN-X4-BIL however on the website it's just a K50IN. I updated all the drivers from Nvidia.

29-Aug-2009 05:55:48 [---] Local time is UTC -7 hours
29-Aug-2009 05:55:50 [---] Can't load library nvcuda.dll
29-Aug-2009 05:55:50 [---] ATI GPU 0: RV770 (CAL version , 512MB, 1232GFLOPS)

29-Aug-2009 02:17:15 [---] No coprocessors

(duplicate)

According to what I've seen on some other BOINC projects, some of the later 6.6.* BOINC versions default to having GPU use turned off, if you haven't made any effort to tell those versions to turn it on. That's what I found with 64-bit BOINC 6.6.36 for Windows and at least one other 6.6.* version for Windows.

Also, the G105M on my laptop is too low-end a version to use under GPUGRID. If you're using a version of BOINC recent enough to show the GPU type, and it shows 1.1 for the compute capability, don't expect very many BOINC projects to be able to use it. Collatz Conjecture and possibly GPU-SETI appears to be the only ones generating enough workunits it can run to keep it busy. In case someone suggests Milkyway@home, I've already found that their GPU application requires compute capability 1.3 in order to get some double precision capability that project requires. GPUGRID appears to be satisfied with either compute capability 1.2 or 1.3, but not some of the lower end cards with compute capability 1.2.

MW requires double precision and only the 200 plus cards have that ... though at this time you are better off if you want to run MW to get an ATI card as the GTX260 (for example) takes 3:30 to run one task for MW and the slightly less expensive (last I looked) HD4870 does them in 0:55 seconds ... (well, unless you have two cards in the system and the other card is running Collatz which is likely a bug ... though I have not figured out why)

MW requires double precision and only the 200 plus cards have that ... though at this time you are better off if you want to run MW to get an ATI card as the GTX260 (for example) takes 3:30 to run one task for MW and the slightly less expensive (last I looked) HD4870 does them in 0:55 seconds ... (well, unless you have two cards in the system and the other card is running Collatz which is likely a bug ... though I have not figured out why)

MW requires double precision and only the 200 plus cards have that ... though at this time you are better off if you want to run MW to get an ATI card as the GTX260 (for example) takes 3:30 to run one task for MW and the slightly less expensive (last I looked) HD4870 does them in 0:55 seconds ... (well, unless you have two cards in the system and the other card is running Collatz which is likely a bug ... though I have not figured out why)

Paul,

You may want to check for signs that the MW and Collatz workunits actually run one after the other on the same video card, instead of on separate cards.