Posts by [AF>France>IDF>Paris]Kasur

1) Message boards : Number crunching : Ignore Opencl and force CUDA version (Message 8596) Posted 17 days ago by [AF>France>IDF>Paris]Kasur Post: Hi, Now that I'm running CUDA, I realize that, with a 3080 Ti in CUDA, I'm going to do less work than with an RX 6750 XT in OpenCL (it's really less powerful). I pass a unit in 300s. in Cuda. Another cruncher with a 3080 works a unit in 200s in OpenCL, my card passed a few units in OpenCL in 2 hours each with all drivers up to date. Mine : https://moowrap.net/result.php?resultid=209133416 line 13 : See my systeme at 32bits, (Windows and Boinc are 64), other 7950X in OpenCL have this. line 14 : Tag my GPU as a AMD iGPU (that I have one like this, but only Nvidia is checked in settings and it's the nvidia app launched) Since I'm the only one with this issue I'll reinstall Windows (And erased the traces of my old GPU).
2) Message boards : Number crunching : Ignore Opencl and force CUDA version (Message 8594) Posted 18 days ago by [AF>France>IDF>Paris]Kasur Post: Hi, While waiting for the project to allow you to choose the applications worked on. If your graphics card is much faster in CUDA but the project only sends you OpenCL. If Link's solutions scare you (thanks a lot anyway). You can use <no_opencl> in your cc_config. <cc_config> <options> <no_opencl>1</no_opencl> </options> </cc_config> https://boinc.berkeley.edu/wiki/Client_configuration This will disable OpenCL calculation and the project will send you Cuda.[/url]