Message boards : 
            Number crunching : 
        How to compel CUDA application ?
Message board moderation
    
| Author | Message | 
|---|---|
        
      JohnMDSend message Joined: 10 Mar 14 Posts: 7 Credit: 25,253,692 RAC: 0  | 
        
         
 "opencl_nvidia_101" WU's do not use my nvidia unit. How can I insist on CUDA WU's ?  | 
| 
        
     Send message Joined: 4 Mar 16 Posts: 2 Credit: 210,088 RAC: 0  | 
        
         
 All I found was this website on the topic.  https://boinc.berkeley.edu/wiki/GPU_computing  | 
| 
        
     Send message Joined: 22 Oct 11 Posts: 33 Credit: 481,684,431 RAC: 36,659  | 
        
         
 "opencl_nvidia_101" WU's do not use my nvidia unit. How can I insist on CUDA WU's ? What nvidia unit do you have? Does BOINC 'see' the nvidia unit upon boot (see BOINC-log)?   
 | 
| 
        
     Send message Joined: 22 Oct 11 Posts: 33 Credit: 481,684,431 RAC: 36,659  | 
        
         
 I see you have both a Intel UHD 620 GPU and a NVIDIA GeForce MX150, does BOINC see both? The following assumes your BOINC data directory is C:\ProgramData\BOINC. Step by step: 1. Start up Notepad 2. Paste the following code into your cc_config.xml file, located in C:\ProgramData\BOINC. If it doesn't exist, make one using notepad, save as .txt and alter the suffix to .xml (ignore the warnings). <cc_config>
    <options>
        <use_all_gpus>1</use_all_gpus>
    </options>
</cc_config>3. Save it into C:\ProgramData\BOINC as cc_config.xml (careful, Notepad likes to put .txt on the end of the filename) 4. Shut down your BOINC client if its running 5. Start BOINC up again You should have a couple of messages in the startup about config: and it should use both GPUs Are you using the standard driver, or the one downloaded from nVidia? For full CUDA performance use the last nVidia-supplied driver for your card.   
 | 
| 
        
     Send message Joined: 31 Mar 18 Posts: 11 Credit: 2,496,595,784 RAC: 623,669  | 
        
         
 Would be nice if they updated the CUDA app to use newer versions of CUDA too.  Would be much faster and use the GPU fully.  |