Here is a quick guide for ATI/AMD GPU users to get SETI crunching on your GPUs.
![smile.gif]()
Also, after your first few GPU workunits have been validated, don't forget to update the GPU Credits Database HERE so others can get an idea of time per workunit and estimated PPD
![wink.gif]()
For anyone else who wanted to crunch GPU only and never quite got it working, here is my app_info.xml file for reference:
- Stop and close your BOINC client (make sure it is stopped and closed properly, not just minimised)
- Sign up to the SETI project HERE and set your SETI your SETI project preferences as follows:
Code:
Code:Resource share = 100 Use CPU = no Use NVIDIA GPU = yes Maximum CPU % for graphics = 5 Run only the selected applications [email protected] Enhanced = yes Astropulse v5 = yes Astropulse v5.05 = yes Accept work from other applications = yes - Sign in to boincstats and click the "host list" heading in the left menu
- under "host name", click the host that you want to run the SETI project on
- Under that host name, put a check in ONLY the "attach" and "do not use no cpu" boxes and then click "set"
- Open your BOINC client and it will now download the initial SETI project files
- After the SETI project files have finished downloading, no SETI tasks should be downloaded and no workunits should start to crunch - if they are downloaded and crunching, you have not set the SETI project preferences properly and it is downloading CPU tasks - go fix it and set the SETI project preferences as per step 2.
- Stop and close your BOINC client (again, make sure it is stopped and closed properly, not just minimised)
- Now it is time to set up the GPU apps - go download the r177 multibeam GPU only app files from HERE and also download the r516 astropulse GPU only app files from HERE
- Once downloaded, extract both archives and copy the files into your SETI project folder:
ie. C:\ProgramData\BOINC\projects\setiathome.berkeley.edu - Now create a new text file and name it app_info.xml (it has to be a .xml file). Note: if you have "hide extensions" on in Windows, it will end up being a .txt, so disable that so you cans ee the full file name and make sure it ends with .xml or it won't work.
- In youe app_info.xml file, add the following:
Code:
Code:<app_info> <app> <name>setiathome_enhanced</name> </app> <file_info> <name>MB_6.10_win_SSE3_ATI_HD5_r177.exe</name> <executable/> </file_info> <file_info> <name>MultiBeam_Kernels.cl</name> <executable/> </file_info> <app_version> <app_name>setiathome_enhanced</app_name> <version_num>610</version_num> <platform>windows_intelx86</platform> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>0.05</max_ncpus> <plan_class>ati13ati</plan_class> <cmdline>-period_iterations_num 1 -instances_per_device 1</cmdline> <flops>20987654321</flops> <file_ref> <file_name>MB_6.10_win_SSE3_ATI_HD5_r177.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>MultiBeam_Kernels.cl</file_name> <copy_file/> </file_ref> <coproc> <type>ATI</type> <count>1</count> </coproc> </app_version> <app> <name>astropulse_v505</name> </app> <file_info> <name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r516.exe</name> <executable/> </file_info> <file_info> <name>AstroPulse_Kernels.cl</name> <executable/> </file_info> <app_version> <app_name>astropulse_v505</app_name> <version_num>506</version_num> <platform>windows_intelx86</platform> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>0.05</max_ncpus> <plan_class>ati13ati</plan_class> <cmdline>-instances_per_device 1 -hp -unroll 10 -ffa_block 4096 -ffa_block_fetch 2048</cmdline> <flops>30987654321</flops> <file_ref> <file_name>ap_5.06_win_x86_SSE2_OpenCL_ATI_r516.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>AstroPulse_Kernels.cl</file_name> <copy_file/> </file_ref> <coproc> <type>ATI</type> <count>1</count> </coproc> </app_version> </app_info> - Now save and close the app_info.xml file to the SETI project folder as well:
ie. C:\ProgramData\BOINC\projects\setiathome.berkeley.edu - Note 1: if you ARE NOT using HD58xx or HD69xx series GPUs, you need to edit the app_info.xml file and CHANGE the 2x instances of "MB_6.10_win_SSE3_ATI_HD5_r177.exe" to "MB_6.10_win_SSE3_ATI_r177.exe" otherwise your CPU usage be high and it may cause your GPU driver to continually restart, the BOINC client to crash or give errors when validating the workunit.
- Note 2: I have set "count" to 1 so it will crunch 1 workunit per GPU, and for me, that seems to offer the best estimated speed per instance and lowest CPU usage - YMMV. If you want it to crunch 2 workunits per GPU, set it to 0.5 and set "instances per device" to 2. If you want 4 workunits per GPU then set it to 0.25 and set "instances per device" to 4 - but keep in mind the more workunits per GPU the more CPU usage and depending on how powerful your CPU is, to many workunits per GPU could bottleneck your GPUs and impact GPU usage. Also, it was mentioned in the SETI forums that crunching more then 1 workunit per GPU can cause workunit computation and/or validation errors.
- Note 3: I initially had an issue with it using 100% CPU usage and only ~70% GPU usage, and I had to change "instances per device" to 1 and that dropped CPU usage down to 55% and increased GPU usage to ~94% - it seems that each workunit uses approx 25% CPU usage as well (I have 2 GPU's, so 25% per workunit x2 is ~50% CPU usage total) - again YMMV. You can try and reduce CPU usage further by increasing the "period_iterations_num" value, however it didn't do anything for me.
- Note 4: The Astropulse tasks are rare, so don't worry if it doesn't download any for days at a time. But the multibeam tasks are quite abundant, so you should start to download them within the first hour. If you don't have any downloaded within an hour, check your configuration and check your SETI project preferences are setup as per step 2.
Also, after your first few GPU workunits have been validated, don't forget to update the GPU Credits Database HERE so others can get an idea of time per workunit and estimated PPD
For anyone else who wanted to crunch GPU only and never quite got it working, here is my app_info.xml file for reference:
