<html><body><div>Hi,<br><br>I would like to ask the community if anyone witnessed a difference between the coincidence rate when running the simulation on CPU and GPU? I have got a PET geometry with voxelized phantom and when using CPU, I am getting coincidence output file of size around 700 kb from 1M primaries. When running the same geometry on GPU with the same number of primaries, I am getting files of 2 MB, i.e. + roughly 185 % detected coincidences.<br>Can anyone tell me what can be the cause and how to fix it?<br><br>I am using those source definitions if that helps:<br>CPU:<br>/gate/source/addSource zdroj voxel<br>/gate/source/zdroj/reader/insert image<br>/gate/source/zdroj/imageReader/translator/insert linear<br>/gate/source/zdroj/imageReader/linearTranslator/setScale 1. Bq<br>/gate/source/zdroj/imageReader/readFile phantom_data/SPECT128.h33<br>/gate/source/zdroj/setType backtoback<br>/gate/source/zdroj/gps/particle gamma<br>/gate/source/zdroj/gps/energytype Mono<br>/gate/source/zdroj/gps/monoenergy 0.511 MeV<br>/gate/source/zdroj/setPosition -28.2590976 -28.2590976 -28.259072 cm<br>/gate/source/zdroj/gps/angtype iso<br><br>GPU:<br>/gate/source/addSource zdroj GPUEmisTomo<br>/gate/source/zdroj/attachPhantomTo pacient<br>/gate/source/zdroj/setGPUBufferSize 15000000<br>/gate/source/zdroj/setGPUDeviceID 1<br>/gate/source/zdroj/reader/insert image<br>/gate/source/zdroj/imageReader/translator/insert linear<br>/gate/source/zdroj/imageReader/linearTranslator/setScale 1. Bq<br>/gate/source/zdroj/imageReader/readFile phantom_data/SPECT128.h33<br>/gate/source/zdroj/setType backtoback<br>/gate/source/zdroj/gps/particle gamma<br>/gate/source/zdroj/gps/energytype Mono<br>/gate/source/zdroj/gps/monoenergy 0.511 MeV<br>/gate/source/zdroj/setPosition -28.2590976 -28.2590976 -28.259072 cm<br>/gate/source/zdroj/gps/angtype iso<br><br>Best regards,<br><br>Jiri Ters<br></div></body></html>