This thread has been locked.
If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.
Dear Champs,
Is there any benchmark data of TIDL on TDA2P and TDA2?
Do you think the TIDL performance of 2x EVE of TDA2P is similar with 4x EVEs of TDA2 because the CPU clock of TDA2P EVE is 40% higher than TDA2?
My customer validated their algorithm on 4x EVEs of TDA2 using TIDL and feel there is not much CPU resource remained. Now, because of internal ISP, they are thinking of TDA2P but are afraid if their algorithm can be run on 2 x EVE of TDA2P.
But, because the core clock of TDA2P EVE is higher than TDA2(900Mhz vs 650Mhz), I think the TIDL performance would be similar between 2x EVE of TDA2P and 4x EVE of TDA2.
Could you please let me know your idea on this total TIDL performance of TDA2P and TDA2?
Thanks and Best Regards,
SI.
Hi SI,
TIDL experts are looking at your questions, but in meantime you can check those threads for some details of TIDL performance evaluation:
TDA2EVM5777: Theoretical Performance Evaluation for Custom Deep Learning Model to run TDA2xx Platform...
TDA2EVM5777: Comparison between Sparse and Dense Model ? - Processors forum - Processors - TI E2E support...
Regards,
Yordan
Hi,
By default we are not validating TIDL_OD usecase on TDA2px so we do not have any performance number for this on TDA2px plaform.
We only run this usecase on TDA2xx platform and with TI's provided input files, you can get a FPS of 22-24 of Display.
As we did not run this on TDA2px so we can not comment anything related to its performance on TDA2px.
But its very easy to port the usecase on TDA2px and you can check the performance yourself.
Please use OPP_HIGH(Higher clock speed for cores) MLO for running this usecase for both platforms.
Regards,
Praveen