Need help for real world deployment

We intend to use Nvidia GPU for real world deployment of intelligent video analytics. Which is the best mid range GPU which is good for inference , as we don’t have budget to buy the likes of tesla T4 , we plan to use RTX2080 TI , or jetson Xavier developer kit. Is it recommended to use these products for daily runtime i.e 24X7 365 days in the first place ?
We want to process 80-100 streams at 10 FPS

Hi hrsht.sarma,

Please refer to DeepStream SDK web page for the performance evaluation on different platforms as reference:


For long time usage, the RTX2080 TI and Jetson Xavier should be reliable to meet your requirement under the well temperature, humidity controlled environment.

Thanks

1 Like

Is there any script which is available to automatically shut down the system once temperature reaches certain threshold

Hi hrsht.sarma,

We have thermal design guide to provide the system-level thermal, mechanical and qualification requirements for the Jetson Series modules, you can find them from https://developer.nvidia.com/embedded/downloads

1 Like

Will it be possible for you to give recommendations for graphic card which is specifically good for inference, Can’t afford tesla p at the moment

*bump

wanted to know which GPU is preferred for inference

Hi hrsht.sarma,

I can’t give you the exact answer for which card is specifically good for inference due to this is the performance and budget question. You need consider your product outlook, performance and budget to find the suitable one for your project.