Documentation about interoperating between Xavier A and Xavier B

Please provide the following info:
Hardware Platform: [DRIVE AGX Pegasus™ Developer Kit]
Software Version: [DRIVE Software 10]
Host Machine Version: [native Ubuntu 18.04]
SDK Manager Version: []

Dual Xavier could be powerful, but I can’t find documentation on how to optimally use both concurrently. There is a IPC sockets sample, but it doesn’t explain performance between SOCs. Any tips on best practices for optimising load distribution between CPU A/B and GPU A/B?

Hi @gpu_power,

That’s flexible and highly depends on your application and use cases. But “1.4. DRIVE Software”. has the information of how DRIVE Software 10 uses it. FYI.

Hi @VickNV
Thanks for pointing to which DRIVE components are installed on Xavier A and Xavier B.
Is there any more information about developing across the two processors? Specifically, is there shared memory between the two? Can both Xaviers access both dGPUs? Example of CUDA across multiple iGPUs and dGPUs would be useful.

NvSciIpc may be what you’re looking for. Please take a look at “Inter-Process Communication” documentation. Thanks!