nvgpuswitch.py may misleading me, I don’t have a Clara AGX kit.
I setup A4500 to pcie 4.0 x8 on the side with extra power supply.
Now i am troubled in switch Xavier to dGPU mode, or does a dGPU mode exists? All i need is installing driver like in x86 platform?
This use case is not supported, the AGX Xavier Devkit platform do not support dGPU via PCIe.
hi, kayccc:
thanks for reply, do you mean AGX Xavier Devkit platform does not support dGPU via side PCIe slot? why? could a connector like ConnectX solve this problem?
That’s not a problem, but a feature which is not implemented on AGX Xavier Devkit platform.
Sure, that’s not a problem. In fact, I wan’t to confirm using a AGX Xavier industrial module with some kind of connector can support dGPU? Since we can’t use Pegasus or Clara platfrom for some reason.
apt update && apt install -y cuda-11-1 nvidia-container-runtime nvidia-l4t-* tensorrt python*-libnvinfer* graphsurgeon-tf onnx-graphsurgeon uff-converter-tf
orin support dGPU
hi, i found the dicuss about orin platform, which support dGPU. Does the only way to use my dGPU is turn to Orin?
Again, this is not supported, no matter with any kind of connector, the dGPU is not able to work with AGX Xaveir devkit.
That’s IGX which is not sold for end customers, and it supports A6000 GPU only.
ok, thanks a lot.
After check https://repo.download.nvidia.com/jetson/, I found no offical driver for dGPU. Is that the reason why AGX Xavier not support dGPU? If so, I still can turn to open-gpu-kernel for my project. Does this way still not make sense?
I don’t know, and not able to comment.
allright, thanks anyway.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.
