What GPU to add?

My Jetson TX2 dev kit comes in a few days. With the Xavier announcement, I think I still want to start working with the TX2 instead of waiting.

What GPU can I add to this board, on a budget? I know cards prices are high now, but I would like something to where I can experiment with some simple image classification NN’s and also I have been working on some highly parallelized code for algorithmic trading and I want to try moving the functionality to using a GPU.

Hi slackmoehrle, Jetson TX2 features an integrated NVIDIA 256-core Pascal GPU with 1TFLOPS+ performance and support for FP16. Here’s an example of using it for image classification with TensorRT: https://github.com/dusty-nv/jetson-inference

Note that although the TX2 devkit has a PCIe x4 slot, JetPack-L4T doesn’t support the discrete GPU PCIe kernel drivers, only the userspace drivers for it’s integrated GPU.

Thanks, so it isn’t possible to add another GPU to this board. What do folks use the PCIx4 slot for then?

Network cards, USB3 cards, disk capacity in various flavors, FPGAs, so on.

@linuxdev of course, thanks, that should have been more obvious.

For FPGA’s is Intel still the main company to buy from? Any recommendations on a card to buy?


I have no experience with that. Most of the questions I’ve seen were based on Xilinx (often related to getting DMA working with SMMU enabled).