AWS instance

Hi,

I am planning to instantiate AWS EC2 instance that is exactly the same as that of jetson nano. This is to test inference logic before actually testing in the device. If we start facing any issues, i want to upgrade the instance to TX1/TX2 before actually purchasing the device.
Can someone suggest EC2 instance i can make use of to replicate Jetson nano environment in AWS cloud?
Thanks

Hi,

Since EC2 doesn’t have Jetson hardware (iGPU+ARM), you will need to use docker to simulate the environment:
However, since the underlying capability is different, it’s possible to have some gap for inference on a real Jetson.

Thanks.

Thanks @AastaLLL for quick response. When we select EC2, do we required to consider the instance that’s having GPU? If so, how many GPUs do we have to select while instantiating the instance

Hi,

You will need a GPU to run CUDA (even with docker) tasks.

However, since the environment of dGPU and iGPU are quite different.
It’s hard to simulate the performance on EC2 compared to Jetson.

As a result, you can choose a GPU that meet your requirement directly.

Thanks.