I have a Nano 4G dev kit that I frequently get overcurrent alerts, throttling, and temperature alerts. I’m just wondering if this is normal operation for this unit. I’m consuming RTSP so a fair amount of network I/O. I run a display mouse and keyboard, but also still have these problems just using nomachine remote.
I’m using the 5V 4A power supply with the barrel jack, I moved J28 as it says. My power supply setting is “MAXN” which I really don’t know what that means but it was the default. Does all of that sound right ? Do I need more power ? Cooling fan ? I don’t feel like we should be working the unit this hard for what we are doing (basic detection).
As we evaluate this product for potential production use, power and heat are definitely something to consider. I was hoping we could run 2 RTSP cameras on one Nano, but running two models for detection using jetson inference slows our frame rate down considerably. In summary, is this typical expected performance and power usage, and is there a production unit that addresses this in a server/rack configuration ?
It’s a custom python app that is similar to Dusty’s jetson-inference detection app. I can get overcurrent alerts just by consuming from an RTSP camera continuously and doing detection, so maybe it’s network i/o. I’m also writing video files to disk, and converting from cuda to cv2 to draw my own bounding boxes.
Unfortunately I can’t post the code. At any rate it seems OK if I’m only consuming one RTSP camera stream on the nano. I was hoping we could get 2 but I think it’s just too much resource consumption running two instances of our code. Thank you.
I did still wonder if anyone has used hardware other than a development kit they would recommend. At a minimum it needs an ethernet connector, but none of the other peripherals unless an hdmi port is required to get a desktop output window. I didn’t have any luck getting this to work without an hdmi cable attached or a dummy plug.
