I am designing a new Jetson Nano Board, and I want to put two M.2 Key E ports on it instead of one. Can it be done? Does it matter which PCIe lane I connect too?
Hi pkoump, the Jetson Nano module has one PCIe controller available, so if you wanted to connect multiple PCIe endpoint devices, you would need to use a PCIe switch to bifurcate the lanes.
So, even though the NANO has 4 PCIe ports, I cannot use the other three. There is one PCIe only controller, or I cannot use the other 3 PCIe ports because they are all connected on one controller and that is the limiting factor?
The Design guide states:
Tegra contains a PCIe controller that brings one interface up to four lanes to the module pins for use on the carrier board. A second single-lane PCIe interface is used on-module for Ethernet.
So, even though I have one controller interfacing 4 lanes, I can have only one M.2 Key E interface.
Nano has one external PCIe port, and that port can have up to four lanes.
That’s correct, as you can only have one endpoint device connected per controller, you would need to use a PCIe switch to connect multiple devices.
From the discussion, the controller is a PCIex4 with single controlling clock limiting it to one PCIe up to four lane. If I got this correct, what switch does the NVIDIA suggest since it is not on the supported component list. Also, will the switch (say a ports switch) need a clock generator for each port or one clock for all?
I don’t believe we have done another Nano design with PCIe switch, so we don’t have official suggestion, however recommend looking into PCIe gen2 switches or consulting with Jetson hardware design ecosystem: https://developer.nvidia.com/embedded/community/ecosystem#hardware_design