Versions Compared

Old Version 9

changes.mady.by.user Patricia Liu

Saved on

compared with

New Version 10

changes.mady.by.user Anthony Luo

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Yup. Title. I’m gonna have to learn how to use confluence.

Communication with Jetson vs SPI/CAN/ETC/WHATEVER WE WANT

  • needs much more research.

  • need to confirm what data is being transferred. Is the camera module being handled by the jetson entirely? are we just getting position from cv? what’s the plan……

There is a tool called Jetson_IO that can be used to make configuring GPIO pins easier.

I/O

Notes

SPI

I2C

UART

CAN

  • built in CAN interfacing is only available to the more expensive Jetson device modules.

  • CAN controller is not available in Jetson Nano, but you can use an MCP251x chip (SPI to CAN interface), which works with the SPI interface on Jetson Nano.

I2S

USB?

  • Could be a viable alternative.

  • Need to consider how many datalines for USB our chip has.

Some quick links

RTOS (zp) on Jetson?

  • Concurrent Redhawk Linux is supported on Jetson TX1/TX2, Nano, and Jetson AGX Xavier. This is an RTOS version of Linux.

  • Seems like there is already a FreeRTOS too.

  • zephyr doesn't seem to have any board support for ARM Cortex-A57, or Carmel ARMv8.2.

CV on RTOS?

  • might have issues ← come back to this.

RPi vs Jetson

Below we find our comparisons between our different device considerations. These are the specs for the device module, not the developer kit.

Benchmark

RPI

Jetson Nano

Jetson Xavier

JEtson TX2

Physical specifications

RPI (RPI4 Model B )

Jetson Nano

Jetson AGX Xavier

Jetson TX2

Power consumption

Can range between 2.56W to 7.30W

Between 5W to 10W

Between 10W to 30W

Between 7.5W to 15W

CPU

  • Broadcom BCM2711B0 quad-core A72 (ARMv8-A) 64-bit @ 1.5GHz

  • Quad-core ARM Cortex-A57 MPCore processor @ 1.43 GHz

  • 8-core Carmel ARM v8.2 64-bit CPU, 8MB L2 + 4MB L3

  • 4-core ARM Cortex-A57 @ 2 GHz and 2-core Denver2 @ 2 GHz

RAM (volatile memory) and Memory storage

  • 2GB, 4GB or 8GB LPDDR4-3200 SDRAM

  • Uses a 4 GB LPDDR4 (Low-Power Double Data Rate) RAM that offers 25.6 GB/s

  • Apparently microSD for the developer kit. 16 GB eMMC 5.1 for the actual module

  • 32GB 256-Bit LPDDR4x | 136.5GB/s

  • 32GB eMMC 5.1

  • Uses a 8 GB 128-bit LPDDR4 RAM that offers 58.3 - 59.7 GB/s

  • 32 GB eMMC 5.1 Flash Storage

GPU

  • Broadcom VideoCore VI

  • NVIDIA Maxwell architecture with 128 NVIDIA CUDA® cores

  • 0.5 TFLOPS (FP16) in terms of AI performance

  • 512-core Volta @ 1.37 GHz

  • 11 TFLOPS (FP16)
    22 TOPS (INT8)

Has tensor cores (64)

  • NVIDIA Pascal™ architecture with 256 NVIDIA CUDA cores

  • 1.3 TFLOPS (FP16) in terms of AI performance

Networking and Display (USB, ethernet, etc.)

  • 2.4 GHz and 5.0 GHz IEEE 802.11ac wireless, Bluetooth 5.0, BLE

  • Gigabit Ethernet

  • 2 USB 3.0 ports; 2 USB 2.0 ports.

  • 2 × micro-HDMI ports (up to 4kp60 supported)

  • Gigabit Ethernet and M.2 Key E

  • HDMI 2.0 and eDP 1.4 (1 HDMI, 1 display port)

  • 4 USB 3.0, 1 USB 2.0 Micro-B

  • Gigabit Ethernet

  • 3 multi-mode DP 1.2/eDP

  • USB 2.0 (4?), 3 USB 3.1

  • 1.4/HDMI 2.0 a/b

  • no wireless - can use usb wifi adapter though

  • USB 3.0 Type A and USB 2.0 Micro AB

  • Has wifi (Connectivity to 802.11ac Wi-Fi and Bluetooth-Enabled Devices)

  • Gigabit Ethernet and M.2 Key E

  • 2 multi-mode DP 1.2/eDP

  • 1.4/HDMI 2.0

  • 2 x4 DSI (1.5Gbps/lane)

Supported communication interfaces

GPIOs

Supports GPIO, I2C, I2S, SPI, UART

Supports UART, SPI, CAN, I2C, I2S, DMIC & DSPK, GPIOs

Supports GPIOs, I2C, I2S, SPI, CAN, and TTL UART with flow control

Cost

Starting from $35 (increases depending on RAM)

$149

$999

Starting at $299

Jetson TX2 NX is $199 but is smaller and has less RAM.

Notes:

Between RPI4 and Jetson Nano, RPI4 is slightly better in terms of Display and CPU. RPI also has built-in wifi (which is nice). The general computing performance between both devices is relatively similar; however, when it comes to machine learning, 3D graphics, or very complex algorithms, Jetson Nano is better thanks to its GPU.