NVIDIA introduces lower cost Jetson TX2 NX SO-DIMM module

NVIDIA Jetson AI modules and developer kits range from the entry-level Jetson Nano module (5W, ~0.5 TOPS) to the high-end Jetson Xavier AGX module (30W, 32 TOPS). The higher-end modules usually come with a 400-pin board-to-board, while cheaper/cost-down modules like Jetson Nano and Jetson Xavier NX feature a 260-pin SO-DIMM connector and small form factor.

But so far all Jetson TX2 modules came with a 400-pin connector, but this has changed with the introduction of the Jetson TX2 NX SO-DIMM system-on-module that is offered as a cost-down version of Jetson TX2 4GB module.

Jetson TX2 NX SO-DIMM ModuleNVIDIA Jetson TX2 NX specifications with highlights in bold or strikethrough showing the difference against Jetson TX2 4GB:

  • Processor – Dual-Core NVIDIA Denver 2 64-Bit CPU and Quad-Core Arm Cortex-A57 MPCore processor @ up to 2.0 GHz
  • GPU – 256-core NVIDIA Pascal GPU @ up to 1.3 GHz
  • AI Performance – 1.33 TFLOPS
  • Video
    • Encode
      • H.265 – 1x 4Kp60, 3x 4Kp30, 4x 1080p60, 8x 1080p30
      • H.264 – 1x 4Kp60, 3x 4Kp30, 7x 1080p60, 14x 1080p30
    • Decode
      • H.265/H.264 – 2x 4Kp60, 4x 4Kp30, 7x 1080p60, 14x 1080p30
  • System Memory – 4 GB 128-bit LPDDR4 51.2GB/s
  • Storage – 16 GB eMMC 5.1
  • 260-pin SO-DIMM edge connector
    • Display I/F
      • 2x multi-mode DP 1.2/eDP 1.4/HDMI 2.0
      • 2 x4 1 x2 DSI (1.5Gbps/lane)
    • Camera I/F
      • Up to 6x 5x cameras (12 via virtual channels)
      • 12 lanes MIPI CSI-2
      • D-PHY 1.1 1.2 up to 30 Gbps
    • Networking – Gigabit Ethernet
    • PCIe – 1×1 + 1 x4 1 x2  PCIe Gen2 , or 1 x1 + 1 x1 + 1 x2
    • USB – 1x USB 3.1 (5Gbps), 3x USB 2.0
    • Other I/Os – 1x SDIO, 2x SPI, 3x UART, 4x I2S, 4x I2C, 1x CAN, GPIOs
  • Supply Voltage – 5V
  • Power Modes – 7.5 or 15W
  • Dimensions – 69.6 mm x 45 mm

The main differences are the form factor, and fewer interfaces for display, camera, and PCIe. NVIDIA  Jetson TM TX2 NX places itself right between Jetson Nano and Jetson Xavier NX in terms of performance, and shares form-factor and pin compatibility with the two modules.

Jetson SoftwareNVIDIA explains the new module is suitable for mass-market AI products, and retains software compatibility with the JetPack SDK including libraries and samples for accelerators embedded into the chip. Major AI frameworks such as Tensorflow and Caffe are supported, and the company provides pre-trained AI models from NVIDIA NGC and the NVIDIA Transfer Learning Toolkit can help with quickly developing AI applications. You can download documentation and software by registering for a free developer account.

Samples of Jetson TX2 NX modules are available now for $200 on Arrow or 140 Euros on Silicon Highway Direct. [Update: It’s also listed on Seeed Studio with Jetson Mate carrier board as an option. Further details may be found on the Jetson product page.

Thanks to Abu for the tip.

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard

9 Replies to “NVIDIA introduces lower cost Jetson TX2 NX SO-DIMM module”

  1. A newbie here, can you kindly recommend any 260 pin sodimm “mother board” to plug this in?

  2. I really have no need for this but i do want it.

    I wonder if this is the SoC that is going to be used in the next Switch

    1. People thinking the next Switch will have PS4 performance will be very disappointed if this is. Nintendo would need Xavier NX for that.

      1. I havent seen anything remotely close to ps4 level in even the most optimistic discussion on the rumored ‘switch pro’. Unless people are mixing up a dlss 4k solution with actually being able to render natively at 4k (which even the ps4 couldnt do).

  3. Well damn, so that means that all this time, the NX in Xavier NX really had nothing to do with Nintendo at all. Might have been a red herring.
    Thinking about it now, “NX” probably just means “Nano X” in reference to the Jetson Nano which has a cutdown Tegra X1 that doesn’t have a special name.

  4. I don’t understand why they continue to push their totally outdated, 9-years old, power-inefficient A57 here. Either they need power and they put something better, or they don’t need power and they put a way more efficient A53 or A55. It’s even possible than a 2 GHz A55 would deliver more performance with less power. They are simply trading their customer’s power bill for their laziness. It might be a mentality in the company not to care at all about power efficiency, that probably affects the rest of their product line after all, when you see how thick the power cables have to be to power their modern GPUs…

Leave a Reply

Your email address will not be published. Required fields are marked *

Boardcon Rockchip and Allwinner SoM and SBC products
Boardcon Rockchip and Allwinner SoM and SBC products