ThinkRobotics Jetson Nano Dev Kit (SUB) Review: The eMMC-Equipped Nano for Learners and Makers
by Gaurav Sarraf
The NVIDIA Jetson Nano was the board that put GPU-accelerated edge AI within reach of students, hobbyists, and small-scale developers when it launched. Years on, the platform continues to sell because it covers an important category: a genuine CUDA-enabled development environment at a price point well below the Jetson Orin family.
The ThinkRobotics Jetson Nano Dev Kit (SUB) is a variant of the Jetson Nano B01 that makes one meaningful hardware change over the standard NVIDIA Developer Kit: the Nano module it ships with includes 16GB of onboard eMMC storage. This removes the requirement for a microSD card, which has been one of the most frequent friction points for new Jetson Nano users. This review covers what that difference means in practice, the full specs, JetPack support status, who should buy it, and how it sits relative to the official B01 and the current Jetson Orin Nano lineup.
What Is the Jetson Nano Dev Kit (SUB)?
The SUB designation refers to a Jetson Nano developer kit built around NVIDIA's official Jetson Nano 4GB production module, which includes 16GB of eMMC flash storage directly on the module. The carrier board matches the layout of the B01 reference design in size, connector placement, and peripheral set, which means it is compatible with the same cameras, HATs, cases, and accessories designed for the B01 ecosystem.
The compute specifications of the Jetson Nano module are unchanged from the standard B01: the same quad-core Arm Cortex-A57 CPU, the same 128-core Maxwell GPU, and the same 4GB of LPDDR4 memory. What changes is how the operating system is stored and how the board boots.
Gigabit Ethernet (10/100/1000Base-T, PoE-ready with external module)
Wireless
M.2 Key E slot (Wi-Fi/BT card not included)
GPIO
40-pin Raspberry Pi-compatible header
Other Interfaces
I2C, SPI, UART, I2S
Power Supply
5V/4A via DC barrel jack, or 5V/2A via Micro-USB
Power Consumption
5W (5W mode) or up to 10W (10W MAXN mode)
JetPack Support
JetPack 4.6.1 (latest supported, Ubuntu 18.04)
Operating System
Ubuntu 18.04 LTS
The eMMC vs microSD Difference
This is the defining characteristic of the SUB version. The standard NVIDIA Jetson Nano B01 Developer Kit ships with a Nano module that has no onboard storage. Every user must purchase a microSD card, flash the JetPack image, and boot from that card. In practice, microSD cards create three categories of problems that eMMC solves.
Standard B01 approach
microSD Boot
!Card quality varies; cheap cards cause slow boot times and filesystem corruption under heavy write loads
!The exposed microSD slot is easy to accidentally remove or damage during project builds
!Read and write throughput is a bottleneck for data-logging or model-access applications
!Reflashing is simple (Balena Etcher) but the card can fail or corrupt unexpectedly
SUB version
16GB eMMC (onboard)
✓Soldered directly onto the module. No physical removal or accidental damage possible
✓Faster read/write speeds: roughly double that of a typical Class 10 microSD card
✓MicroSD slot retained for expansion alongside the eMMC
Reflashing requires SDK Manager: Updating or reinstalling JetPack on the eMMC module requires a host Ubuntu machine running NVIDIA SDK Manager and a USB connection to the board in recovery mode. This is more involved than re-writing a microSD card with Balena Etcher, but provides a more stable long-term storage solution. Plan for this before buying if you switch JetPack versions frequently.
JetPack and Software Support
The Jetson Nano platform is now on JetPack 4.6.1 as its final supported release. NVIDIA has confirmed that JetPack 5.x and 6.x are not supported on the Jetson Nano module. Understanding this ceiling is important before committing to the platform for a long-term project.
JetPack
4.6.1 (final)
OS
Ubuntu 18.04
CUDA
10.2
cuDNN
8.2
TensorRT
8.0.1
JetPack 4.6.1 is a stable and functional environment for the vast majority of learning, computer vision, and robotics applications. Most popular tutorials, ROS packages, and AI example repositories targeting the Jetson Nano were written for JetPack 4.x and run without modification.
Learning OpenCV with CUDA acceleration, TensorRT inference, YOLOv4 object detection
Classification and detection with TensorFlow and PyTorch (JetPack 4.x-compatible versions)
ROS Melodic natively on Ubuntu 18.04; ROS Noetic via Docker
NVIDIA Deep Learning Institute (DLI) course projects and examples
University AI course projects that already target JetPack 4.x
JetPack ceiling to know: Some newer Python packages, updated PyTorch versions, and frameworks that now require Ubuntu 20.04 or later will not install natively. For teams building new production systems in 2025 who need the latest Ubuntu LTS or JetPack 6.x capabilities, the Jetson Orin Nano Super Developer Kit is the appropriate platform.
Use Cases
🧠
Learning GPU-Accelerated AI
One of the lowest-cost ways to get hands-on time with CUDA, TensorRT, and real-time inference. Running YOLOv4 at 25 to 30 FPS on a live camera feed, training a simple image classifier with PyTorch, and deploying a speech recognition model are all achievable on the Nano with JetPack 4.6.1.
🎓
University AI and Robotics Courses
Many Indian engineering colleges continue to use the Jetson Nano as their standard AI hardware lab platform. The well-established JetPack 4.x curriculum materials and NVIDIA DLI online courses provide a complete structured learning path. The eMMC eliminates the shared-lab card corruption issues that plague standard B01 deployments.
🤖
ROS Robotics Projects
The Jetson Nano supports ROS Melodic on Ubuntu 18.04 natively, and ROS Noetic through Docker. For students building autonomous mobile robots, the Nano provides the CUDA compute needed for perception and planning workloads that a Raspberry Pi cannot handle. The 40-pin GPIO, dual CSI ports, and USB 3.0 interfaces connect directly to standard robotics peripherals.
👁️
Computer Vision Prototyping
Object detection, face recognition, lane tracking, and visual inspection projects running OpenCV with CUDA acceleration perform well on the Nano's 128-core Maxwell GPU. Dual CSI camera ports support stereo camera setups for depth estimation and binocular tracking applications.
🛠️
Maker and DIY AI Projects
For individuals building AI-driven door locks, plant monitoring systems, bird feeders with species identification, or home automation systems with computer vision, the Nano's 5W power mode means it can run from a standard power bank for portable deployments.
🏛️
Existing JetPack 4.x Deployments
For teams maintaining systems already built around JetPack 4.x software, or institutions with existing Nano-based curriculum, the SUB version provides a more reliable hardware base for those same workflows with the eMMC upgrade.
Pricing in India
The ThinkRobotics Jetson Nano Dev Kit (SUB) is available through Think Robotics with local warranty support in the range of Rs. 21,000 to Rs. 22,500, competitive with international pricing given import duties and local support. Find current pricing at the product page.
A power supply is not included in the standard kit. A 5V/4A DC barrel jack adapter is required for full 10W operation. The 5V/2A micro-USB supply can be used for lower-power configurations but may cause instability when connecting multiple USB peripherals simultaneously.
What Users Are Saying
Feedback from the NVIDIA Developer Forums, JetsonHacks, the Medium engineering community, and verified buyers covers consistent themes about eMMC reliability and the practical differences users encounter moving from microSD to the production module.
eMMC flash gives NVIDIA the flexibility of being able to spec which actual memory chips and cell type are being used. For industrial applications, the module can contain high-rated SLC. The developer gets the comfort of knowing that there is at least one configuration they can depend on. On a development board, you're much more likely to fry other components before you ever reach the life of the eMMC.
Jetson Nano's microSD speeds look pretty reasonable at 84.63 MB/sec. But the weakest link in the package is the storage IO speeds, particularly during neural network training where camera frames stored on the microSD need to be brought into DRAM first. The eMMC module sku avoids this bottleneck entirely.
I wish it had eMMC from the start and not having to wait. The onboard eMMC on the production module addresses one of the most complained-about limitations of the developer kit, which was the unreliability of microSD cards in sustained-use scenarios.
The Jetson device is failing to boot because of multiple hardware errors. The system cannot communicate with several key components, including a thermal sensor and a Wi-Fi module. Note that Wi-Fi requires a separate M.2 card and is not built into the module.
Both the Nano and Xavier NX production modules come equipped with 16GB of on-board eMMC storage. When the latest JetPack and SDK application components are fully loaded, they can use up to 14GB of storage, so additional storage flexibility via the microSD or USB is invaluable for most projects.
For a university project using the Jetson Nano 4GB with eMMC 16GB developer kit, the current OS is L4T R32.6.1 (JetPack 4.6) booting from eMMC. The microSD can be mounted for additional storage but the OS boots from eMMC by default on this module variant.
Setup infoNVIDIA Developer Forums, university project thread, 2025
How It Compares
SUB vs
Official Jetson Nano B01 Dev Kit
The official B01 uses a Nano module without onboard storage, requiring a microSD card. The SUB uses NVIDIA's production module with 16GB eMMC, removing the microSD dependency for the OS. The carrier board layout, interface set, and compute performance are otherwise identical.
The Pi 5 is faster in CPU-heavy tasks and has no CUDA GPU, no TensorRT support, and no dedicated AI accelerator. For AI inference workloads and computer vision with GPU acceleration, the Jetson Nano SUB is the correct choice. For Python scripting, GPIO projects, and IoT without GPU compute, the Pi 5 is faster and better-supported.
The Orin Nano Super delivers 67 TOPS versus the Nano's 472 GFLOPS, runs JetPack 6.2 on Ubuntu 22.04, and supports generative AI workloads the Nano cannot handle. For anyone starting a new project in 2025, the Orin Nano Super is the recommended starting platform.
Orin for new projects; Nano for budget/legacy
Before You Buy
✅
Full B01 hardware compatibility. Same carrier board layout, same camera connectors, GPIO header, and case ecosystem as the standard B01. All B01-compatible cameras, HATs, and accessories work without modification.
✅
MicroSD slot retained for expansion. If the 16GB eMMC fills up with JetPack, application code, and model files, a microSD card provides additional storage. Many users boot from eMMC and store large datasets on microSD or USB drives.
✅
Dual CSI camera ports for stereo vision. Two 2-lane MIPI CSI-2 ports support simultaneous camera streams for stereo depth estimation, binocular tracking, or dual-angle monitoring applications.
✅
5W power mode for portable AI projects. At 5W, the board can run from a capable power bank. Suitable for field deployments, portable vision systems, and battery-powered maker projects.
⚠️
JetPack 4.6.1 is the final release. No JetPack 5.x or 6.x support on this hardware. Budget for this software ceiling in any long-term project plan. Ubuntu 18.04 will also reach community end-of-life for security patches.
⚠️
Power supply not included. A 5V/4A DC barrel jack adapter is needed for 10W operation. Using the micro-USB supply limits the board to 5W and may cause instability when connecting multiple USB peripherals simultaneously.
⚠️
Wi-Fi and Bluetooth not built in. An M.2 Key E Wi-Fi/Bluetooth card must be added separately for wireless connectivity. This is an easy addition but adds cost and is easy to overlook at checkout.
⚠️
Reflashing requires a host Ubuntu machine. SDK Manager on a host machine is needed to reflash the eMMC. Keep a host Ubuntu machine available or use a VM if you plan to update JetPack versions or recover from a corrupted image.
Conclusion
The Most Practical Jetson Nano for Learning and Education in India
The ThinkRobotics Jetson Nano Dev Kit (SUB) is the most practical way to buy a Jetson Nano in India for learning and education. The 16GB eMMC module removes the microSD card dependency that causes the most common setup frustrations with the standard B01, while retaining full compatibility with the B01 hardware ecosystem.
The 472 GFLOPS Maxwell GPU, 4GB LPDDR4 RAM, dual CSI camera ports, and JetPack 4.6.1 software stack provide a complete CUDA learning environment at an accessible price for individual developers and student teams. The JetPack 4.x software ceiling is the main limitation for multi-year production deployments, and in those cases the Jetson Orin Nano Super is the more forward-looking platform. Find the Jetson Nano Dev Kit (SUB) at Think Robotics with local warranty and support across India.
The SUB version uses NVIDIA's production Nano module with 16GB of onboard eMMC storage, so it boots without a microSD card. The standard B01 uses a module with no onboard storage and requires a microSD card for the OS. All other hardware including the CPU, GPU, RAM, and carrier board interfaces is identical. The eMMC provides faster read/write speeds, better durability, and eliminates the physical vulnerability of an exposed card slot.
No, for normal use. The 16GB eMMC holds the JetPack operating system and standard packages. If you need more storage for datasets or large model files, the microSD slot accepts a card for expansion alongside the eMMC. Many users run the OS from eMMC and store training data or large files on a microSD or USB drive.
No. JetPack 4.6.1 running on Ubuntu 18.04 is the final supported release for the Jetson Nano module. JetPack 5.x and 6.x are not supported on this hardware. This is the most important consideration for long-term project planning. For JetPack 6.x, Ubuntu 22.04, and access to generative AI capabilities, the Jetson Orin Nano Super Developer Kit is the correct platform.
No to both. A 5V/4A DC barrel jack adapter is needed for full 10W operation. Using the micro-USB supply limits the board to 5W and may cause instability with multiple USB peripherals connected. Wi-Fi and Bluetooth require a separately purchased M.2 Key E card. Budget for both before ordering if wireless connectivity is part of your project.
For new projects in 2025, the Orin Nano Super is the better long-term choice: 67 TOPS, JetPack 6.2, Ubuntu 22.04, and support for generative AI models. The Jetson Nano SUB is the right choice for budget-constrained buyers, existing JetPack 4.x workflows, educational courses built around the Nano platform, and any application where the Maxwell GPU with CUDA is sufficient for the workload.
Get the Jetson Nano Dev Kit (SUB) from Think Robotics India
eMMC-equipped production module, full B01 ecosystem compatibility, CUDA learning environment. Local warranty and support across India.
3D printed parts are of low finish. Lot.of sharp edges on parts. Also the fitment of the parts are not happening because of excess material. Most often the parts have to be filed or reworked. Screws are given in exact number as needed. If one screw is missed, then the assembly is impossible
The consignment was not dispatched as per committed date and had to pursue for dispatch information. The courier didn,t deliver us despit lying in Jaipur and we had to personally collect it from Courier Hub adding to inconvenience. The logistics need to be improved. Just a feedback.
Bought this from this website, stopped working within a month. Support made me wait for weeks, slow response.
At the end the Abhiram from the technical team tells me I should take it to some local electrical repair person who might be able to change the element.
Disgusting quality and service. I was told it does come under warranty but still no solution.
I've sent them photos, videos, nothing. Disgusting service.
