Skip to main content
EDGE AI · NVIDIA JETSON

Jetson Orin Nano vs Orin NX: What Is the Difference?

The Jetson Orin Nano delivers up to 67 TOPS in 7 to 25 W with 4 to 8 GB of memory, suited to entry-level, low-power vision. The Orin NX steps up to 157 TOPS in 10 to 40 W with 8 to 16 GB, adds deep learning accelerators and a hardware video encoder, and handles multi-camera, heavier inference.

01

Orin Nano and Orin NX at a glance

The Jetson Orin Nano and Jetson Orin NX are NVIDIA’s two smaller Orin modules. They share the same 260-pin SO-DIMM connector and footprint but sit at different performance levels. With the Super power modes in JetPack 6.2, the Orin Nano gives up to 67 TOPS in a 7 to 25 W envelope, while the Orin NX reaches up to 157 TOPS in 10 to 40 W, with more memory.

The Nano suits single-camera, entry-level vision and cost-sensitive robots. The NX handles several cameras, larger models and applications that need hardware video encoding. Because the modules are pin-compatible, a carrier board designed for both lets you change tier without changing the rest of the system.

02

Orin Nano vs Orin NX specifications

Each module comes in two memory sizes. The figures below are NVIDIA’s published specifications with the Super modes enabled.

Specification Orin Nano 4GB Orin Nano 8GB Orin NX 8GB Orin NX 16GB
AI performance (Super) 34 TOPS 67 TOPS 117 TOPS 157 TOPS
AI performance (original) 20 TOPS 40 TOPS 70 TOPS 100 TOPS
GPU 512 CUDA, 16 Tensor Cores 1,024 CUDA, 32 Tensor Cores 1,024 CUDA, 32 Tensor Cores 1,024 CUDA, 32 Tensor Cores
CPU 6-core Arm Cortex-A78AE 6-core Arm Cortex-A78AE 6-core Arm Cortex-A78AE 8-core Arm Cortex-A78AE
Deep learning accelerator None None 1× NVDLA v2 2× NVDLA v2
Memory 4 GB LPDDR5, 64-bit 8 GB LPDDR5, 128-bit 8 GB LPDDR5, 128-bit 16 GB LPDDR5, 128-bit
Hardware video encode None (CPU encode) None (CPU encode) 1× 4K60 H.265 1× 4K60 H.265

TOPS figures are sparse INT8 peaks across the GPU and, on the NX, the DLAs. The extra NX TOPS come partly from the DLAs and partly from higher GPU clocks, so a GPU-only model gains less than the headline ratio suggests.

03

Where the Orin NX pulls ahead

Beyond raw TOPS, the NX carries more memory, more CPU cores on the 16GB version and a higher power ceiling, so it sustains larger models and several concurrent pipelines. If a single camera and a single model fit comfortably on the Nano, there is no need for the NX. Once the workload grows, the NX is the step up without changing form factor.

  • //
    Multiple cameras. Concurrent multi-camera perception needs the NX’s extra compute and memory bandwidth headroom.
  • //
    Larger models. 8 to 16 GB lets you run bigger models, or several models at once, that the Nano cannot hold.
  • //
    DLA offload. The NX’s deep learning accelerators can run a second model alongside the GPU at low power.
  • //
    Future headroom. If the application is likely to grow, the NX buys room without a redesign.
04

Power and cooling

The two modules draw very different power at their top modes: up to 25 W for the Orin Nano and up to 40 W for the Orin NX. In a fanless enclosure that difference decides the size of the heatsink and how warm a cabinet the system can sit in. A design rated for the Nano at 25 W may need to cap an NX at a lower power mode, which gives away part of its advantage.

Plan the thermal design for the module and power mode you intend to ship, and validate it at the highest ambient temperature on site. A module that throttles under heat delivers less than its rating, whichever tier you choose.

05

Video encoding: the difference people miss

The Orin Nano has no NVENC hardware video encoder. It can decode video in hardware, but encoding H.264 or H.265 runs on the CPU cores, which limits how many streams it can record or re-stream at once. The Orin NX keeps a hardware encoder that handles a 4K60 H.265 stream, or several lower-resolution streams.

For applications that only analyze video and send results, this rarely matters. For recorders, video gateways and systems that stream annotated video to an operator, it often decides the choice on its own.

06

Pin-compatible by design

Because the Nano and NX share a connector and form factor, a carrier board designed for one generally accepts the other. That lets you prototype on a Nano and move to an NX, or offer both tiers on one design, without a board change. Check that the carrier’s power supply and cooling are rated for the NX at its higher power modes, and that the board support package covers both modules.

Our Jetson Orin Nano computers and Jetson Orin NX computers cover both tiers, and many Jetson Orin NX carrier boards take either module. Our guide to choosing a Jetson module compares them with the AGX Orin and Thor.

07

Quick answers on Orin Nano vs Orin NX

Orin Nano vs Orin NX: common questions

Is the Orin NX faster than the Orin Nano?

Yes. The Orin NX 16GB is rated at up to 157 TOPS against 67 TOPS for the Orin Nano 8GB, with more CPU cores and memory. Real gains on a GPU-only model are usually smaller than that ratio.

Can an Orin Nano carrier board take an Orin NX?

Usually, yes. They share the 260-pin SO-DIMM connector and pinout. Confirm that the carrier’s power, cooling and software support cover the NX before swapping.

Does the Orin Nano have a hardware video encoder?

No. It decodes in hardware but encodes on the CPU. The Orin NX has a hardware encoder for H.264 and H.265.

Do both modules support Super mode?

Yes. JetPack 6.2 adds higher-power Super modes to both, raising the Orin Nano 8GB from 40 to 67 TOPS and the Orin NX 16GB from 100 to 157 TOPS.

Which has a deep learning accelerator?

Only the Orin NX. The Orin NX 8GB has one NVDLA and the 16GB version two. The Orin Nano relies on its GPU alone.

We size the Jetson module to the workload and build the carrier board, storage, cooling and JetPack image around it as standard. Tell us about your cameras, models and power budget, and we will recommend the Nano or NX build that fits.