NVIDIA Jetson Orin Nano Super: What the 67 TOPS Update Means

The Jetson Orin Nano Super is not new silicon. It is a free software update that raises the Jetson Orin Nano from 40 to 67 TOPS — a 1.7x jump in AI performance — by unlocking higher clocks, more power, and more memory bandwidth. Existing Orin Nano designs gain it with a reflash; new ones ship on the Super profile. The developer kit also came down in price, which is why it became the default entry point to edge AI.

In late 2024 NVIDIA released a software update that re-tuned the Jetson Orin Nano rather than replacing it. The same module runs faster because the update raises the GPU clock to 1020 MHz, lifts the power ceiling from 15 to 25 W, and increases memory bandwidth from 68 to 102 GB/s.

The headline result is up to 67 TOPS of AI performance, against 40 TOPS before — a 1.7x increase on hardware that was already in the field. The CPU and GPU cores are unchanged; what changed is how hard the module is allowed to run. The Super profile reaches the module through a JetPack release, so how JetPack versions decide what a Jetson module can run is worth reading alongside our overview of the NVIDIA Jetson platform.

The gain is entirely from the software profile. The 1024 CUDA cores, 32 tensor cores, and 6-core Arm Cortex-A78AE CPU are the same parts as before.

Numbers only matter if they change what the module can do. The jump from 40 to 67 TOPS, with more memory bandwidth, shifts the Orin Nano from running a single model comfortably to handling more demanding edge AI: faster object detection, more camera frames per second, and small generative and transformer models that were borderline before.

NVIDIA positioned the Super specifically around generative AI at the edge, and the extra bandwidth matters as much as the extra TOPS for those workloads, because they are often memory-bound rather than compute-bound. For classic vision and inference, the practical effect is more headroom on the same module — a heavier model, a higher frame rate, or a second pipeline alongside the first.

The developer kit is the starting point, not a deployable product. The kit is NVIDIA’s reference Jetson carrier board with an Orin Nano module fitted, meant for evaluation and software development — the fastest way to benchmark the Super profile and build an application.

A deployed system pairs the bare module with a Jetson carrier board built for the product’s I/O, environment, and lifecycle, in an enclosure that can actually dissipate 25 W. So the Super developer kit is where a project starts; it is not what ships. We cover the module versus a developer kit in full, and take the module you validated on the kit into a deployable platform.

The Super narrows the gap between the Orin Nano and the next module up, but it does not close it. At 67 TOPS the Orin Nano Super is roughly 40 percent of the Orin NX at 157 TOPS, and it tops out at 8 GB of memory against the NX maximum of 16 GB.

For single-camera vision and entry-level inference it is now a stronger, cheaper option than before. When the workload needs multiple cameras, larger models, or more memory, the Orin NX is still the step up. Our Orin Nano vs Orin NX FAQ covers the choice.

The one thing the Super refresh does not do for free is cooling. The extra performance comes from running the module harder, at up to 25 W, so a thermal design sized for the original 15 W profile will throttle before it reaches 67 TOPS.

We build Orin Nano systems on the Super profile with the cooling to hold it there under load, and confirm the module sustains Super mode rather than dropping back to the old numbers. If you are weighing the Orin Nano Super against the rest of the range, our guide to choosing a Jetson module walks the lineup.