Among NVIDIA's production modules compatible with the Auvidea JNX46 carrier board, the Jetson Orin NX 16GB has the highest AI performance: 100 TOPS in default mode and 157 TOPS in Super mode (INT8, sparse), paired with 16 GB of LPDDR5 memory running at 102 GB/s.
Here's the full comparison across the three supported module families:
NVIDIA production modules for the JNX46 carrier board auvidea.eu +6auvidea.eu ↗“compatibility.supported_jetson_families=Jetson Xavier NX, Jetson Orin Nano, Jetson Orin NX”nvidia.com ↗“100 TOPS”nvidia.com ↗“70 TOPS”nvidia.com ↗“40 TOPS”nvidia.com ↗“20 TOPS”developer.nvidia.com ↗“up to 21 INT8 TOPS”developer.nvidia.com ↗“up to 21 INT8 TOPS”
| Module | Max AI TOPS (default) | Max AI TOPS (Super) | Memory |
|---|---|---|---|
| Jetson Orin NX 16GB | 100 TOPS | 157 TOPS | 16 GB LPDDR5 |
| Jetson Orin NX 8GB | 70 TOPS | 117 TOPS | 8 GB LPDDR5 |
| Jetson Orin Nano 8GB | 40 TOPS | 67 TOPS | 8 GB LPDDR5 |
| Jetson Orin Nano 4GB | 20 TOPS | 34 TOPS | 4 GB LPDDR5 |
| Jetson Xavier NX 16GB | 21 TOPS | — (no Super figure) | 16 GB LPDDR4x |
| Jetson Xavier NX 8GB | 21 TOPS | — (no Super figure) | 8 GB LPDDR4x |
A few notes on the figures:
- The TOPS values are INT8 and sparse for the Orin-family modules, and the JNX46 itself lists 100 TOPS / 157 TOPS (Super), which corresponds to the Orin NX 16GB module. auvidea.eu +1auvidea.eu ↗“INT8 TOPS (sparse, Base)”auvidea.eu ↗“Super Mode support (JetPack 6.2)”
- Super mode applies only to the Orin NX and Orin Nano modules; the Xavier NX modules have no Super-mode figure stated by their manufacturer. developer... +1developer.nvidia.com ↗“up to 21 INT8 TOPS”developer.nvidia.com ↗“up to 21 INT8 TOPS”
So the answer: Jetson Orin NX 16GB — 100 TOPS (default) / 157 TOPS (Super), using LPDDR5 memory (16 GB).