Here's the direct comparison based on the catalog:
| Attribute | Jetson Orin Nano 8GB | Jetson Orin NX 8GB |
|---|---|---|
| Memory bandwidth | 68 GB/s (default mode) · 102 GB/s (Super mode) | 102 GB/s (default mode) · 102 GB/s (Super mode) |
| NVDLA engines | 0 | 1 |
| Max power envelope | 7–15 W (default mode) · 7–25 W (Super mode) | 10–20 W (default mode) · 10–40 W (Super mode) |
Key differences:
- Memory bandwidth: In Super mode both modules offer the same 102 GB/s LPDDR5 bandwidth, but in default mode the Orin NX 8GB runs at 102 GB/s while the Orin Nano 8GB runs at 68 GB/s. nvidia.com +1nvidia.com ↗“memory_bandwidth_gb_per_s=68”nvidia.com ↗“memory.memory_bandwidth_gb_per_s=102”
- NVDLA engines: The Orin Nano 8GB has 0 NVDLA engines; the Orin NX 8GB has 1, giving it a dedicated deep-learning accelerator for offloading vision/AI inference workloads from the GPU. nvidia.com +1nvidia.com ↗“compute.nvdla_engines=0”nvidia.com ↗“compute.nvdla_engines=1”
- Power envelope: The Orin NX 8GB reaches a higher maximum power range (10–40 W in Super mode) versus the Orin Nano 8GB's 7–25 W, reflecting its higher overall compute capability (up to 117 INT8 TOPS in Super mode vs 67 TOPS for the Orin Nano). nvidia.com +3nvidia.com ↗“power_envelope_super=7-25 W”nvidia.com ↗“power_envelope_super=10-40 W”nvidia.com ↗“117 INT8 TOPS (Super mode)”nvidia.com ↗“67 INT8 TOPS (Super mode)”
Both are Ampere-architecture SoMs in the same 260-pin SO-DIMM form factor (69.6×45 mm) with 8 GB LPDDR5. nvidia.com +1nvidia.com ↗“260-pin SO-DIMM (69.6x45mm)”nvidia.com ↗“compute.gpu_architecture=Ampere”