The 16GB module has 2 more CPU cores, 1 more NVDLA engine and 40 more INT8 TOPS in Super mode than the 8GB module.
Jetson Orin NX 8GB vs 16GB nvidia.com +5nvidia.com ↗“compute.cpu_cores=6”nvidia.com ↗“compute.nvdla_engines=1”nvidia.com ↗“performance.max_ai_tops_super=117 TOPS”nvidia.com ↗“compute.cpu_cores=8”nvidia.com ↗“compute.nvdla_engines=2”nvidia.com ↗“performance.max_ai_tops_super=157 TOPS”
| Spec | Jetson Orin NX 8GB | Jetson Orin NX 16GB |
|---|---|---|
| CPU cores (Arm Cortex-A78AE) | 6 | 8 |
| NVDLA engines | 1 | 2 |
| AI performance, Super mode (INT8) | 117 TOPS | 157 TOPS |
| AI performance, default mode (INT8), for reference | 70 TOPS | 100 TOPS |
- CPU: The 8GB has a 6-core Cortex-A78AE CPU. The 16GB has an 8-core one. nvidia.com +1nvidia.com ↗“6-core Cortex-A78AE”nvidia.com ↗“8-core Cortex-A78AE”
- NVDLA: The 16GB has twice as many deep-learning accelerator engines (2 vs. 1).
- Super-mode AI TOPS: Both modules have a Super-mode power envelope of 10–40 W. Within it, the 16GB reaches up to 157 TOPS and the 8GB up to 117 TOPS. On Orin, these figures are sparse INT8 TOPS. nvidia.com +1nvidia.com ↗“power.power_envelope_super=10-40 W”nvidia.com ↗“power.power_envelope_super=10-40 W”
The GPU is the same on both modules: a 1024-CUDA-core Ampere GPU. nvidia.com +1nvidia.com ↗“compute.gpu_cuda_cores=1024”nvidia.com ↗“compute.gpu_cuda_cores=1024”