OwnAI Compute
Book a call
NVIDIA Jetson · Orin NX

Jetson Orin NX 8GB

1024-core NVIDIA Ampere architecture GPU · 6-core Arm Cortex-A78AE, 1.5 MB L2 + 4 MB L3 · 8 GB 128-bit LPDDR5

Module price in your region
$1,197 excl. tax
Stock: 36

Manufacturer specifications

Processor architecture
NVIDIA Orin SoC, Arm v8.2 64-bit
CPU
6-core Arm Cortex-A78AE, 1.5 MB L2 + 4 MB L3
GPU
1024-core NVIDIA Ampere architecture with 32 Tensor Cores
AI performance
117 TOPS (INT8, sparse (Super mode))
Memory
8 GB 128-bit LPDDR5
Memory bandwidth
102 GB/s
Storage
None on module; boots from NVMe · 2 M.2 NVMe slot(s) on our carrier
Networking
1× Gigabit Ethernet
Power
10 W – 40 W (configurable power modes)
Dimensions
69.6 × 45 mm module (260-pin SO-DIMM)
Cooling
Active heatsink with fan
Expansion
PCIe Gen4 1×4 + 3×1, 3× USB 3.2 Gen2, 4-lane CSI, 2 × 4K display
Supported frameworks
  • JetPack 6
  • CUDA 12
  • cuDNN
  • TensorRT
  • TensorRT-LLM (selected models)
  • PyTorch
  • ONNX Runtime
  • DeepStream
  • Isaac ROS
  • Ollama
  • llama.cpp
  • MLC
  • Docker

Source: NVIDIA product pages. Figures are manufacturer specifications, not our measurements.

Theoretical vs real-world

117 TOPS is a peak figure at INT8, sparse (Super mode) with structured sparsity. Chat-style LLM generation is limited by the 102 GB/s memory bus and by how much model fits in 8 GB. Rough rule: tokens/s ≈ bandwidth ÷ model bytes.

40
~3B INT4
estimated tokens/s
15
~8B INT4
estimated tokens/s
—
~13B INT4
does not fit

Estimates assume ~0.6 bytes/parameter at INT4 and ~70% achieved bandwidth. Real results vary by runtime.

Measured LLM inference

We only publish figures we or NVIDIA have measured. None are published yet for this module; use the estimates above for planning.

Rejoining the server...

Rejoin failed... trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please retry or reload the page.