NVIDIA GPU搭載
モジュール |
Jetson AGX Orin |
Jetson Orin NX |
Jetson Orin NANO |
MXM RTX3000 |
 |
 |
 |
|
| キャリアボード |
|
|
|
| システムソリューション |
|
|
|
NVIDIA GPU搭載
モジュール |
Jetson AGX Orin |
Jetson Orin NX |
Jetson Orin NANO |
MXM RTX3000 |
| 仕様詳細 |
| AI Performance |
200 / 275 (INT8) TOPS |
16GB: Up to 100 (Sparse) INT8 TOPs and 50 (Dense) INT8 TOPs 8GB: Up to
70 (Sparse) INT8 TOPs and 35 (Dense) INT8 TOPs |
8GB: Up to 40 (Sparse) INT8 TOPs and 20 (Dense) INT8 TOPs 4GB: Up to 20
(Sparse) INT8 TOPs and 10 (Dense) INT8 TOPs |
5.3 TFLOPS Peak FP32 Performance |
| GPU |
NVIDIA Ampere architecture with 2048 NVIDIA® CUDA® cores and 64
tensor cores |
1024 NVIDIA® CUDA® cores | 32 Tensor cores | End-to-end lossless
compression | Tile Caching | OpenGL® 4.6 | OpenGL ES 3.2 | VulkanTM
1.1 | CUDA 10 ONX 16GB: Maximum Operating Frequency: 918 MHz ONX 8GB: Maximum
Operating Frequency: 765 MHz |
Nano 8GB: 1024 NVIDIA® CUDA® cores | 32 Tensor cores Nano 4GB:
512 NVIDIA® CUDA® cores | 16 Tensor cores |
NVIDIA® Quadro® RTX3000
1920 NVIDIA® CUDA® Cores, 30 RT cores, and 240 Tensor Cores |
| CPU |
8 / 12-core Arm Cortex-A78AE v8.2 64-bit CPU 2MB L2 + 4MB L3 / 3MB L2 +
6MB L3 |
Arm Cortex-A78AE v8.2 64-bit CPU + 16 GB LPDDR5 Arm Cortex-A78AE v8.2 64-bit
CPU + 8 GB LPDDR5 |
Six-core (ON 8GB and ON 4GB) Cortex A78AE ARMv8.2 (64-bit) heterogeneous
multi-processing (HMP) CPU architecture | 2x clusters (1x 4-core cluster
+ 128 KB L1 + 256KB L2 per core + 2MB L3) + 1x 2- core cluster (128 KB
L1 + 256KB L2 per core + 2MB L3) | System Cache: 4 MB (shared across all
clusters) |
|
| Memory |
32GB / 64GB 256-bit LPDDR5 204.8 GB/s |
ONX 16GB: 16 GB 128-bit LPDDR5 DRAM ONX 8GB: 8 GB 128-bit LPDDR5 DRAM |
8GB: 8GB 128-bit LPDDR5 DRAM 4GB: 4GB 64-bit LPDDR5 DRAM |
6GB GDDR6 memory, 192-bit (Bandwidth: 336 GB/s) |
| Storage |
64GB eMMC 5.1 |
Supports external NVMe |
Supports external NVMe |
|
| Video Encode |
2x 4K60 | 4x 4K30 | 8x 1080p60 | 16x 1080p30 (H.265) |
1x4K60 | 3x4K30 | 6x1080p60 | 12x1080p30 |
1080p30 Supported via CPU Cores with Software |
|
| Video Decode |
1x 8K30 | 3x 4K60 | 7x 4K30 | 11x 1080p60 | 22x 1080p30 (H.265) |
1x8K30 | 2x4K60 | 4x4K30 | 9x1080p60 | 18x1080p30 |
1x4K60 | 2x4K30 | 5x1080p60 | 11x1080p30 |
|
| DL Accelerator |
2x NVDLA v2.0 |
ONX 16GB: 2x NVDLA | ONX 8GB: 1x NVDLA |
- |
|
| Vision Accelerator |
PVA v2.0 |
PVA v2.0 |
- |
|
| Mechanical |
100mm x 87mm 699-pin Molex Mirror Mezz Connector Integrated Thermal Transfer
Plate |
69.6 mm x 45 mm | 260 pin SO-DIMM Connector |
69.6 mm x 45 mm | 260 pin SO-DIMM Connector |
Standard MXM 3.1 Type B
82 (W) x 105 (D) x 4.8 (H) mm |
