Research Article
Performance Evaluation of Deep Learning Algorithm Using High-End Media Processing Board in Real-Time Environment
Table 2
Compares various Jetson embedded boards in terms of performance features taken from [
17] adapted from [
39].
| Comparison of different NVIDIA Jetson boards | Characteristics | Jetson Nano | Jetson TX1 | Jetson TX2 | Jetson Xavier NX | Jetson AGX Xavier |
| CPU (central processing unit) | Quad-core ARM A57, 1.43 GHz | Quad-core A57 processor | Quad-core ARM A57 BD Dual-core Denver 2 64-bit CPU | 6-core NVIDIA Carmel ARM 64-bit CPU 6 MB L2+4 MB L3 | 8-Core ARM 64-bit CPU, 8 MB L2+4 MB L3 |
| GPU | 128 core NVIDIA Maxwell | NVIDIA Maxwell GPU with 256 CUDA cores | N Pascal with 256 NVIDIA CUDA cores | NVIDIA Volta, 384 CUDA cores, and 48 tensor cores | 512-CUDA cores and 64 tensor cores |
| DL accelerator (deep learning accelerator) | None | None | None | 2× NVDLA engines | (2×) NVDLA engine |
| Memory | 2 GB 64-bit LPDDR4 | 4 GB LPDDR4 memory | 8 GB 128-bit LPDDR4 | 8 GB 128-bit LPDDR4× 59.7 GB/s | 16 GB 256-bit LPDDR4x |
| Storage | MicroSD | 16 GB eMMC 5.1 | 32 GB eMMC 5.1 | MicroSD | 32 GB eMMC 5.1 |
| Video encoder | 4Kp30-4× 720p30 (H.264/H.265) | 4 K at a rate of 30 | at a rate of 30 (HEVC) | , 20× 1080p30 (H.264) | at a rate of 60 (HEVC) |
| Video decoder | 4Kp60-2×18× 720p30 (H.264/H.265) | 4 K rate of 30 | at 30, 12-bit support | K30-6× 4 K60 1080p 30 (H.265) 22× 1080p30 (H.264) | at a rate of 30 12-bit support |
| Cost | $99 | $520 | $599 | $399 | $699 |
|
|