Imagine you need an edge computer that can run AI models without a bulky server. The Firefly CORE-K3JD4, 8GB/128GB RISC-V Edge System on Module (SoM) gives you that power in a tiny board. It is built for developers who struggle with slow inference and limited storage on traditional single-board computers. With this module you can deploy real-time AI at the edge and keep your product responsive.
Firefly CORE-K3JD4, 8GB/128GB RISC-V Edge System on Module (SoM)
Edge projects often hit a wall because the processor cannot handle large neural networks and the board does not have enough memory. When the device stalls, the whole system fails to meet latency targets and the user experience suffers. This module solves the problem with an 8-core RISC-V AI processor, 8 GB LPDDR5 RAM and up to 128 GB UFS storage, all in a 70 mm 45 mm board that runs on 12-19 V. The result is fast inference, ample storage and a small footprint that fits into any enclosure.
Key Features
- 8-core A100 AI engine delivering 60 TOPS (SquareINT4).
- 8-core X100 64-bit RISC-V CPU at 2.4 GHz, 130 KDMIPS.
- Supports BF16, FP16, FP8, INT8, INT4 data formats.
- Integrated GPU with Vulkan 1.3, OpenCL 3.0, OpenGL ES 3.2 support.
- 8 GB LPDDR5 RAM (options up to 32 GB).
- 128 GB UFS 2.2 storage (options up to 512 GB).
- 8 PCIe 3.0 lanes, hot-pluggable, RC & EP modes.
- Video decode 4K @ 120 fps, encode 4K @ 60 fps.
- Multiple audio interfaces: 6 full-duplex IS, 4 half-duplex IS, 2 DP/eDP.
- Power input 12 V to 19 V, compact 69.6 mm 45.0 mm 4.5 mm size, 23 g weight.
- Runs Ubuntu 26.04, Fedora, OpenHarmony, and other Linux-based OS.
Specifications:
| AI Performance |
8-core A100 delivers 60TOPS AI computing power (SquareINT4), supporting BF16, FP16, FP8, INT8, INT4 and other data formats.
Equipped with RVV vector instruction set, native FP8 inference and multi-modal algorithm acceleration. Supports 30B local models
with inference performance exceeding 10Tokens/S@30B, adapting to various AI algorithms and model deployment.
|
| SoC |
SpacemiT Key Stone K3 |
| CPU |
8-core X100 64-bit RISC-V AI Processor, max main frequency 2.4GHz, 130KDMIPS general computing power,
full RVA23 Profile support, single-core Specint2006 >9.0/GHz, equivalent to ARM-A76
|
| GPU |
Integrated 3D graphics engine supporting Vulkan 1.3, OpenCL 3.0, OpenGL ES 1.1/3.2, EGL 1.5 |
| RAM |
8GB / 16GB / 32GB LPDDR5 |
| Storage |
128GB / 256GB / 512GB UFS2.2 |
| PCIe |
8 × PCIe 3.0 (8Gbps/Lane), PCIe x8 supports RC & EP modes, hot-pluggable |
| Codecs |
Video decoding: 4K@120fps H.264/VP9Video encoding: 4K@60fps H.264 |
| Power |
12V (Supports 12V~19V wide voltage input) |
| Audio |
6 × Full-duplex I²S (48kHz sampling rate)4 × Half-duplex I²S (2 connected to DP/eDP controller,
48kHz sampling rate)2 × DP/eDP audio interfaces (192kHz sampling rate)
|
| OS |
Ubuntu 26.04, Buildroot Linux with QT GUI, Fedora, Bianbu OS, Deepin, OpenKylin, OpenHarmony 6.1, OpenEuler |
| Dimension |
69.6mm × 45.0mm × 4.5mm |
| Weight |
≈ 23g |
| Environment |
Operating Temperature: -20℃～60℃Storage Temperature: -20℃～70℃Operating Humidity: 10%～90%RH (No condensation) |
| USB |
3 × USB3.0 HOST (combo with PCIe), 1 × USB3.0 DRD (Type-C, including USB2.0 OTG), 1 × USB2.0 HOST |
