Specifications
SoC
ESWIN EIC7700X
CPU
4× SiFive P550 (RISC-V RV64GBCH), up to 1.8 GHz
Microarchitecture
13-stage, triple-issue, out-of-order, ECC (SECDED)
Cache
32 KB L1 I + 32 KB L1 D per core, 256 KB L2, 4 MB shared L3
NPU
19.95 TOPS@INT8, INT16 / FP16 support
GPU
IMG AXM-8-256 — 0.25 TFLOPS, 1 TOPS, 8 Gpixels; OpenGL ES 3.2, Vulkan 1.2, OpenCL 1.2/2.1 EP2
Memory
8 / 16 / 32 GB LPDDR5 6400 MT/s
Storage
SPI flash, M.2 SATA SSD, eMMC connector, microSD, SATA3 connector
Expansion
1× PCIe x8 connector (PCIe Gen3 x4 signalling)
Discrete GPU Support
Validated up to AMD Radeon RX 7900 XTX
Networking
2× Gigabit Ethernet
Wireless
1× M.2 E-Key (SDIO + UART) for Wi-Fi / BT module
USB
4× USB 3.0 HOST + 2× front-panel USB3
Display
1× HDMI 2.0
Video
H.265 decode up to 8K@50fps or 32× 1080p@30; H.265 encode up to 8K@25fps; JPEG up to 32K × 32K
Virtualization
RISC-V Hypervisor (H) extension
AI Toolchain
PyTorch, TensorFlow, PaddlePaddle, ONNX; high-precision LLM inference
Form Factor
Mini-ITX, 170 × 170 mm
Power
Standard 24-pin ATX or 12V DC (5.5 × 2.5 mm)
Overview
The Milk-V Megrez is a Mini-ITX "RISC-V AI PC" built on the ESWIN EIC7700X SoC. Four SiFive P550 cores at up to 1.8 GHz, a 19.95 TOPS INT8 NPU and an IMG AXM-8-256 GPU bring local LLM and computer-vision inference to a standard desktop form factor.
It fits standard Mini-ITX cases and ATX power supplies, exposes a PCIe x8 connector validated with AMD Radeon RX 7900 XTX graphics, and offers four storage paths (microSD, eMMC, M.2 SATA and SATA3). The Jiachen Project ecosystem supplies distributions and packages.
Key Benefits
19.95 TOPS of local NPU inference with a quad-core SiFive P550 host, 8K video pipelines and discrete-GPU support — a native RISC-V desktop that runs on-device generative AI workloads without an x86 host.
Applications
On-device LLM and generative AI, computer vision and video analytics, RISC-V desktop and software development, edge inference nodes, and educational AI platforms.
Request a Quote — MILK-V MEGREZ — RISC-V AI PC MINI-ITX BOARD
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →