Specifications

Overview

The LattePanda Mu Ultra is a credit-card-sized x86 compute module purpose-built for on-device AI. It keeps the 69.6 x 60 mm footprint of the earlier LattePanda Mu (Alder Lake-N) module but moves to Intel Core Ultra 200V "Lunar Lake" silicon, combining CPU, GPU and NPU into a single module that reaches up to 115 TOPS of combined peak INT8 AI performance.

Two processor options are offered. The Core Ultra 5 226V pairs eight cores and threads running up to 4.5 GHz with an Intel Arc 130V GPU (7 Xe cores, up to 53 TOPS INT8) and an Intel AI Boost NPU rated up to 40 TOPS, for 97 TOPS overall. The Core Ultra 7 256V lifts the CPU to 4.8 GHz, the GPU to Arc 140V with 8 Xe cores and 64 TOPS, and the NPU to 47 TOPS, reaching 115 TOPS combined. Both variants carry 16 GB of on-package LPDDR5X-8533 memory with up to 11.6 GB allocatable as VRAM for larger language models and KV cache — a key enabler for running local LLMs and vision models without a discrete card.

Integration is designed to be familiar to embedded teams: rich expansion pins expose PCIe 4.0, USB 3.2 Gen 2 and HDMI/DisplayPort, an AMI UEFI BIOS on 256 Mbit SPI flash handles boot, and Windows 11 and Ubuntu 24.04 are both supported. The module is largely compatible with existing LattePanda Mu carrier boards, and DFRobot's DFR1293 Mini Carrier Board adds an M.2 2230/2280 NVMe socket, HDMI 2.0 output, dual DisplayPort over USB-C, audio and networking, turning the module into a pseudo-single-board computer. Target applications include robotics, industrial automation, portable instruments, vision AI and edge AI terminals.