ArmSoM
RK3576 AI SINGLE-BOARD COMPUTER

ArmSoM Sige5

Build at the edge.
Keep your options open.

Octa-core RK3576 compute and 6 TOPS AI, paired with dual Gigabit Ethernet, WiFi 6 and NVMe expansion for connected prototypes and embedded applications.

8 coresApplication processor
6 TOPSIntegrated NPU
16 GBMax. memory option
WiFi 6Onboard wireless
COMPUTE. CONNECT. CREATE.SIGE SERIES
Official ArmSoM Sige5 RK3576 product banner
Dual Gigabit92 × 62mm

Application compute and practical I/O.
A compact board for your next build.

ENGINEERED FOR YOUR NEXT BUILD

A complete board. More room to build.

Combine application processing, network connections and expansion in one development platform.

8-core application compute

The RK3576 processor runs your Linux applications, services and embedded development tools.

4 × Cortex-A72 up to 2.2GHz + 4 × Cortex-A53 up to 1.8GHz

6 TOPS on-device AI

Use the Rockchip NPU toolchain to convert supported models and validate inference on the target board.

6 TOPS · integrated NPU

Two wired network connections

Connect separate network segments or develop data and control services using the onboard Ethernet ports.

2 × Gigabit Ethernet

Expand beyond onboard storage

Use the M.2 Key M connection for NVMe storage alongside onboard eMMC and a MicroSD card.

M.2 Key M · PCIe 2.1 ×1

Connect your display

Develop a desktop, media application or information display with the documented hardware video resources.

4K display applications

Keep the workbench connected

Onboard wireless, USB and the 40-pin header provide familiar connections for development and peripherals.

WiFi 6 + Bluetooth 5.3 (SYN43752)
ROCKCHIP · 8nm PROCESS

RK3576

Application processing with dedicated media and AI resources.

CPU8 application cores
GPUMali-G52 MC3
NPU6 TOPS
MemoryUp to 16GB
NetworkingDual Gigabit
StorageeMMC + NVMe
BUILT TO WORK TOGETHER

Bring the parts of your application together.

Sige5 combines the RK3576 application processor with onboard memory, storage interfaces and wired and wireless connectivity.

Use the CPU for application logic and dedicated media or inference hardware where supported by your software. Choose a documented operating-system image, then validate your peripherals and workload.

Explore the hardware specifications ↗
FROM MODEL TO APPLICATION

Put AI to work on the board.

Develop an on-device inference workflow with the Rockchip RKNN toolchain.

6TOPS

Integrated NPU · INT8 compute

RKNNLocal inferenceEdge applications
  1. 01

    Prepare your model

    Check model format and operator compatibility with the target Rockchip NPU.

  2. 02

    Convert and validate

    Use the matching RKNN tools to convert, quantize and check the model.

  3. 03

    Integrate your application

    Run inference on the board and test latency, accuracy and thermal behavior.

NPU compute capability is not an application benchmark. Results depend on model, precision, software and cooling.

CONNECT YOUR DESIGN

Ready for the rest of your system.

Network, storage, displays and peripheral development on a compact SBC.

Dual Ethernet

Dual Gigabit

WiFi 6 + Bluetooth

Onboard wireless

M.2 Key M

NVMe storage expansion

HDMI output

4K display output

USB host + OTG

Peripheral and programming I/O

40-pin GPIO

UART / SPI / I²C / I²S / PWM

MicroSD + eMMC

System and data storage

Fan + RTC

Cooling and clock connections

ONE PLATFORM. MANY DIRECTIONS.

What will you build?

A flexible starting point for developers, makers and embedded product teams.

01

Edge applications

Combine local processing, network data and supported AI inference.

02

Media & displays

Develop a desktop, media player or information display with hardware video processing.

03

Network services

Build connected-device services and storage-backed local applications.

04

Embedded prototyping

Bring up sensors, peripherals and software before product integration.

THE DETAILS THAT MATTER

Technical specifications

Use this overview for selection and the full documentation for design-level details.

ArmSoM Sige5 specifications
SpecificationArmSoM Sige5
ProcessorRK3576 · 8nm
CPU4 × Cortex-A72 up to 2.2GHz + 4 × Cortex-A53 up to 1.8GHz
GPUMali-G52 MC3
NPU6 TOPS (INT8)
MemoryLPDDR4x options up to 16GB; see store for available configurations
Storage32GB / 64GB / 128GB eMMC options; MicroSD and M.2 NVMe expansion
NVMe expansionM.2 Key M · PCIe 2.1 ×1
Wired networking2 × Gigabit Ethernet
WirelessWiFi 6 + Bluetooth 5.3 (SYN43752)
Video decodingUp to 8K30 or 4K120; H.264 / H.265 / VP9 / AV1 / AVS2
Video encodingUp to 4K60 H.264 / H.265 and JPEG
Display outputsHDMI 2.1 up to 4K120; MIPI DSI up to 2K60; DP 1.4 up to 4K120
USBUSB host and Type-C OTG; consult board revision and image documentation
Expansion40-pin GPIO: UART / SPI / I²C / I²S / PWM; pin and voltage verification required
Other connectionsFan and RTC battery connectors; headphone and HDMI audio
PowerDedicated Type-C power input; follow the Sige family power guide
Dimensions / weight92 × 62 × 14.6mm / 43g
Operating temperature0°C to 80°C
PCB8-layer PCB

Board revisions and software images can affect interface modes. Current memory/storage options are listed in the store; check the product documentation before integration.

KEEP BUILDING

Documentation & developer resources

Hardware files, system images and source code, all within reach.

DebianAndroidBuildrootArmbian

Operating-system versions and supported peripherals vary by image. Use the current product documentation to choose a release.

BEFORE YOU BUILD

Frequently asked questions

What processor does Sige5 use?

Sige5 uses the Rockchip RK3576 with 4 × Cortex-A72 up to 2.2GHz + 4 × Cortex-A53 up to 1.8GHz. It is a complete single-board computer, with onboard network and display connections.

Can I expand the storage?

Yes. The documented storage interfaces include onboard eMMC, MicroSD and M.2 Key M · PCIe 2.1 ×1. Check mechanical clearance, SSD compatibility and system support before installation.

Which Type-C port is for power?

Use the dedicated power input identified in the hardware diagram. The other Type-C connector serves data or OTG functions. Follow the Sige family power guide.

Can I attach Raspberry Pi accessories?

The 40-pin header exposes familiar peripheral signals. Check the exact pin assignment, voltage, driver and device-tree configuration before attaching an accessory.

How do I start with AI inference?

The documented NPU capability is 6 TOPS. Use the matching Rockchip toolchain and image for supported models, then validate model accuracy and real application performance.

YOUR NEXT PRODUCT STARTS HERE

Build with ArmSoM Sige5.

Explore available options or talk to us about your embedded application.