Innodisk EXEC-Q911 COM-HPC Mini kit is powered by a Qualcomm Dragonwing IQ-9075 Edge AI SoC

Innodisk has recently introduced the EXEC-Q911, a COM-HPC Mini starter kit designed for edge AI applications powered by a Qualcomm Dragonwing IQ-9075 System-on-Chip (SoC), also known as the QCS9075, which delivers up to 200 TOPS of AI performance.

The platform includes 36GB LPDDR5X memory and 128GB UFS 3.1 storage, supports dual 2.5GbE, two 4-lane MIPI CSI-2 camera interfaces, DisplayPort 1.2 and eDP outputs, and multiple expansion options via M.2 (PCIe Gen4 x4/x2) slots. Designed for industrial applications, it operates in a wide -40°C to 85°C operating temperature range, accepts 9–36V DC input, integrates TPM 2.0 security, and offers various I/Os, USB 3.2 Gen 2, CAN FD, RS-232/422/485, GPIO, SPI, and I²C, with longevity guaranteed until 2038.

Innodisk EXEC-Q911 specifications:

• Compute-on-Module – Innodisk EXMA-Q911 COM-HPC Mini module
  • SoC – Qualcomm DragonWing IQ-9075
    • CPU
      • Octa-core Kryo Gen 6 (Cortex-A78C-based) application cores @ up to 2.36 GHz
      • Quad-core Cortex-R52 real-time cores @ up to 1.85GHz
    • GPU – Adreno 663 GPU delivering up to 1.2 TFLOPS FP32 with secure GPU compute; supports Vulkan 1.2, OpenGL ES 3.2, OpenCL 2.0 FP, Adreno NN Direct
    • VPU – Adreno VPU 765
      • Video Decode
        • AV1 / HEVC / H.265 / H.264 / VP9 / MPEG-2
        • 1x 8Kp60 / 2x 8Kp30 / 4x 4Kp60 / 8x 4Kp30 / 16x 1080p60 / 32x 1080p30
      • Video Encode:
        • H.264 / H.265 / HEIF / HEIC
        • 2x 4Kp60 / 4x 4Kp30 / 8x 1080p60 / 16x 1080p30
      • Concurrent 2x 4Kp60 encode + 2x 4Kp60 decode
    • NPU
      • 200 TOPS @ INT8 (Sparse Computing Power)
      • Hexagon Tensor Processor (HTP) with quad HVX and dual HMX
      • Supports deep-learning frameworks: TensorFlow, PyTorch, ONNX, Paddle, Caffe, DarkNet, etc.
  • System Memory – 36GB LPDDR5X (MT62F3G32D8DV-023)
  • Storage
    • 128GB UFS 3.1 flash (MTFC128GBCA VTC-AIT)
    • 32GB eMMC storage (SDINBDA6-32G-ZA1)
    • 32MB NOR Flash (MX25U25643G)
  • Networking –  RTL8221 LAN PHY
  • USB – USB 2.0 hub controller (GL854G)
  • Security – TPM 2.0 via NPCT760AABYX
  • Host interface – Standard 400-pin connector for I/Os
• Carrier board
  • Storage – M.2 Key-M 2280 socket for NVMe SSD
  • Video Output
    • 2x DisplayPort 1.2
    • eDP interface (40-pin connector, 0.5 mm pitch)
  • Camera I/O
    • 2x 4-lane MIPI-CSI 22-pin FPC connectors on the module
    • Supports GMSL cameras via GMSL to MIPI-CSI adapter board (not confirmed by the company)
  • Audio
    • 3.5 mm audio jack (Mic In/Line Out) via WM8904  audio codec
    • Stereo Speaker output via 1 x 4-pin header, internal NAU8223YG 3.1w amp
  • Networking
    • 2x 2.5GbE RJ45 Ethernet ports
    • M.2 Key-B 3052 socket for Wi-Fi/Bluetooth
    • M.2 Key-B socket and Nano SIM card slot for 4G LTE/5G cellular
  • USB
    • 1x USB 3.2 Gen 2 Type-C port with OTG support (on the backside)
    • 3x USB 3.2 Gen 2 Type-A ports (on the backside)
    • 2x USB 2.0 Type-A ports
    • 1× USB 2.0 via 8-pin 1.0 mm-pitch internal header
  • Serial
    • 1× RS-232 / RS-422 / RS-485 port via 9-pin header on the carrier board via F81439A chip
    • 1× CAN-FD interface (3-pin header, 2.0 mm pitch, TPT1042V controller)
  • Expansion
    • M.2 Key-M 2280 socket (PCIe Gen 4 x4) for SSD
    • M.2 Key-B 3052 socket (USB 3.2 Gen 2, USB 2.0) typically for 5G/LTE
    • M.2 Key-E 2230 socket (PCIe Gen 4 x2, USB 2.0) for Wi-Fi/Bluetooth
    • 8× GPIO via 10-pin header (1.0 mm pitch, includes VCC and GND, up to 0.5 A)
    • 1× general-purpose SPI and I2C via a single 12-pin header (2.0 mm pitch, 3.3 V)
  • Debug – UART debug port (3-pin header, 2.54 mm pitch, 3.3 V)
  • Misc
    • Emergency Download mode (EDL) jumper (3-pin header, 2.0 mm pitch)
    • CPU fan connector (4-pin, 2.54 mm pitch, PWM control, 12 V)
    • Boot mode selection toggle switch (4-pole)
    • Watchdog Timer
    • RTC with CR2032 coin cell
    • Power and reset buttons
    • PWM fan connector 12V
    • CPU temperature, fan speed, module voltage, and current monitor
• Power Supply
  • 9V to 36V DC via 2-pin pluggable terminal block (12V @ 2.6A typical)
  • GMSL camera power header (4-pin, 2.0 mm pitch, 12 V / 3 A)
  • eDP power jumper
• Dimensions – 146 x 102 x 56.99 mm (Carrier Board with COM-HPC Mini module)
• Weight – 1704 grams (with heatsink and fan)
• Temperature – Operating: -40°C to 85°C; storage: -40°C to 95°C
• Humidity – Operating & Storage: 5% to 95%, non-condensing
• Shock – 30G, IEC 60068-2-27, Half Sine, 11ms Duration (Operating)
• Vibration – 3G, IEC 60068-2-64, Random Vibration, 5 ~ 500Hz, 1hr/axis (Operating)
• ESD – EN61000-4-2, Air-15kV, Contact-8kV

Note:  Marked in bold in the specifications list are the features that differ from the earlier EVT (Engineering Validation Test) design. Basically, if you managed to get your hands on an early prototype unit, it won’t perfectly match the final mass-production specs. The finalized design adds OTG support to the USB Type-C port, tweaks the pinouts for the internal SPI, I2C, and USB 2.0 headers, and locks in the specific configurations for the COM ports, boot mode toggle switch, and Nano SIM slot.

Qualcomm has introduced a new brand name, “Dragonwing,” for its industrial processors to distinguish them from the consumer “Snapdragon” line. Under this new structure, the QCS9075 has been renamed to IQ-9075 (part of the Dragonwing IQ9 series). The chip silicon itself remains the same, but the marketing name has shifted to “Dragonwing IQ-9075” to indicate it is part of their top-tier industrial portfolio.

On the software side, the EXEC-Q911 supports Yocto Linux (Kernel 6.6) and Ubuntu 24.04 (Kernel 6.8). Innodisk provides the IQ Studio toolkit on GitHub, which gives access to BSPs, benchmarking tools, and AI sample applications distributed as Docker images or IPK packages.

For AI workloads, the module uses the Qualcomm AI Stack and Qualcomm AI Hub, which gives access to over 100 validated models such as YOLO, ResNet, and UNet-Seg. Additionally, it supports frameworks, including PyTorch, TensorFlow, TensorFlow Lite, and ONNX via the ONNX Runtime (ORT) or QAIRT (QNN) to accelerate inference on the NPU, CPU, and GPU.

We recently wrote about the Quectel SP895BD-AP AIoT module, which features the Qualcomm “Dragonwing” Q-8750 (QCS8550) and an 80 TOPS NPU, but Innodisk is targeting even higher performance with the industrial-grade Qualcomm Dragonwing IQ-9075 COM-HPC Mini module. We also wrote about other COM-HPC Mini modules which use Intel Core Ultra processors, such as the ADLINK COM-HPC-mMTL (Meteor Lake) and AAEON HPC-ARHm (Meteor / Arrow Lake). These platforms are designed for general-purpose workloads that support up to 64GB of RAM, and offer combined CPU, GPU, and NPU AI performance in the 34 to 96 TOPS range. In comparison, Innodisk’s EXEC-Q911 module offers up to 100 TOPS (Dense) from its dedicated Hexagon NPU alone, making it better suited for power-efficient, AI-centric edge workloads.

The Innodisk EXEC-Q911 starter kit (EXEC-Q911-00A1-W1) is now sampling. Pricing information has not been publicly disclosed, but you can find more details on the product page. 

Debashis Das

Debashis Das is a technical content writer and embedded engineer with over five years of experience in the industry. With expertise in Embedded C, PCB Design, and SEO optimization, he effectively blends difficult technical topics with clear communication

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.