iDhi™ – Edge AI Platform
iDhi™ 8250/5165 is built using ISQ5165/ISQ8250 small-form-factor System on Modules (SoM), coupled with a customizable carrier board that integrates high-speed storage, LAN/WLAN/WWAN connectivity, and interfaces to cameras, displays, and third-party sensors.
The SoM leverages Qualcomm’s QRB5165 and QCS8250 SoCs, providing up to 15 TOPS of compute in Linux and Android environments for industrial enterprise and robotic applications.
Use Cases
- Worker Safety and Compliance
- MedTech Solutions
- Tele-Operation
- Industry 4.0
Features
- Multiple Input/Output Global
- Global Connectivity
- Optimized AI on the Edge
- Built-in Robotics
- Robust Video Processing
- Cloud Agnostic and ModelOps Capabilities
SPECIFICATIONS – iDhi™ – 8250/5165
HARDWARE
Components
Descriptions
Processor (QCS8250 / QRB5165)
Kryo 585-64bit with 4MB L3 Cache, (Kyro gold 2.419 GHz with 256KB L2)*3, (Kyro gold 2.842 GHz with 512KB L2)*1 and (Kyro silver 1.8 GHz with 128KB L2) *4
Memory and Storage
System memory: 8GB LPDDR5 POP UFS memory: 128GB
OS
Android 10, Ubuntu, ROS2
Connectivity: WWAN
4G/5G through RM500Q-GL(Option)
Connectivity: WLAN
Wi-Fi 6 – 802.11ax 2×2 MU-MIMO Wi-Fi/BT 5.1
Peripheral Interfaces
Camera: USB, DVI, HDMI MIPI – CSI
Display: DVI, HDMI, DisplayPort and MIPI – DSI
High-Speed I/O: USB 3.x, PCIe
Connectivity LAN
Gigabit Ethernet
Expandable Storage
NVME/SATA SSD slot
Hi-Performance Expansion
M.2 slot with USB3.x & 2-lane PCIe Gen 3
Antennas
2X2 MIMO for 4G/5G , 2X2 for Wi-Fi/BT
Power Input
12V DC
Video Encoding
Up to 4K 120 fps 10-bit: H.265
GPU
Adreno 650, Fmax 587MHz-2X2K 60fps UI, OpenGL ES 3.2, Vulkan1.1, DX12FL12.1, OpenCL 2.0 full profile
Digital-signal Processing/NPU
Compute DSP with Hexagon Vector extensions (dual-HVX) and Hexagon CP2.0, Dedicated NPU230
Security
SPU 240, Secure boot3.0 using Android Strongbox and iUICC, secure debug, secure key provisioning, TrustZone, Qualcomm® Trusted Execution Environment 5.0, hardware-supported Keystore Crypto engine v5 (CE5), DRBG/PRNG (FIPS- compliant), inline crypto engine (FIPS- compliant) Secure file system, Fast trusted storage, DRM support in Hardware
Audio
Built-in Audio Codec, Support for Multiple Digital MIC/MIC-Arrays, Stereo Speaker outputs, Qualcomm Fluence Pro audio signal processing (echo, noise, beamforming, beam tracking), Enhanced third party front-end audio processing
SoM Interface
Custom SMARC
Dimension
Compliance to 3GPP Standard, TEC (India) ready, CE (EU) ready, FCC (NA) ready
Peripheral Compatibility
Camera: Stryker – (DVI interface), AIDA – 100 (HDMI interface),
Customized IR
Camera (USB 3.1)
Displays: Waveshare AMOLED FHD HDMI display with touch
Storage: Western Digital NVMe SSD – SN500, SN550
WWAN: QUECTEL – RM500Q – GL (Worldwide 4G/5G support)
SOFTWARE
Components
Descriptions
OS
Andriod- 10, QCS8250LA1-0-AP-Standard-OEM/master/r00013. 1 or Ubuntu
Compute Power
15 TOPS cumulative (Octa-core APSS, DSP, NPU, GPU, DPU, VPU)
Imaging
OpenCL 2.0 profile, OpenGL 3.2 profile, Vulkan 1.1, OpenCV accelerated by fastCV
Board Support
Optimized bootloader and drivers for all listed peripherals
Package (BSP)
Fluence/Acoustic voice suite (use case like Keyword spotting, Voice-based UI)
Analytics
Object detection | Object tracking | Image segmentation
Public Cloud Connectivity
Azure IoT Hub
Device Management
Edge Agent
AWS Green grass, Azure IoT Edge, and Customized OEM specific edge agents
MLOps
AWS Sagemaker, Azure ML and iFusion
OTA Updates
Android OS updates, Security patches, Device specific application updates
SDKs
Components
Descriptions
Hexagon DSP SDK
Designed to optimize the features and performance of multimedia software
Helps allow audio, imaging, embedded vision, and heterogeneous computing acceleration on the Hexagon DSP embedded in Qualcomm Technologies processors to create compelling multimedia user experiences
Qualcomm® Neural Processing SDK
Optimizes deep learning processing performance across available resources to achieve a superior edge computing experience
Qualcomm® Computer Vision SDK
Offers a mobile-optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation
Qualcomm Spectra
480 Image Signal
Processor (ISP)
Supports up to 12 cameras by D-PHY and 18 cameras by C-PHY (seven concurrent)
SNPE SDK
Enables trained AI/ML models execution on the edge
ROS 2
Pre-integrated Robotics Operating System 2 on ISQ5165 only with selected drivers for sensors and outputs
SPECIFICATIONS – iDhi™ – 4290
HARDWARE
Components
Descriptions
Processor
(QCS4290)
Kryo 260-64bit (Kyro gold 2 GHz with 1MB L2)*4 and (Kyro silver 1.8
GHz with 512KB L2) *4
Memory and Storage
System memory: 8GB LPDDR4 + eMMC memory: 128GB
OS
Android 12
Connectivity: WWAN
4G/5G through RM500Q-GL(Option)
Connectivity: WLAN
Wi-Fi 6 – 802.11ax-ready 1×1 MU-MIMO Wi-Fi/BT 5.1
Peripheral Interfaces
Camera: USB, HDMI, MIPI – CSI
Display: HDMI
High-Speed I/O: USB 3.1
Connectivity LAN
Not Supported
Expandable Storage
Not Supported
Hi-Performance Expansion
M.2 slot with USB3.1
Antennas
2X2 MIMO for 4G/5G , 1X1 for Wi-Fi/BT
Power Input
12V DC
Video Encoding
Up to 1080p 60 fps 8-bit: H.264
Up to 1080p 60 fps 8-bit: H.265
GPU
Adreno 610 at 950Mhz , OpenGL ES 3.2, Vulkan1.1,OpenCL 2.0
Digital-signal Processing/NPU
Compute DSP with Hexagon Vector extensions (dual-HVX at 1GHz)
Image Signal Processor
Spectra 340T
Security
Secure boot, Crypto engines, Key provisioning security, Qualcomm TEE,
Qualcomm Content Protection (Widevine), Secure Periphery, Debug
security
Audio
Built-in Audio Codec (MP3; AAC; HE AAC v1, v2; FLAC; APE; ALAC;
AIFF)
SoM Interface
Custom SMARC
Dimension
SoM – 100mmX50mm. Carrier – 23.8 cm X 20 cm
Peripheral Compatibility
Camera: AIDA – 100 (HDMI interface),
Displays: Waveshare AMOLED FHD HDMI display with touch
WWAN: QUECTEL – RM500Q – GL (Worldwide 4G/5G support)
SOFTWARE
Feature
Specification
OS
Andriod or Ubuntu
Edge AI
Gstreamer/NNStreamer Framework / TF-Lite Environment / SSD Mobilenet v2 Neural Net
AI inferences/sec
Estimated – DSP: 65 inf/sec, GPU: 5 inf/sec, CPU: 1 inf/sec
Imaging
OpenCL 2.0 profile, OpenGL 3.2 profile, Vulkan 1.1, OpenCV accelerated by fastCV
Board Support
Optimized bootloader and drivers for all listed peripherals
Analytics
Object detection | Object tracking | Image segmentation
Cloud Connectivity
Microsoft Azure IoT Hub / Customized Private Cloud
Device Management
Azure IoT Edge Service, and Customized Edge agents
Edge AI
Gstreamer/NNStreamer Framework / TF-Lite Environment / SSD Mobilenet v2 Neural Net
MLOps
Azure Cognitive Cloud / Innominds iFusion
Hexagon DSP SDK
Enables audio, imaging, embedded vision, vector compute on the Hexagon DSP core
Qualcomm® Neural
Processing SDK (SNPE)
Optimizes deep learning processing performance across available compute resources. Enables trained AI/ML models execution on the edge
Qualcomm® Computer
Vision SDK
Optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation
OTA Updates
Android / Linux OS updates, Security patches, Device specific application updates
SDKs
Components
Descriptions
Hexagon DSP SDK
Designed to optimize the features and performance of multimedia software
Helps allow audio, imaging, embedded vision, and heterogeneous computing acceleration on the Hexagon DSP embedded in Qualcomm Technologies processors to create compelling multimedia user experiences
Qualcomm® Neural Processing SDK
Optimizes deep learning processing performance across available resources to achieve a superior edge computing experience
Qualcomm® Computer Vision SDK
Offers a mobile-optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation
Qualcomm Spectra
480 Image Signal
Processor (ISP)
Supports up to 12 cameras by D-PHY and 18 cameras by C-PHY (seven concurrent)
SNPE SDK
Enables trained AI/ML models execution on the edge
ROS 2
Pre-integrated Robotics Operating System 2 on ISQ5165 only with selected drivers for sensors and outputs
iDhi™ – Edge AI Platform
iDhi™ 8250/5165 is built using ISQ5165/ISQ8250 small-form-factor System on Modules (SoM), coupled with a customizable carrier board that integrates high-speed storage, LAN/WLAN/WWAN connectivity, and interfaces to cameras, displays, and third-party sensors.
The SoM leverages Qualcomm’s QRB5165 and QCS8250 SoCs, providing up to 15 TOPS of compute in Linux and Android environments for industrial enterprise and robotic applications.
Use Cases
- Worker Safety and Compliance
- MedTech Solutions
- Tele-Operation
- Industry 4.0
Features
- Multiple Input/Output Global
- Global Connectivity
- Optimized AI on the Edge
- Built-in Robotics
- Robust Video Processing
- Cloud Agnostic and ModelOps Capabilities
SPECIFICATIONS – iDhi™ – 8250/5165
HARDWARE
| Components | Descriptions |
|---|---|
| Processor (QCS8250 / QRB5165) | Kryo 585-64bit with 4MB L3 Cache, (Kyro gold 2.419 GHz with 256KB L2)*3, (Kyro gold 2.842 GHz with 512KB L2)*1 and (Kyro silver 1.8 GHz with 128KB L2) *4 |
| Memory and Storage | System memory: 8GB LPDDR5 POP UFS memory: 128GB |
| OS | Android 10, Ubuntu, ROS2 |
| Connectivity: WWAN | 4G/5G through RM500Q-GL(Option) |
| Connectivity: WLAN | Wi-Fi 6 – 802.11ax 2×2 MU-MIMO Wi-Fi/BT 5.1 |
| Peripheral Interfaces | Camera: USB, DVI, HDMI MIPI – CSI Display: DVI, HDMI, DisplayPort and MIPI – DSI High-Speed I/O: USB 3.x, PCIe |
| Connectivity LAN | Gigabit Ethernet |
| Expandable Storage | NVME/SATA SSD slot |
| Hi-Performance Expansion | M.2 slot with USB3.x & 2-lane PCIe Gen 3 |
| Antennas | 2X2 MIMO for 4G/5G , 2X2 for Wi-Fi/BT |
| Power Input | 12V DC |
| Video Encoding | Up to 4K 120 fps 10-bit: H.265 |
| GPU | Adreno 650, Fmax 587MHz-2X2K 60fps UI, OpenGL ES 3.2, Vulkan1.1, DX12FL12.1, OpenCL 2.0 full profile |
| Digital-signal Processing/NPU | Compute DSP with Hexagon Vector extensions (dual-HVX) and Hexagon CP2.0, Dedicated NPU230 |
| Security | SPU 240, Secure boot3.0 using Android Strongbox and iUICC, secure debug, secure key provisioning, TrustZone, Qualcomm® Trusted Execution Environment 5.0, hardware-supported Keystore Crypto engine v5 (CE5), DRBG/PRNG (FIPS- compliant), inline crypto engine (FIPS- compliant) Secure file system, Fast trusted storage, DRM support in Hardware |
| Audio | Built-in Audio Codec, Support for Multiple Digital MIC/MIC-Arrays, Stereo Speaker outputs, Qualcomm Fluence Pro audio signal processing (echo, noise, beamforming, beam tracking), Enhanced third party front-end audio processing |
| SoM Interface | Custom SMARC |
| Dimension | Compliance to 3GPP Standard, TEC (India) ready, CE (EU) ready, FCC (NA) ready |
| Peripheral Compatibility | Camera: Stryker – (DVI interface), AIDA – 100 (HDMI interface), Customized IR Camera (USB 3.1) Displays: Waveshare AMOLED FHD HDMI display with touch Storage: Western Digital NVMe SSD – SN500, SN550 WWAN: QUECTEL – RM500Q – GL (Worldwide 4G/5G support) |
SOFTWARE
| Components | Descriptions |
|---|---|
| OS | Andriod- 10, QCS8250LA1-0-AP-Standard-OEM/master/r00013. 1 or Ubuntu |
| Compute Power | 15 TOPS cumulative (Octa-core APSS, DSP, NPU, GPU, DPU, VPU) |
| Imaging | OpenCL 2.0 profile, OpenGL 3.2 profile, Vulkan 1.1, OpenCV accelerated by fastCV |
| Board Support | Optimized bootloader and drivers for all listed peripherals |
| Package (BSP) | Fluence/Acoustic voice suite (use case like Keyword spotting, Voice-based UI) |
| Analytics | Object detection | Object tracking | Image segmentation |
| Public Cloud Connectivity | Azure IoT Hub |
| Device Management Edge Agent |
AWS Green grass, Azure IoT Edge, and Customized OEM specific edge agents |
| MLOps | AWS Sagemaker, Azure ML and iFusion |
| OTA Updates | Android OS updates, Security patches, Device specific application updates |
SDKs
| Components | Descriptions |
|---|---|
| Hexagon DSP SDK | Designed to optimize the features and performance of multimedia software Helps allow audio, imaging, embedded vision, and heterogeneous computing acceleration on the Hexagon DSP embedded in Qualcomm Technologies processors to create compelling multimedia user experiences |
| Qualcomm® Neural Processing SDK | Optimizes deep learning processing performance across available resources to achieve a superior edge computing experience |
| Qualcomm® Computer Vision SDK | Offers a mobile-optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation |
| Qualcomm Spectra 480 Image Signal Processor (ISP) |
Supports up to 12 cameras by D-PHY and 18 cameras by C-PHY (seven concurrent) |
| SNPE SDK | Enables trained AI/ML models execution on the edge |
| ROS 2 | Pre-integrated Robotics Operating System 2 on ISQ5165 only with selected drivers for sensors and outputs |
SPECIFICATIONS – iDhi™ – 4290
HARDWARE
| Components | Descriptions |
|---|---|
| Processor (QCS4290) |
Kryo 260-64bit (Kyro gold 2 GHz with 1MB L2)*4 and (Kyro silver 1.8 GHz with 512KB L2) *4 |
| Memory and Storage | System memory: 8GB LPDDR4 + eMMC memory: 128GB |
| OS | Android 12 |
| Connectivity: WWAN | 4G/5G through RM500Q-GL(Option) |
| Connectivity: WLAN | Wi-Fi 6 – 802.11ax-ready 1×1 MU-MIMO Wi-Fi/BT 5.1 |
| Peripheral Interfaces | Camera: USB, HDMI, MIPI – CSI Display: HDMI High-Speed I/O: USB 3.1 |
| Connectivity LAN | Not Supported |
| Expandable Storage | Not Supported |
| Hi-Performance Expansion | M.2 slot with USB3.1 |
| Antennas | 2X2 MIMO for 4G/5G , 1X1 for Wi-Fi/BT |
| Power Input | 12V DC |
| Video Encoding | Up to 1080p 60 fps 8-bit: H.264 Up to 1080p 60 fps 8-bit: H.265 |
| GPU | Adreno 610 at 950Mhz , OpenGL ES 3.2, Vulkan1.1,OpenCL 2.0 |
| Digital-signal Processing/NPU | Compute DSP with Hexagon Vector extensions (dual-HVX at 1GHz) |
| Image Signal Processor | Spectra 340T |
| Security | Secure boot, Crypto engines, Key provisioning security, Qualcomm TEE, Qualcomm Content Protection (Widevine), Secure Periphery, Debug security |
| Audio | Built-in Audio Codec (MP3; AAC; HE AAC v1, v2; FLAC; APE; ALAC; AIFF) |
| SoM Interface | Custom SMARC |
| Dimension | SoM – 100mmX50mm. Carrier – 23.8 cm X 20 cm |
| Peripheral Compatibility | Camera: AIDA – 100 (HDMI interface), Displays: Waveshare AMOLED FHD HDMI display with touch WWAN: QUECTEL – RM500Q – GL (Worldwide 4G/5G support) |
SOFTWARE
| Feature | Specification |
|---|---|
| OS | Andriod or Ubuntu |
| Edge AI | Gstreamer/NNStreamer Framework / TF-Lite Environment / SSD Mobilenet v2 Neural Net |
| AI inferences/sec | Estimated – DSP: 65 inf/sec, GPU: 5 inf/sec, CPU: 1 inf/sec |
| Imaging | OpenCL 2.0 profile, OpenGL 3.2 profile, Vulkan 1.1, OpenCV accelerated by fastCV |
| Board Support | Optimized bootloader and drivers for all listed peripherals |
| Analytics | Object detection | Object tracking | Image segmentation |
| Cloud Connectivity | Microsoft Azure IoT Hub / Customized Private Cloud |
| Device Management | Azure IoT Edge Service, and Customized Edge agents |
| Edge AI | Gstreamer/NNStreamer Framework / TF-Lite Environment / SSD Mobilenet v2 Neural Net |
| MLOps | Azure Cognitive Cloud / Innominds iFusion |
| Hexagon DSP SDK | Enables audio, imaging, embedded vision, vector compute on the Hexagon DSP core |
| Qualcomm® Neural Processing SDK (SNPE) |
Optimizes deep learning processing performance across available compute resources. Enables trained AI/ML models execution on the edge |
| Qualcomm® Computer Vision SDK |
Optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation |
| OTA Updates | Android / Linux OS updates, Security patches, Device specific application updates |
SDKs
| Components | Descriptions |
|---|---|
| Hexagon DSP SDK | Designed to optimize the features and performance of multimedia software
Helps allow audio, imaging, embedded vision, and heterogeneous computing acceleration on the Hexagon DSP embedded in Qualcomm Technologies processors to create compelling multimedia user experiences |
| Qualcomm® Neural Processing SDK | Optimizes deep learning processing performance across available resources to achieve a superior edge computing experience |
| Qualcomm® Computer Vision SDK | Offers a mobile-optimized computer vision (CV) library, enabling use cases like object detection, object tracking, & image segmentation |
| Qualcomm Spectra
480 Image Signal Processor (ISP) |
Supports up to 12 cameras by D-PHY and 18 cameras by C-PHY (seven concurrent) |
| SNPE SDK | Enables trained AI/ML models execution on the edge |
| ROS 2 | Pre-integrated Robotics Operating System 2 on ISQ5165 only with selected drivers for sensors and outputs |