Avnet Silica is now an exclusive DEEPX EMEA distributor, helping companies cut costs, save power, and bring intelligence to every device. Anywhere, anytime.
DEEPX is a leading AI hardware company revolutionising how artificial intelligence is deployed and accessed worldwide. Founded in 2015 with the vision of "AI for Everyone & Everywhere," they develop cutting-edge AI processors that bring powerful intelligence directly to edge devices, eliminating the need for cloud connectivity and data centres.
Their flagship products include the DX-M1, which delivers 25 TOPS of AI performance with exceptional power efficiency, and the DX-V3, a specialised AI Vision SoC supporting up to 12 camera streams with 13 TOPS of compute power at just 5W of consumption. These breakthrough semiconductors achieve up to 20 times better performance per watt than traditional solutions while maintaining GPU-level AI accuracy.
They collaborate with more than 300 global organisations across smart cameras, surveillance systems, autonomous vehicles, robotics platforms, and industrial automation. Operating across the US, China, Taiwan, and Europe, DEEPX is building the foundational platform for ubiquitous AI, making advanced intelligence accessible in every device, everywhere, regardless of connectivity or power constraints.
Why are we partnering?
Innovation happens when the right technologies meet the right ecosystem. That’s why DEEPX and Avnet Silica are joining forces. To accelerate the adoption of edge AI across Europe.
DEEPX delivers ultra-efficient, high-performance AI semiconductors designed for on-device intelligence. Avnet Silica has demonstrated expertise in several areas within AI over the past decade, and also offers unmatched expertise in power management, embedded systems, and industrial IoT, complemented by a robust distribution network and comprehensive engineering support.
Together, we offer:
- Complete Edge AI Solutions – Combining DEEPX processors with Avnet Silica’s power technologies for optimised performance.
- Faster Time-to-Market – Access to Avnet Silica’s design services and technical support ensures smooth integration.
- Scalable Innovation – From robotics and smart mobility to industrial automation, our partnership enables AI everywhere.
This collaboration positions both companies at the forefront of next-generation intelligent systems, where efficiency, performance, and reliability drive the future.
Featured DEEPX products
Technology
Artificial Intelligence Overview
Head over to our Artificial Intelligence overview page for more AI articles, applications and resources.
- DX-M1 M.2 LPDDR5x2
- DX-M1 AI Accelerator
- DX-V3 AI Vision SoC (Coming soon)
- DX-H1 V-NPU PCIe Card
For Every AIoT
The DEEPX DX-M1 M.2 module brings server-grade AI inference directly to edge devices. Delivering 25 TOPS of performance at just 2-5W, the module achieves 20x better performance efficiency (FPS/W) than GPGPUs while maintaining GPU-level AI accuracy.
Applications
- Edge Camera Systems: Real-time surveillance and intrusion detection without cloud latency, ensuring immediate threat response and local privacy protection
- Smart Mobility: Autonomous navigation and collision avoidance without network dependency, ensuring reliable operation regardless of connectivity conditions
- Smart Factory: On-site quality control and defect detection with superior thermal efficiency, operating reliably in harsh industrial environments without cooling systems
- Smart Cities: Live traffic management and crowd monitoring with reduced bandwidth costs, processing thousands of camera feeds locally
- Smart Retail: Real-time customer analytics and inventory management through local video processing, protecting customer privacy while optimising operations
- Smart Homes: Local video analytics and scene recognition without sending personal data to cloud servers, ensuring family privacy
- Robotics: Live traffic management and crowd monitoring with reduced bandwidth costs, processing thousands of camera feeds locally
- Drones: Lightweight, power-efficient AI processing for aerial surveillance and autonomous flight control without draining battery life
- Edge Computing: High-performance AI inference at the network edge with minimal power consumption, reducing server loads and network bandwidth requirements
| Feature | AI Accelerator | Details |
|---|---|---|
| Processor | INT8 Performance | 25 TOPS (=200 eTOPS / INT8) |
| Signal Interface | PCI Express |
PCle Gen.3 x4 / Bandwidth: 4GB/s *Compatible to PCIE x1 |
| Power | Power Consumption | 2W min., 5W max. for DEEPX supported models |
| Operating | Temperature |
-25 ~ 85°C (Throttling) -25 ~ 65°C (Non_Throttling) |
| Environment | Humidity | 40 °C @ 85% relative humidity (non-condensing) |
| Thermal Solution | Cooling | Heatsink (Option) |
| Physical |
Form Factor Dimensions Power Range |
M.2 2280 (Key M) 22mm x 80mm x 4.1mm 3.3V ± 5% |
| Software Support |
Windows Linux Framework |
Windows 11, 10 64 bit Ubuntu 22.04, 20.04 LTS Support TensorFlow, TensorFlow Lite, ONNX, Keras, PyTorch by Dataflow compiler converted |
| System Support | CPU Platform | x86, ARM Based Architecture |
Intelligence liberated: powerful, efficient, and universal at the edge
The DX-M1 AI chip is more than silicon; it is the declaration that intelligence belongs everywhere, not just in the cloud. At only 2–5W, it delivers the force of 200 GPU-class TOPS, transforming the edge from a passive receiver into an active decision-maker. at just 2-5W through breakthrough efficiency design.
Through its innovative memory design, the DX-M1 enables multiple AI models to work side by side, not as a luxury, but as a new standard. What once demanded costly, power-hungry systems is now within reach of every device, every factory, every city.
This is not a step forward in chips. It is a step forward in civilisation, where real-time intelligence becomes universal, affordable, and inevitable.
Applications
- Edge Camera Systems: Real-time surveillance and intrusion detection without cloud latency, ensuring immediate threat response and local privacy protection
- Smart Mobility: Autonomous navigation and collision avoidance without network dependency, ensuring reliable operation regardless of connectivity conditions
- Smart Factory: On-site quality control and defect detection with superior thermal efficiency, operating reliably in harsh industrial environments without cooling systems
- Smart Cities: Live traffic management and crowd monitoring with reduced bandwidth costs, processing thousands of camera feeds locally
- Smart Retail: Real-time customer analytics and inventory management through local video processing, protecting customer privacy while optimising operations
- Smart Homes: Local video analytics and scene recognition without sending personal data to cloud servers, ensuring family privacy
- Robotics: Live traffic management and crowd monitoring with reduced bandwidth costs, processing thousands of camera feeds locally
- Drones: Lightweight, power-efficient AI processing for aerial surveillance and autonomous flight control without draining battery life
- Edge Computing: High-performance AI inference at the network edge with minimal power consumption, reducing server loads and network bandwidth requirements
| Features | Details |
|---|---|
| Processor | 25 TOPS (=200 eTOPS / INT8) |
| Signal Interface | PCIe GEN3 x4 / Bandwidth: 4GB/s (*Compatible to PCIe x1) |
| Power | 2W min., 5W max. for DEEPX supported AI models |
| Operating Temperature | -25 ~ 85°C (Throttling), -25 ~ 65°C (Non-Throttling) |
| Software and Framework Support | Windows 11, 10 64-bit Ubuntu 22.04, 20.04 LTS Support Yocto Project Support TensorFlow, ONNX, Keras, PyTorch by Dataflow compiler converted |
| System Support | x86, ARM Based Architecture |
Coming soon: The compact eye of AI - efficient, perceptive, and everywhere
The DX-V3 is not simply a chip; it is the eye of the intelligent world. Delivering 13 TOPS of AI performance at only 5W, DX-V3 brings the sharpness of perception once confined to powerful systems into the smallest, most efficient form.
By uniting AI computation, image and video processing, and connectivity into one platform, DX-V3 eliminates fragmentation and makes perception both sustainable and universal. It is not just hardware; it is the compact eye of Physical AI that enables trust, vigilance, and progress everywhere.
Applications
- Edge Camera Systems: Real-time surveillance and intrusion detection without cloud latency, ensuring immediate threat response and local privacy protection
- 3D Sensing: Real-time 3D map of the environment, essential for applications requiring depth perception
- Driver Monitoring Systems (DMS): Real-time driver alertness and behaviour analysis for automotive safety compliance and accident prevention
- Autonomous Platforms: Real-time decisions without network latency for safe navigation and autonomous operation
- Robotics: Extended battery life and compact form factor enable longer autonomous missions with complex AI processing for mobile robots
- Drones: Lightweight, power-efficient AI processing for aerial surveillance and autonomous flight control without draining battery life
- AI Security: High-performance AI inference at the network edge with minimal power consumption, reducing server loads and network bandwidth requirements
- Smart Safety Monitoring: Local video analytics and scene recognition without sending personal data to cloud servers, ensuring family privacy
- Smart Retail: Real-time customer analytics and inventory management through local video processing, protecting customer privacy while optimising operations
| Features | Details |
|---|---|
| Processor | Up to 13 TOPS (INT8) for real-time AI inference |
| Powerful CPU |
Quad-core Cortex-A53 CPU @ 1.2GHz Full Linux OS support |
| Advanced ISP & Video Processing |
Integrated 12MP ISP with 3A (AE/AF/AWB), and noise reduction Supports H.264/H.265 encoding/decoding up to 4K @ 60fps |
| Multi-Channel Video Input/Output |
Captures video from 4x MIPI CSI-2 and legacy BT.656/1120 input Output via MIPI DSI, CSI, and parallel interface for display or streaming |
| High-Speed I/O & Storage | Includes USB 3.0, Gigabit Ethernet, and SD/eMMC interface |
| Edge AI Integration | All-in-one SoC optimised for compact, real-time edge AI vision |
| DSP | 75 GFLOPS (SLAM/Radar support) |
Power redefined: massive AI video at 40W, making intelligence sustainable at scale
Integrated decoding, encoding, and transcoding turn the DX-H1 V-NPU into a complete foundation for real-time, large-scale monitoring. What once required rooms full of expensive, power-hungry servers can now be placed into the hands of any organisation.
The barrier to advanced AI falls away, and what remains is possibility, accessible, sustainable, and inevitable.
Applications
- Security: Comprehensive perimeter monitoring and threat detection with advanced behavioural analysis, providing 24/7 surveillance while maintaining data privacy through local processing
- Smart Buildings: Intelligent building management with integrated security and occupancy monitoring through local video analytics, ensuring tenant privacy while reducing infrastructure requirements
- Smart Factory: On-site quality control and defect detection with superior thermal efficiency, operating reliably in harsh industrial environments without cooling systems
- Smart Cities: Real-time decisions without network latency for safe navigation and autonomous operation
- Smart Cities: Live traffic management and crowd monitoring with reduced bandwidth costs, processing thousands of camera feeds locally
- Smart Safety Monitoring: Real-time workplace safety compliance and hazard detection with instant alert systems, enabling immediate response to safety violations and emergency situations
- Smart Transportation: Autonomous navigation and collision avoidance without network dependency, ensuring reliable operation regardless of connectivity conditions
- Smart Retail: Real-time customer analytics and inventory management through local video processing, protecting customer privacy while optimising operations
- Content Delivery Network (Smart Search): Transform video content libraries into searchable databases with natural language queries, reducing search time from hours to seconds
- Logistics & Warehouses: Real-time inventory tracking and automated sorting with high-density camera deployments, enabling efficient operations and safety monitoring while reducing server infrastructure costs
| Features | Details |
|---|---|
| AI Performance | 50 TOPS (=400 eTOPS / INT8) for real-time AI inference |
| SRAM |
Genesis NPU: Internal SRAM, Codec Chip: 256 KB to 512 KB |
| CPU |
Codec Chip: 4x Cortex-A76 (up to 2.4 GHz ) Codec Chip: 4x Cortex-A55 (up to 1.8 GHz) |
| Memory |
Genesis NPU: Total = LPDDR5 8GB (4GB + 4GB) Codec Chip: Total = LPDDR5 16GB (8GB + 8GB) |
| Storage | eMMC 32GB |
| Power Consumption | 40W / 12V |
| PCIe | PCI Express Gen3 16X Using Data x8 Lanes |
| Video Decoding | H.264 64CH (1080P @ 30FPS) |
| Video Encoding | H.264 32CH (1080P @ 30FPS) |
| Display | HDMI 2.0 |
| Dimensions | 167mm x 56mm x 18.8mm |
| Host Hardware | x86, ARM-Based Architecture |
Why DEEPX?
Unmatched AI Performance
DEEPX delivers 240% higher performance than
GPGPU (ResNet-50 benchmark) while achieving
20x better power efficiency (FPS/W). This performance
-per-watt ensures reliable AI processing in power-constrained
environments without degradation or compromise.
Thermal Efficiency
DEEPX chips maintain high performance within
-25°C to 85°C, meeting industrial standards without
thermal throttling. This guarantees consistent operation
in real-world edge environments where reliability is critical.
Seamless Scalability
Designed for effortless integration, DEEPX solutions
fit IoT, robotics, and industrial automation. With
a compact footprint, low power needs, and flexible
deployment, manufacturers can scale AI without
redesigning infrastructure, accelerating innovation
with minimal complexity.
Industry-Leading TCO
Each DX-M1 saves up to 94% in electricity costs over five years compared to GPGPUs. Even if competitors offered chips for free, DEEPX remains more economical thanks to reduced power and cooling requirements, protecting your long-term investment.
GPU-Level AI Accuracy
DEEPX achieves GPU-level accuracy using optimised Int-8 quantisation instead of power-hungry FP32. While industry standards target less than 1% accuracy loss, DEEPX consistently delivers reliable predictions you can trust.
Where DEEPX outperforms the rest
Thermal efficiency improves reliability and longevity
A major advantage of DEEPX technology lies in its exceptionally low heat generation during operation. This characteristic significantly simplifies system design because thermal management is often one of the most challenging aspects of hardware integration. By minimising heat output, DEEPX reduces or even eliminates the need for bulky and costly cooling components such as fans, large heat sinks, or liquid cooling systems. This not only lowers the overall bill of materials and power consumption but also enables more compact, lightweight designs, ideal for edge devices, embedded systems, and environments where space and energy efficiency are critical. Additionally, reduced thermal complexity improves reliability and longevity, as fewer moving parts and less thermal stress mean fewer points of failure over time.
The 'Butter Benchmark' test, where DEEPX compare their AI accelerator 'DX-M1' against a competitor's AI chip from a thermal efficiency perspective.
Every TOP counts through architectural efficiency
DEEPX solutions are built with a singular focus: delivering efficient, high-performance AI inference without unnecessary complexity. At the core is a custom Neural Processing Unit designed specifically for inference workloads, avoiding the compromises of general-purpose GPUs. This purpose-built architecture is complemented by an intelligent memory design, which combines compact on-chip SRAM with high-speed LPDDR5 DRAM positioned close to the processor. The result is reduced latency and consistent performance without the drawbacks of oversized memory blocks. Powering this hardware is the DXNN® SDK, which compiles and optimises models directly for the NPU, eliminating redundant operations and maximising throughput per watt. Together, these innovations enable DEEPX products to deliver faster, leaner, and more reliable AI for real-world applications.
Through optimised architecture, every TOP counts, so fewer TOPs can match or exceed competitor chips' practical throughput for vision tasks like YOLO, segmentation, and detection.
Avnet Silica benchmark results
Avnet Silica put DEEPX's DX-M1 AI chip up against two very similar chips from well-known AI semiconductor manufacturers. We benchmarked various aspects, including latency, FPS, power usage, TOPs, and more.
| Performance | DX-M1 (DEEPX) | Product (Competitor A) | Product (Competitor B) |
|---|---|---|---|
|
Latency in ms (single) *Lower is better |
2.93 | 9.96 | 10.24 |
|
Latency in ms (multi) *Lower is better |
0.76 | 9.96 | 12.15 |
|
FPS (single) *Higher is better |
652 | 251 | 217 |
|
FPS (multi) *Higher is better |
2,528 | 251 | 711 |
|
FPS/Power (single) *Higher is better |
93 | 33 | 31 |
|
FPS/Power (multi) *Higher is better |
356 | 33 | 100 |
|
FPS/TOP (single) *Higher is better |
26.10 | 3.74 | 16.68 |
|
FPS/TOP (multi) *Higher is better |
101 | 4 | 55 |
Contact us for full benchmarking information
Interested in learning more about this benchmarking information, including the application we used to carry out the benchmarking? Speak to our AI experts!
Datasheets and resources |
Hardware
Hardware for AI
Each tier of AI deployment, from sensor nodes and edge devices to near-edge infrastructure and centralised data, presents distinct hardware considerations.
AI-Vision
Computer Vision & Recognition Systems
Computer vision systems and recognition systems must process large volumes of optical sensor data in real time, making them another category of product that can benefit from artificial intelligence in general, and machine learning (ML) in particular.
Technologies
Power
Our EMEA power FAEs are helping hundreds of businesses across EMEA shift to more efficient, smarter and cleaner power technologies. Explore our power capabilities and browse resources on wide bandgap technology, thermal management and more.
Contact us
Have a question?
Int. Freecall - 00800 412 412 11 | Product or shop-related inquiries: OnlineSupportEU@avnet.com | For anything else, head over to our contact form.