The Hailo-8™ edge AI processor, featuring up to 26 tera-operations per second (TOPS), significantly outperforms all other edge processors. Its area and power efficiency are far superior to other leading solutions by a considerable order of magnitude – at a size smaller than a penny even including the required memory. With an architecture that takes advantage of the core properties of neural networks, our neural chip allows edge devices to run deep learning applications at full scale more efficiently, effectively, and sustainably than other AI chips and solutions, while significantly lowering costs.
Its features:
- Holistic solution comprised of hardware and software
- Proprietary novel structure-driven Dataflow architecture
- Unprecedented performance, enabling:
- Processing of FHD stream in real-time (without down-sampling)
- Typical power consumption of 2.5W
- High power efficiency
- Smaller size and lower solution cost
- Can be used as a standalone processor or as a co-processor
- Scalable performance via chip cascading
- Easy hardware integration (no additional external memory)
- Seamless integration with popular software frameworks
- Automotive and Industrial grade solution
The Hailo-8™ M.2 Module is an AI accelerator module for AI applications, compatible with NGFF M.2 form factor M, B+M and A+E keys. The AI module is based on the 26 tera-operations per second (TOPS) Hailo-8™ AI processor with high power efficiency. The M.2 AI accelerator features a full PCIe Gen-3.0 2-lane interface (4-lane in M-key module), delivering unprecedented AI performance for edge devices. The M.2 module can be plugged into an existing edge device with M.2 socket to execute in real-time and with low power deep neural network inferencing for a broad range of market segments. Leveraging Hailo’s comprehensive Dataflow Compiler and its support for standard AI frameworks, customers can easily port their Neural Network models to the Hailo-8 and introduce high-performance AI products to the market quickly.
Its features:
- AI processor: Hailo-8™ AI processor with up to 26 TOPS and best-in-class power efficiency
- Supported Frameworks: TensorFLow, TensorFlow Lite, ONNX, Keras, Pytorch
- Form factor: NGFF M.2 Key M, NGFF M.2 Key B+M, NGFF M.2 Key A+E
- Dimensions: 20 x 42 mm with breakable extensions to 22 x 60 and 22 x 80 mm, 22 x 30 mm
- Interface: PCIe Gen-3.0, 4-lanes (up to 32Gbs), PCIe Gen-3.0, 2-lanes (up to 16Gbs)
- Supported OS: Linux, Windows
- Certification: CE, FCC
The Hailo-8R™ Mini PCIe Module is an AI accelerator module for AI applications, compatible with PCI Express Mini (mPCIe) form factor. The Mini PCIe AI accelerator delivers industry-leading AI performance for edge devices – up to 13 tera-operations per second (TOPS) with industry-leading power efficiency. The module can be plugged into an existing edge device with an mPCIe Full-Mini socket to execute in real-time and with low-power deep neural network inferencing for a broad range of market segments. Leveraging Hailo’s comprehensive Dataflow Compiler and its support for standard AI frameworks, customers can easily port their Neural Network models to the Hailo-8™ and introduce high-performance AI products to the market quickly.
Its features:
- Performance: up to 13 TOPS
- Best-in-class power efficiency
- Supported Frameworks: TensorFlow, TensorFlow Lite, Keras Pytorch, ONNX
- Form factor: PCI Express Mini (mPCIe)
- Dimensions: 30 x 50.95 mm
- Interface: PCIe Gen-3.0, 1-lane
- Supported OS: Linux, Windows
- Certification: CE, FCC
Hailo-8™ starter kit
Intelligent Video Analytics (IVA) has penetrated many applications across industry verticals. While penetration growth is high, ease of development remains a challenge in the implementation of advanced IVA in edge devices.
Hailo-8™ AI accelerator module can be used to enhance existing AI-enabled IVA applications and for the development and prototyping of intelligent video analytics solutions thanks to its capabilities:
- Delivers data center-class performance to edge devices with out-of-the-box passive thermal solutions for most system form factors
- Best-in-class real-time performance utilizing the Hailo-8™ 26 TOPS compute power
- Industry-leading power efficiency with a typical power consumption of 5W
- Highest cost-efficiency (TOPS/$) compared with existing solutions
- Robust software toolchain supports state-of-the-art deep learning models and applications out-of-the-box
- Enables fast time to market using M.2 standard form factor modules
- High user privacy by processing all deep learning models locally and transferring to cloud-only partial (if at all) of the captured video
Suitable for various applications:
- Edge AI Boxes and video analytics servers for Smart City, Intelligent Transportation System (ITS), Security, Surveillance, Smart Retail, Smart Building, and more
- Edge servers, industrial PCs, and gateways
- Industrial and commercial robots
- In-Vehicle Computers and ADAS/AV sensing
- The Hailo-8™ starter kit includes:
- Hailo-8™ M.2 Module (key M, key A+E or key B+M)
- Thermal management kit includes the metal cubes, thermal pads, and thermal adhesive tape
- Screw and screwdriver
Its features:
- M.2 module with Hailo-8™ AI accelerator processor
- Optional form-factors: M.2 key M, key B+M, key A+E
- Can be plugged into an existing edge device with an M.2 socket to execute deep learning inferencing in real-time
- Unprecedented performance, enabling:
- Advanced analytics algorithms can run on high-resolution & high-frame-rate video, resulting in
- High accuracy and ability to detect small objects
- Detection of fast-moving objects
- Significant reduction in false alerts
- Generation of comprehensive metadata
- Process video and generate alerts in real-time, allowing to take immediate actions
- Process simultaneously multiple advanced AI applications and multiple video streams
- Save on transmission to cloud and cloud storage & services usage for AI analytics
- High power efficiency
- Fanless, consumes 3.3W when running ReSent-50 with input resolution of 224×224 @ 915FPS
- Comprehensive software suite suitable for various computer vision tasks like object detection, object classification, semantic segmentation, post estimation, face detection, instance segmentation, depth estimation, and facial landmark
Hailo-8™ century evaluation platform
Prototype & Develop high-performance, high-efficiency edge AI video analytics systems with multiple Hailo-8™ AI processors
Today’s intelligent vision systems are high-resolution, high-frame-rate and multi-sensor, requiring more powerful and accurate real-time AI processing power.
To support developers of high-performance edge video analytics systems across industry verticals, the Hailo industry-leading Hailo-8™ AI processor is now available in a multi-chip configuration delivering up to 104 TOPS of AI performance at a typical power consumption of 25W. Designed for development, prototyping and evaluation, this standard form factor PCIe card is supported by the robust Hailo software toolchain and a suite of dedicated AI developer tools and out-of-the-box applications.
Its features:
- 4x Hailo-8 M.2 22X42 mm Acceleration Modules on a 16-lane PCIe Gen 3 adapter board
- Dimensions: 111mm x 210mm
- Standard PCIe form factor enters slot with no auxiliary power requirement
- Deliver up to 104 TOPS of AI performance at a typical power consumption of 25W
- Best-in-class PCI AI acceleration card – better performance than GPU, at lower power and lower cost
- Dedicated enablement S/W package and AI developer tools, with out-of-the-box support for state-of-the-art NN models
- Designed to enable development & prototyping of high-performance and efficiency real-time AI video analytics solutions, such as high-throughput Edge AI Boxes, dedicated video edge servers, industrial PCs and gateways
Hailo-8™ Century is also available as a reference design for our partners to design their own PCIe AI accelerator cards with up to 6 Hailo-8™ AI processors.
