Business Fortune
Founded in 2018, Untether AI specializes in AI inference acceleration tailored for energy efficiency across diverse environments, from edge devices to cloud platforms. Their innovative approach revolves around an at-memory architecture for deep learning, strategically placing processing capabilities alongside memory to minimize power consumption and data movement. This pioneering technology liberates enterprises from traditional data center constraints, enabling them to deploy AI workloads flexibly and efficiently across various operational landscapes.
While competitors in the AI inference accelerator sector often make bold claims, Untether AI distinguishes itself through product innovations and a commitment to practical applications.
Revolutionizing AI Inference with High-Performance Accelerator Solutions
Untether AI offers cutting-edge AI inference solutions designed to meet the demands of OEMs and on-premise data centers, enabling deployment of artificial intelligence across diverse environments. Leveraging their innovative at-memory architecture, Untether AI delivers superior performance, unparalleled energy efficiency, and cost-effectiveness for AI applications.
AI Inference ICs for Versatile Deployment
Untether AI's AI inference ICs empower OEMs and data centers with high-throughput, energy-efficient AI capabilities. By minimizing data movement and optimizing power consumption, these ICs ensure optimal performance across various neural network models and deployment scenarios.
AI Accelerator Cards: Power and Efficiency Combined
Untether AI's AI accelerator cards translate the robust performance of their ICs into a PCI-Express form factor suitable for high-demand inference tasks. These cards excel in delivering exceptional performance while maintaining industry-leading energy efficiency, making them ideal for critical applications in edge computing and data centers.
tsunAImi tsn200: Efficiency at the Edge
The tsunAImi® tsn200 AI accelerator card integrates the runAI200 device, offering unmatched performance and power efficiency for edge applications like video analytics. With a low-profile design and 75-watt TDP, the tsunAImi tsn200 sets new standards in efficiency, ensuring optimal performance even in power-sensitive environments.
tsunAImi tsn800: High-Performance Computing for Data Centers
Designed for high-performance servers and data centers, the tsunAImi® tsn800 AI accelerator card features four runAI200 devices, delivering over 2 PetaOps of INT8 performance. With a 300-watt TDP, this PCIe card excels in processing vision-based CNNs, natural language processing tasks, and financial analytics with unmatched compute density and efficiency.
Market Applications: Vision AI, Government, Financial
Untether AI's products cater to diverse sectors including Vision AI, Government, and Financial industries, empowering organizations to harness AI capabilities efficiently and effectively. Their solutions drive innovation and enhance operational performance across critical applications.
Unlocking Seamless AI Development with the imAIgine® Software Development Kit
Untether AI introduces the imAIgine® Software Development Kit (SDK), revolutionizing the deployment of AI models from training to inference across its cutting-edge hardware solutions. This comprehensive SDK simplifies the integration process for developers, ensuring smooth transitions from model development to high-performance inference.
Seamless Integration with Leading Frameworks
The imAIgine® SDK offers seamless integration with industry-standard frameworks such as TensorFlow, PyTorch, and ONNX. Developers can import trained models effortlessly with a single click, leveraging the SDK's capabilities to optimize performance and accuracy for deployment on tsunAImi® accelerator cards and runAI® family devices.
Accelerated Model Deployment
With its Model Garden and custom kernel development flow, the imAIgine® SDK accelerates the deployment of custom neural networks tailored for high-performance computing applications. This streamlined process empowers developers to efficiently translate pilot models into robust production-ready solutions.
Advanced Optimization Tools
The imAIgine® SDK includes powerful automated tools like the imAIgine Compiler and Toolkit, designed to enhance efficiency and performance during the development lifecycle:
- imAIgine Compiler: Import TensorFlow or PyTorch graphs directly and optimize them without compromising accuracy. Developers can specify performance metrics, silicon utilization, and power consumption targets to achieve optimal deployment conditions.
- imAIgine Toolkit: Evaluate functionality and performance using extensive profiling and simulation tools, ensuring applications meet stringent performance requirements before deployment.
- imAIgine Runtime: Provides a C-based API for seamless integration into deep learning applications, monitoring tsunAImi® acceleration cards' health and temperature to maintain operational integrity and prevent thermal issues.
Facilitating Rapid Development Cycles
From pilot models to production, the imAIgine® SDK supports rapid development cycles with robust automation and software support. Developers benefit from accelerated deployment timelines, enabling faster time-to-market for AI-driven applications.
Empowering AI Innovation
The imAIgine® SDK empowers developers to harness the full potential of AI by simplifying complex workflows and optimizing performance across Untether AI's hardware solutions. With its user-friendly interface and advanced capabilities, the SDK drives innovation in AI development, setting new standards in efficiency and effectiveness.
Chris Walker| CEO
Chris Walker is an innovative and growth driven semiconductor executive having led major initiatives and business units throughout his 30 year career at Intel corporation. His expertise extends across operations, finance, business development/M&A, product, and marketing leadership. Recently he has been a senior advisor with Cerberus Capital supporting investments in technology fund and management consulting.
.webp)
