Products
DM2 Series
This solution represents an ideal choice for organizations seeking to implement advanced AI capabilities while working within space constraints, without compromising on performance or security. Its evolution from an OCP project to a commercial product demonstrates its versatility and effectiveness in meeting the growing demand for compact, powerful AI acceleration solutions. The DM2 effectively bridges the gap between enterprise-grade AI capabilities and space-efficient computing, making it a compelling choice for organizations looking to embrace AI technology while optimizing their physical infrastructure footprint.
Overview
Transforming Enterprise AI
from Meta to Mainstream
from Meta to Mainstream
The DM2 represents a cutting-edge enterprise GenAI solution, originally designed for Meta's OCP (Open Compute Project) and now adapted to meet the growing demand for AI PC applications. This compact AI accelerator card stands out in the market by offering enterprise-grade AI capabilities in a more space-efficient form factor.
Core Technology:At its heart, the DM2 features a powerful Generative AI Processor designed to optimize LLM (Large Language Model) performance. The processor efficiently offloads over 90% of AI workload from the host CPU, significantly enhancing overall system performance. The card includes an advanced Embedded Engine that delivers more than 10x improvement in vector similarity searches, crucial for RAG (Retrieval-Augmented Generation) processes.
Memory and Storage:Despite its compact size, the DM2 maintains substantial memory capacity with advanced vector database capabilities. It employs innovative quantization techniques that compress vector data by a factor of four, maximizing storage efficiency. This on-board memory serves as a secure, local vector database, enabling fast and efficient data processing while ensuring sensitive information remains protected.
Specification
| DM2 Series Specification | |
|---|---|
| Specification | Description |
| Support LLM Model | Llama3.2, Mistral NeMo, Phi 3.5, Breeze, TAIDE |
| Total Board Power | Min. = 25W Default = 35W Max. = 40W |
| Thermal Solution | Upon Request |
| Mechanical Form Factor | OCP Dual M.2 |
| Memory Type | LPDDR5 |
| Memory Size | Up to 64GB |
| Memory Clock | 6400 Mbps |
| Ambient Operating Temperature | 0°C to 50°C |
| Storage Temperature | -40 °C to 75°C |
| Operating Humidity | 5% to 85% Relative Humidity |
| Storage Humidity | 5% to 95% Relative Humidity |
Key Benefit
Maximize Savings, Minimize AI Costs
Transform AI Operation Instantly Efficiently
Efficiently Compact, No Hardware Hassle
Streamline Operations, Maximize Output
Offline Security, Zero Data Leaks
Low Power, High Sustainability
Key Feature
Space Optimization
The DM2's compact form factor provides significant advantages for organizations with limited physical space or those looking to maximize density in their computing environments. This benefit is particularly valuable for small offices, edge computing applications, or environments where space efficiency is a primary concern, while still maintaining the same level of AI processing capabilities as larger solutions.Cost Efficiency
Like its larger counterpart, the DM2 delivers substantial financial benefits by reducing the total cost of AI implementation. Its compact design potentially offers additional cost savings in environments where space is at a premium. Organizations can achieve advanced AI capabilities without extensive infrastructure investments, making it an economical choice for businesses of all sizes.Flexible Integration
The DM2's smaller form factor enhances its integration flexibility, making it suitable for a wider range of hardware configurations. This allows organizations to implement AI capabilities in environments where larger cards might not be practical, while maintaining the plug-and-play simplicity that makes deployment straightforward and minimally disruptive to existing operations.Optimized Performance
Despite its compact size, the DM2 maintains the same impressive performance benefits as larger solutions. It effectively offloads AI workload from the CPU and accelerates RAG processes, enabling organizations to achieve high-performance AI operations even in space-constrained environments. This optimization ensures that organizations don't have to compromise on performance when choosing a more compact solution.