Best performance per Watt and per dollar
Kalray’s acceleration cards deliver exceptional performance and energy efficiency for AI/ML, edge computing, and data-intensive applications.
Ideal for SMEs, labs, and universities, they redefine computing standards with unmatched power and versatility.
Redefining Data Processing Efficiency
Accelerate Your Data-Intensive Applications with Kalray’s manycore processors and products
Traditional AI and signal processing tasks often rely on specialized GPUs like or specialized ASICs. While powerful, these solutions are costly and consume significant power, posing challenges in scalability and operational expenses. Moreover, ASICs lack the flexibility for updates, limiting adaptability to evolving workloads.
Kalray addresses these challenges with its Data Processing Unit (DPU) technology, including the MPPA® DPU Coolidge™ processor and K300-LP/TC4™ acceleration cards. These solutions offer comparable performance to high-end GPUs, with superior performance per watt and per dollar.
High-Performance Computing
Achieve up to 25 TFLOPs (16-bit) or 50 TOPs (8-bit) for demanding AI and signal processing tasks.
Energy Efficiency
Optimize operations with low power consumption, reducing costs and environmental impact.
Programmability
Utilize a fully programmable environment supporting standard languages like C/C++, Linux, and POSIX.
Scalability
Easily expand capabilities to meet growing data processing demands across various sectors.
Integrated AI Acceleration:
Benefit from dedicated AI processing capabilities, enhancing performance for ML applications.
Cost-Effectiveness
Achieve superior performance per dollar, making HPC accessible to a broader range of organizations.
Kalray’s innovative processors and acceleration cards deliver unparalleled performance and energy efficiency, empowering SMEs, laboratories, and universities to excel in AI, signal processing, and data-intensive applications.
A game changer for data-centric processing tasks and ML.
Kalray K300
Low Power and High performance for edge computing
The Kalray K300 significantly enhances server performance by offloading intensive tasks from the main CPU. It’s an ideal solution for AI and parallel processing algorithms, offering comparable computing power to mid-range GPUs but with significantly lower power consumption. While perfect for Smart Storage acceleration, the K300 is also versatile for various high-performance computing applications.
Key Specifications:
- Processor: Kalray MPPA®3-80 V1.2 @ 1GHz
- SSD Support: Up to 24 x 30TB NVMe SSD
- Interface: X16 PCIe Gen4, 2x QSFP28 100Gb Ethernet
- Performance: Up to 25 FP16 TFLOPS
- Power Consumption: 36W (typical), 42W (max)
Why Choose K300?
- Versatile Use Cases: Supports storage acceleration, AI, and parallel processing.
- High Efficiency: Reduces CPU load while optimizing performance.
- Flexible and Programmable: Compatible with any OS and server.
Kalray TC4
A Paradigm Shift in Compute Acceleration and AI/ML
The TurboCard4 (TC4) revolutionizes compute acceleration for AI-powered smart vision and data-indexing applications. Housing four Coolidge2™ DPUs, TC4 combines classical and AI-based processing technologies, offering up to 100 FP16 TFLOPS at 250W power consumption. Manufactured in France, TC4 delivers unmatched efficiency and performance for the most demanding AI workloads.
Key Specifications:
- Processor: 4x Kalray MPPA®3-80 V1.2 @ 1GHz
- Interface: X16 PCIe Gen4, 2x QSFP28 100Gb Ethernet
- Performance: Up to 100 FP16 TFLOPS
- Power Consumption: 60W (typical), 250W (max)
Why Choose TC4
- High Efficiency: Superior performance with lower power consumption.
- Made in France: Produced in partnership with Asteelflash, ensuring high-quality manufacturing.
- Scalability: Ideal for complex, parallel-processing tasks.
Kalray MPPA® DPU Manycore
The Coolidge processor, part of Kalray’s 3rd generation MPPA® DPU family, features a Massively Parallel Processor Array architecture. It addresses the explosion of data that traditional technologies struggle to handle efficiently.
Key Features:
- Massively Parallel Processor Array: Composed of computing clusters connected via AXIFabric bus grid and RDMANoC (Network-on-Chip) interconnects.
- Efficient Data Transfers: Optimized for diverse data transfer types, enhancing read/write access and network communication.
- Robust Partitioning: Ensures safe operation through configurable memory management units (MMUs) and memory protection units (MPUs).
Kalray’s DPUs excel in managing massive data flows and multiple workloads, providing superior performance per watt compared to traditional CPUs.
FAQs
Kalray’s advanced software solutions, tailored for high-performance data processing.
- Efficient Data Orchestration: Seamlessly manage and orchestrate data across storage and compute resources.
- AI and HPC Optimization: Enhance AI and HPC workloads with efficient data transfer and processing.
- SDKs and APIs: Easily develop custom applications with Kalray’s comprehensive SDKs and open APIs.
