Corsair

Description

Corsair is a first-of-its-kind AI compute platform designed for generative AI inference in datacenters. It delivers unparalleled performance and efficiency for large language models, enabling high interactivity and offline batched use cases at scale.

Key Differentiators

• 20x better power efficiency compared to traditional solutions
• 20x lower latency for generative AI inference
• 10x lower total cost of ownership (TCO)
• Seamless integration with open frameworks like PyTorch and Triton DSL

Features

• Unique Digital-In Memory Compute Architecture (DIMC™) for AI inference acceleration
• Integrated performance memory for blazing fast interactivity
• Off-chip capacity memory for offline batched inference
• Block floating point numerical formats for inference efficiency
• PCIe form factor for easy plug-and-play with existing datacenter infrastructure
• Scalable from card to rack with standard fabrics

Use Cases

• Deploy AI inference workloads efficiently across datacenters
• Enable low-latency interactive applications for generative AI
• Optimize inference for large language models up to 100B parameters
• Support offline batched inference for enterprise workloads

Pricing

Deployment & Platforms

Integrations

Target Market