d-Matrix

d-Matrix utilizes a unique architecture called 3D stacked digital in-memory compute (3DIMC™) that integrates memory and compute for ultra-low latency AI inference. This design allows for efficient processing of large-scale generative AI workloads, significantly reducing latency and increasing throughput.

d-Matrix is powered by its Corsair AI platform, which employs advanced digital in-memory technology and a chiplet-based design. This architecture supports models with up to 100 billion parameters and delivers high throughput and low latency for AI inference.

Key features of d-Matrix include ultra-low latency AI inference, high batched throughput, and energy efficiency. The platform also supports disaggregated GPU pipelines, enabling up to 10x performance improvements over standard GPU-only setups.

d-Matrix focuses on generative AI, leveraging its innovative memory-compute integration to enhance inference capabilities. This allows for the processing of complex AI models efficiently, catering to various enterprise applications.