2026-11-04 · Fonte：Nvidia Blog
The AI inference chipmaker will connect its next-generation Raptor XPUs to NVIDIA's NVLink scale-up, Spectrum-X scale-out networking and MGX rack architecture, joining a growing roster of NVLink Fusion ecosystem partners.
d-Matrix, a chipmaker focused on AI inference silicon, said it will adopt NVIDIA's NVLink Fusion technology to link its next-generation Raptor XPUs into NVIDIA's AI infrastructure platform. The announcement places the inference specialist alongside a growing roster of ecosystem partners that have signed on to the interconnect program, extending NVIDIA's rack-scale connectivity to accelerator designs developed outside its own silicon roadmap.
Under the arrangement, Raptor will connect to NVIDIA NVLink scale-up networking and Spectrum-X scale-out networking, as well as the NVIDIA MGX rack architecture and the company's broader AI platform. The integration is aimed at rack-scale deployment, allowing d-Matrix's inference accelerators to sit within the same chassis-level and cluster-level topology that NVIDIA's own systems occupy, with NVLink Fusion serving as the bridge between the third-party XPU and NVIDIA's connectivity stack.
The move carries weight for the merchant accelerator market. NVLink Fusion extends NVIDIA's scale-up connectivity to third-party accelerator designs, and d-Matrix's participation underscores how inference-focused silicon vendors are plugging into established rack ecosystems rather than standing alone. For d-Matrix, alignment with NVLink scale-up and Spectrum-X scale-out fabric means Raptor-based systems can be deployed within NVIDIA's reference rack architecture, the MGX standard, reducing the need for operators to engineer bespoke interconnect paths.
The growing roster of NVLink Fusion partners cited in the announcement signals NVIDIA's strategy of positioning its networking and rack platform as common infrastructure for a range of accelerators, including competing inference silicon. For data center operators, the d-Matrix integration adds another accelerator option that slots into MGX-based racks and Spectrum-X networks, while for NVIDIA it reinforces the centrality of its platform even as alternative XPUs reach the inference market.
Tags: d-Matrix · NVIDIA · Raptor XPU · NVLink Fusion · AI Inference · Data Center
萬安算交所 AIXX · AIXX · Exchange verificado · Ir ao mercado