d-Matrix Announces NVLink Fusion Integration Plans for Raptor XPU
Редакция ЦОД:Главное
d-Matrix has announced plans to adopt NVIDIA NVLink Fusion for its new Raptor XPU and to join NVIDIA’s AI platform partner ecosystem. The company intends to use NVIDIA MGX infrastructure components, including standardized rack designs, power delivery and liquid cooling, for AI-inference deployments.
- d-Matrix announced plans to use NVIDIA NVLink Fusion with its Raptor XPU.
- The proposed design uses NVIDIA MGX rack, power and liquid-cooling infrastructure.
- Planned integrations include Vera CPU, ConnectX-9 SuperNIC, BlueField-4 DPU and Spectrum-X Ethernet.
- The announcement suggests a shared infrastructure model for GPU, CPU and specialized XPU deployments.
- No capacity, deployment schedule, commercial volume or CAPEX has been disclosed.
d-Matrix plans to connect its new Raptor XPU to NVIDIA AI infrastructure through NVLink Fusion, positioning the inference processor within a common high-speed scale-up domain.
The company also plans to use the NVIDIA MGX ecosystem, including standardized rack configurations, power infrastructure and liquid-cooling systems. Its proposed platform integration includes NVIDIA Vera CPUs, ConnectX-9 SuperNICs, BlueField-4 DPUs and Spectrum-X Ethernet.
According to the announcement, d-Matrix also intends for its racks to operate alongside NVIDIA Vera Rubin NVL72 GPU systems. The proposed configuration would place specialized inference accelerators, CPUs and GPU systems within a more common physical and network infrastructure environment.
For data-center operators, the significance is architectural rather than a confirmed capacity deployment. Shared rack, power, networking and liquid-cooling standards could simplify the design and integration of heterogeneous AI infrastructure, particularly where inference accelerators are deployed alongside GPU-based systems. This could reduce the requirement for separate hardware architectures for each processor type if the integrations are adopted at scale.
The announcement does not disclose power capacity, rack counts, geography, deployment timing, commercial volumes or CAPEX. The integrations are described as plans rather than operational infrastructure.
The announcement points to a more modular AI-factory architecture in which specialized XPUs can operate alongside GPU and CPU systems within shared rack, networking and liquid-cooling environments. If such integrations scale, operators could reduce the need for processor-specific infrastructure designs, although no deployment volumes or operational installations have been confirmed.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.