What is NVIDIA DSX?

NVIDIA DSX is an AI-factory-scale platform that unifies design, simulation, operations, and ecosystem technologies. It defines how AI factories are designed, built, and optimized across the full stack, from chips and systems to infrastructure software, facilities, and partner technologies. DSX combines open software libraries, workflow guides, and reference designs with NVIDIA compute platforms and codesigned infrastructure to enable a broad ecosystem of software and hardware solutions. By aligning every layer through a common architecture, DSX improves tokens per watt, accelerates deployment, and strengthens operational reliability and resiliency.

Why do you need it?

Enterprises need NVIDIA DSX to efficiently build and operate AI factories that are optimized for the demanding requirements of modern AI. As AI models grow in complexity and scale, and the imperative for rapid deployment intensifies, traditional, piecemeal approaches to infrastructure development lead to significant challenges:

Inefficient resource utilization: Difficulty in maximizing computational output (tokens) per unit of power (watts), leading to higher operational costs and environmental impact.

Prolonged deployment cycles: Lengthy and complex processes for designing, validating, and bringing AI infrastructure online, delaying AI innovation and time to market.

Operational fragility: Challenges in ensuring the consistent performance, reliability, and scalability of AI operations across diverse hardware and software components.

Lack of standardization: Absence of unified architectures and reference designs, complicating integration and maintenance across the AI infrastructure stack.

NVIDIA DSX addresses these by providing a holistic, codesigned solution that streamlines the entire AI factory lifecycle, ensuring optimal performance, efficiency, and reliability from the ground up.

Benefits of NVIDIA DSX

NVIDIA DSX delivers a suite of benefits crucial for the success of AI factory-scale operations:

​Accelerated time to revenue: By providing generation-specific, AI factory-validated architectures (like the Vera Rubin DSX reference design) and digital-twin capabilities (DSX Sim), DSX significantly reduces the time to first token. This accelerates the deployment and operationalization of AI factories, allowing businesses to realize value from their AI investments much faster.

Maximized efficiency (tokens per watt): DSX is fundamentally designed to optimize tokens per watt. This is achieved through components like DSX MaxLPS, an AI-factory-efficiency layer that maximizes computing output within a fixed power budget; and DSX Flex, a grid-aware power orchestration layer. This focus ensures maximum computational output for a given power budget, leading to lower operational costs and a reduced carbon footprint.

Enhanced reliability and resiliency: The platform strives to maximize throughput and operations by aligning every layer through a common architecture. DSX OS provides an open-source, modular software platform for Infrastructure as a Service (IaaS) and Platform as a Service (PaaS), ensuring robust and consistent performance. DSX Exchange further enhances reliability by acting as the AI factory IT/OT communications hub, facilitating seamless interaction for operational stability.

Unified full-stack optimization: DSX unifies design, simulation, operations, and ecosystem technologies across the entire AI stack – from chips and systems to infrastructure software and facilities. This comprehensive approach, including open software libraries, workflow guides, and co-designed infrastructure, simplifies deployment, reduces integration complexities, and ensures all components work together seamlessly for peak AI performance.

How does ASUS help?

ASUS plays a crucial role as a leading AI infrastructure partner in the NVIDIA ecosystem by:

​Implementing DSX reference designs: ASUS AI POD with NVIDIA Vera Rubin NVL72 aligns with NVIDIA DSX reference designs. This ensures that the infrastructure is purpose-built and validated to house NVIDIA's GPUs and the DSX software stack, providing optimal compatibility and performance.

Accelerating deployment with DSX SimReady: ASUS is adopting DSX SimReady across its rack-scale platforms. This involves converting their CAD geometry into validated OpenUSD assets embedded with crucial power, thermal, and connection-point metadata. These SimReady assets plug directly into the NVIDIA Omniverse DSX Blueprint and the broader DSX ecosystem.

Enabling digital-twin validation: By developing DSX SimReady assets, ASUS allows customers to design, verify, and optimize AI infrastructure using digital twins before physical construction begins. This pre-validation capability is crucial for delivering pre-validated AI factory hardware to gigawatt-scale customers, significantly compressing data center bring-up timelines.

Providing scalable hardware solutions: ASUS infrastructure solutions empowered by NVIDIA DSX contribute the robust, scalable, and reliable physical computing platforms that enable the efficient and high-performance operation of NVIDIA DSX. ASUS expertise in hardware manufacturing and integration is essential for bringing NVIDIA's vision of optimized AI factories to fruition.