A New 2U Server Form Factor
At this year’s SIGGRAPH conference in Vancouver, Nvidia introduced the RTX Pro 6000 Blackwell Server Edition, a new GPU designed specifically for on-premises data centers. Unlike previous RTX server GPUs, which came only in bulkier 4U–8U enclosures, the new model fits into a 2U rack-mount chassis. This slimmer form factor is aimed at mid-sized facilities seeking to enhance performance without consuming extra floor space.
Industry Partnerships
Nvidia has teamed up with major hardware vendors—Dell, Cisco, HPE, Lenovo, and Supermicro—to integrate the RTX Pro 6000 into their server lines. Dell, for instance, rolled out the PowerEdge R7725 2U, equipped with dual RTX Pro 6000 GPUs and powered by Nvidia’s AI Enterprise software stack. These collaborations ensure organizations can adopt the new hardware in flexible configurations tailored to AI and compute-heavy workloads.
Hardware and Performance
The servers come with Nvidia’s latest BlueField-3 DPUs and ConnectX-8 SuperNICs for high-speed networking and workload management. While optimized for AI model training and inference, the GPUs are also positioned to accelerate traditional CPU-dependent tasks such as data analytics, simulation, video encoding, and rendering.
Nvidia reports that compared to CPU-only 2U systems, the new RTX Pro servers can deliver up to 45 times more performance and 18 times greater energy efficiency—a significant leap that reinforces the broader case for GPU acceleration in enterprise computing.
Beyond Hardware: AI Model Advancements
Alongside the hardware announcement, Nvidia unveiled updates to its Nemotron family of large language models (LLMs). Two new models—Nemotron Nano 2 and Llama Nemotron Super 1.5—are designed to bring more advanced reasoning and multimodal understanding (text + vision) to enterprise AI applications.
- Nemotron Nano 2 focuses on speed, boasting up to 6x faster token throughput compared to other models of its size.
- Llama Nemotron Super 1.5 prioritizes accuracy, delivering top performance in reasoning-intensive tasks and making it well-suited for complex enterprise workflows.
These LLMs are engineered to strengthen agent-based AI systems, improving capabilities in areas such as coding assistance, scientific problem-solving, decision-making, and structured instruction following.
Why It Matters
With its latest server hardware and AI software, Nvidia is positioning itself as the go-to infrastructure provider for enterprises modernizing their data centers. The combination of efficient GPU-based servers and smarter LLMs highlights Nvidia’s vision of enabling businesses to run advanced AI workloads on-premises with reduced costs, smaller footprints, and significantly faster results.
