Redefining Edge and Modular Compute: A Deep Dive into Cisco UCS X-Series Direct

In today's hybrid IT landscape, organizations are no longer confined to the walls of centralized data centers. With the growth of edge workloads, AI applications and hyper-converged infrastructure (HCI), the need for agile, high-performance compute platforms at the edge has become critical.

Enter Cisco's UCS X-Series Direct — a transformative extension of the UCS X-Series architecture that delivers a fully self-contained modular platform purpose-built for edge and remote-site deployments.

UCS X-Series Direct consolidates compute, networking, storage access and management into a single chassis that eliminates the need for top-of-rack switching. By embedding fabric interconnects directly into the chassis, it simplifies deployment, reduces cost and cabling complexity, and still maintains enterprise-grade performance and scalability. This solution is now available in our Advanced Technology Center (ATC) for customer test drives and proofs-of-concept.

A modular system designed for the edge

At the heart of this solution is the UCS X9508 chassis, a 7-rack-unit modular platform capable of hosting up to eight single- or dual-socket compute nodes or up to four 4-socket nodes. The chassis integrates dual UCS 9108 fabric interconnects, enabling unified networking across both LAN and SAN fabrics without the need for external switches. This consolidated design reduces power, cooling and cabling requirements while maintaining up to 1.6 Tbps of aggregate fabric bandwidth.

UCSX Chassis with the 9108 Fabric Interconnect.

The UCS X-Series Direct supports both traditional UCS Manager and Cisco Intersight for cloud-based, policy-driven management. Customers gain the flexibility to run modern workloads — VMs, containers and bare metal — under a single, consistent management experience. Organizations can schedule hands-on access through our ATC to validate these workloads in real-world scenarios.

What makes this system especially compelling is its midplane-free architecture. This design ensures forward compatibility with emerging technologies like PCIe Gen5 and CXL, as well as upcoming Intel and AMD processor families. It's a true investment in long-term modular compute.

Real-world flexibility with Unified Fabric

One of the strongest benefits of UCS X-Series Direct is its integration of the UCS 9108 fabric interconnects directly into the chassis. This not only eliminates the need for top-of-rack switching but also simplifies deployment in space-constrained or cost-sensitive environments like retail stores, manufacturing sites or remote offices. Each compute node receives dual 100 Gb connections, supporting a full-speed unified fabric for both IP and Fibre Channel over Ethernet (FCoE) workloads.

The 9108 Fabric Interconnect diagram — The 9108 Fabric Interconnect.

The fabric interconnects also support fabric extension and uplink redundancy, ensuring high availability and scalability as needs grow. With up to 200 Gb total bandwidth per server node and unified support for both compute and storage traffic, customers can confidently run enterprise workloads — even at the edge. These network configurations can be fully tested in our ATC to ensure optimal design before deployment.

GPU-powered AI at the edge

Cisco UCS X-Series Direct is GPU-ready, offering multiple configuration options that bring AI workloads closer to the data. Each compute node can house up to two GPUs, such as the NVIDIA T4 or Intel Flex Series (140/170), optimized for AI inferencing, video analytics and VDI.

For more intensive AI and rendering needs, customers can integrate the X440p PCIe node, which, when paired with the X-Fabric module, enables direct PCIe connectivity between compute nodes and GPUs housed in the PCIe expansion slots. The X-Fabric is a critical component that allows disaggregated compute-to-GPU architecture within the UCS X-Series chassis.

This configuration supports up to four full-height, double-wide GPUs per X440p node. Supported GPU options include:

NVIDIA L40S for advanced inference and visualization.
NVIDIA A100 for high-performance AI training and deep learning.
NVIDIA H100 NVL for transformer-based workloads and large language models.
NVIDIA A40 and A16 for virtualization and graphics-heavy VDI.

Customers can deploy multiple X440p nodes in the chassis to scale GPU density even further. In the highest configuration, a single UCS X9508 chassis can support up to eight (8) compute nodes and 4 X440p PCIe nodes, delivering up to 16 full-sized GPUs in a single 7RU chassis when paired with X-Fabric.

UCSX Chassis Showing Possible GPU Placement with X-Fabric (440p).

This makes UCS X-Series Direct an ideal platform for AI at the edge, where latency matters and running models locally is faster and more secure than sending data back to a centralized data center. Customers exploring edge AI use cases can leverage our ATC to benchmark and simulate these GPU workloads.

Nutanix HCI: Simplicity meets scalability

Cisco UCS X-Series Direct is the first modular blade platform to support Nutanix hyper-converged infrastructure. This collaboration combines UCS's modular efficiency with the software-defined power of Nutanix Cloud Platform. The result is a simplified, scalable infrastructure stack that is easy to deploy, manage and grow.

Through a jointly validated architecture, Nutanix clusters can be deployed directly onto UCS X-Series compute nodes, with Prism Central managing the software layer and Cisco Intersight managing the physical infrastructure. This dual-management model streamlines lifecycle operations, enables zero-touch provisioning and provides full-stack visibility.

Use cases for this solution range from remote branch infrastructure to manufacturing floor systems, retail analytics and container-based DevOps pipelines. In particular, Cisco and Nutanix have developed a reference architecture dubbed "GPT-in-a-Box," designed to bring AI inferencing capabilities to the edge using X-Series nodes with dual NVIDIA L40S GPUs. Prospective customers can validate this reference architecture in our ATC environment before scaling into production.

External storage integration: FlashStack, FlexPod and Cisco CVDs

For customers requiring external storage options or those with existing investments in enterprise storage platforms, UCS X-Series Direct integrates seamlessly with Cisco Validated Designs (CVDs) built around FlexPod and FlashStack architectures. These CVDs provide prescriptive guidance for deploying Cisco UCS with NetApp and Pure Storage solutions, ensuring best practices, high availability and predictable performance.

FlexPod, combining Cisco UCS, Cisco Nexus and NetApp ONTAP, remains a strong option for hybrid cloud deployments, tiered storage and mission-critical workloads. FlashStack, built with Cisco UCS and Pure Storage FlashArray, provides an all-flash, high-performance architecture ideal for databases, VDI and latency-sensitive applications.

UCS 9108 Fabric Interconnect can connect to storage either directly or though a traditional storage fabric (MDS / Brocade).

Both architectures can be extended to include UCS X-Series Direct as the compute layer. Customers can connect UCS X-Series Direct to external SAN or NAS systems via Fibre Channel or Ethernet using the integrated UCS 9108 Fabric Interconnects. This flexibility supports organizations that want to modernize compute at the edge or core while maintaining centralized storage operations.

Use case scenarios

Organizations looking to modernize their distributed environments are adopting UCS X-Series Direct for a variety of scenarios, such as:

Edge hyperconverged infrastructure: Fully integrated Nutanix clusters deployed at remote sites with no external switches or rack-level infrastructure.
AI Inferencing and video aAnalytics: Low-latency, GPU-enabled workloads that require proximity to data sources.
Virtual Desktop Infrastructure (VDI): Scalable, GPU-powered desktop deployments near end users.
Retail and branch IT: Plug-and-play compute with centralized management and built-in fabric redundancy.

Technical deep dive

Let's look at the specs that make this platform stand out:

Chassis: UCS X9508 (7RU), holds up to 8 compute nodes or 4 4-socket blades.
Fabric: 2 x UCS 9108 Fabric Interconnects (100 Gb, LAN and FC capable).
Compute nodes: X210c M6/M7/M8 and X215c M8; supports both AMD EPYC and Intel Xeon CPUs; up to 2 CPUs, 32 DIMMs, 2 GPUs per node.
PCIe expansion: X440p node with up to 4 full-height GPUs (L40S, A100, H100 NVL).
Management: Cisco Intersight, optional UCS Manager.

Each component is designed for forward compatibility and unified management, ensuring organizations can evolve workloads without forklift upgrades. All system configurations can be pre-validated in our ATC, allowing customers to tailor and test-fit solutions to their needs.

UCSX Direct supports both AMD and Intel proccessors, Nutanix and NVIDIA GPUs.

Conclusion

Cisco UCS X-Series Direct is more than an edge-ready chassis. It's a convergence of innovation, modularity and workload-centric design. From GPU-powered AI at the edge to zero-touch Nutanix HCI clusters in remote offices, it provides the building blocks for tomorrow's hybrid and AI-driven infrastructure.

With growing demand for local processing, simplified operations and intelligent automation, UCS X-Series Direct delivers the future of enterprise IT today. With full availability in our ATC, customers can begin exploring this future now, hands-on.

To experience this platform firsthand and see how it fits into your infrastructure strategy, contact us to schedule a test drive or customized proof-of-concept in our Advanced Technology Center (ATC).