Dell PowerEdge R660

Dell PowerEdge R660 is Dell's dense 1U rack server and the high-density complement to the R760. It delivers the same Intel Xeon Scalable 4th and 5th Gen processors, the same 32 DDR5 DIMM slots and the same memory ceiling in half the rack height — a foundational platform for hyper-converged infrastructure (HCI), dense virtualization farms, Kubernetes worker nodes and the rack-constrained data centers typical of European colocation facilities, UK financial-services data centers and US edge sites. For our enterprise customers across the EU, UK and US, R660 is the rack-density-optimized counterpart to R760 — chosen when every rack U has a billable cost or when the operating model is built around scaling by adding nodes rather than scaling individual hosts.

Form-factor trade-offs. The 1U chassis necessarily limits storage and accelerator capacity relative to the R760. Storage tops out at 10 EDSFF E3.S NVMe SSDs (versus 24 SFF or 12 EDSFF on R760) or 8 SFF SAS/SATA/NVMe drives in the alternative backplane configuration. GPU support is restricted to single-slot or low-profile accelerators such as the NVIDIA L4 inference card, the NVIDIA A2 entry-level inference accelerator and selected FPGA cards — full double-slot training and inference GPUs require R760 or, for dense training, the dedicated XE9680 platform. In return for these limits, operators gain twice the compute density per rack U: a 42U rack populated end-to-end accommodates 40 R660 servers versus 20 R760 servers, with significant implications for total VM density, container worker-node count and overall data-center utilization economics. For hyper-converged infrastructure deployments specifically, where clusters scale by adding nodes, this density advantage compounds quickly — a 24-node HCI cluster occupies 24U of R660s or 48U of R760s, and the rack-and-power savings are material.

Compute and memory subsystem. Identical silicon options to R760. Two Intel Xeon Scalable 4th or 5th Gen processors with up to 64 cores per socket on Emerald Rapids SKUs; 32 DDR5 DIMM slots supporting RDIMMs up to 256 GB; eight DDR5 memory channels per socket sustaining over 600 GB/s of aggregate memory bandwidth; PCIe Gen5 expansion across all primary slots; CXL memory expansion support. For HCI clusters, the high core-and-memory ceiling per node means fewer nodes are required for a given total cluster capacity — reducing the cluster network footprint, the per-cluster software licensing surface and the operational complexity of managing many small hosts.

Hyper-converged infrastructure (HCI) fit. R660 is one of the most heavily deployed HCI platforms in the world. Dell VxRail — the joint Dell-VMware HCI appliance — is built on the PowerEdge R660 chassis and is the reference platform for vSAN HCI deployments at most large enterprises. Nutanix Cloud Platform certifies R660 as a reference node alongside Dell's own XC Core line; Microsoft Azure Stack HCI / Azure Local certifies R660 as a validated node; Red Hat OpenShift Data Foundation supports R660 as a primary platform. A typical HCI cluster starts at three nodes (the minimum for two-copy redundancy on vSAN and Nutanix), scales in single-node increments to dozens of nodes, and tracks Kubernetes-style cattle-not-pets operating models. Production HCI clusters in our EU and US enterprise customer base typically range from 6 to 24 nodes for departmental clusters and from 24 to 60+ nodes for organization-wide platforms. The 1U form factor is particularly important here because HCI scales by adding nodes — 1U keeps the rack footprint manageable as the cluster grows. For VMware vSAN Express Storage Architecture (ESA), the EDSFF NVMe storage tier delivers excellent IOPS density per node and aligns with the ESA reference design.

Container and Kubernetes worker nodes. For Red Hat OpenShift, Amazon EKS Anywhere, Google Distributed Cloud, Rancher Kubernetes Engine 2 (RKE2), SUSE Rancher and CNCF-conformant Kubernetes distributions, R660 is the canonical worker-node platform. The combination of dense compute, OCP 3.0 25/100 GbE networking and EDSFF NVMe storage delivers excellent pod density and pod-network performance. For service-mesh-heavy architectures (Istio, Linkerd, Consul Connect) and observability-heavy environments (Prometheus, OpenTelemetry, Loki), the high core count also accommodates the per-pod CPU overhead of sidecar containers without compromising application capacity. For stateful workloads, OpenShift Data Foundation, Portworx and similar Kubernetes-native storage layers benefit from the EDSFF NVMe storage tier.

iDRAC 9 and OpenManage Enterprise management. Identical management stack to R760. iDRAC 9 provides KVM-over-IP, virtual media, full sensor telemetry, BIOS and firmware management, and the hardware security stack — Silicon-based Root of Trust for immutable hardware attestation, System Lockdown Mode for protection against unauthorized BIOS or firmware modification, Secured Component Verification for supply-chain integrity, and TPM 2.0 for hardware-backed key storage and measured-boot attestation. OpenManage Enterprise orchestrates fleets via template-based provisioning, firmware compliance reporting, Group Manager federation and full integration with VMware vCenter, Microsoft System Center, Red Hat Satellite, Red Hat Ansible, ServiceNow and Terraform. FIPS 140-2 and FIPS 140-3 validated cryptographic modes are available for regulated workloads. For HCI clusters specifically, OpenManage Enterprise's ability to manage firmware compliance across dozens of nodes from a single template is invaluable — a misaligned firmware revision across cluster nodes is a frequent root cause of vSAN, Nutanix and Azure Stack HCI issues.

Networking. The OCP 3.0 modular network adapter slot supports 1, 10, 25, 50 and 100 GbE NICs and InfiniBand HDR/NDR adapters. The slot is field-replaceable without taking the chassis apart, simplifying network upgrades over the multi-year operational lifecycle of an HCI cluster — customers can deploy on 25 GbE today and upgrade to 100 GbE in three years without replacing the server. PCIe Gen5 expansion delivers full bandwidth to additional network adapters in the standard PCIe risers. For HCI clusters running East-West storage-replication traffic (vSAN, Nutanix, Azure Stack HCI), the network bandwidth is frequently the limiting factor; the OCP 3.0 100 GbE option with RDMA over Converged Ethernet (RoCE) v2 is the recommended baseline for new builds.

Power and thermal envelope. Typical R660 platforms run between 400 W and 600 W under load, depending on CPU SKU selection, DIMM population, accelerator presence and storage configuration. Two redundant hot-plug power supplies are available in 800 W, 1100 W, 1400 W, 1800 W and 2400 W Titanium-class ratings — Titanium PSUs deliver 96 percent efficiency at 50 percent load, the highest tier in the 80 PLUS certification. For dense rack deployments, careful planning of total power per rack and the available power-distribution-unit (PDU) capacity is essential. A fully populated 42U rack of R660s under heavy load can exceed 20 kW total draw — requiring high-density three-phase PDUs (typically 32 A or 60 A circuits in EU and UK deployments, or 208 V three-phase in US deployments) and dedicated cold-aisle containment with raised-floor or in-row cooling. For colocation customers, confirming rack power capacity and cooling envelope with the data-center operator is a critical pre-deployment step that Sora Yazilim coordinates as part of every R660 cluster engagement.

Sora Yazilim engagement model. For R660 we frequently engage on hyper-converged or Kubernetes cluster sizing. We model node count against the customer's VM density, container density, storage IOPS, storage capacity, protection-domain failure tolerance and growth-rate assumptions. We run the Dell Live Optics sizing assessment, the VxRail Sizer, the VMware vSAN sizing tool, the Azure Stack HCI capacity calculator or the OpenShift Data Foundation sizer alongside Dell's official PowerEdge configurator, then deliver a co-engineered cluster bill of materials. OEM-grade rack-and-stack installation includes top-of-rack switch configuration, cluster network plumbing, iDRAC and OpenManage baseline configuration, BIOS hardening per CIS Benchmark and NIST 800-53 SI-7 guidance, System Lockdown firmware-compliance lock, cluster bring-up (VxRail Manager day-zero, Nutanix Foundation, vSAN Cluster Quickstart, Azure Stack HCI deployment, OpenShift installer-provisioned infrastructure) and a complete handover dossier. ProSupport Plus 24/7 with 4-hour SLA is co-quoted on every proposal, and we track warranty renewals 90 days before expiry. For multi-rack cluster deployments we coordinate staged delivery, regional logistics (EU regional logistics in Limerick and Łódź, UK regional logistics in regional hubs, US regional logistics across Texas and Tennessee) and on-site installation timelines against the customer's project schedule.

Compliance and regulatory posture. Identical to R760. FIPS 140-2 and FIPS 140-3 validated cryptographic modes via iDRAC 9 and TPM 2.0; secure boot with measured-boot attestation for hardware-anchored chain-of-trust; complete audit-log retention for NIST 800-53 AU-2 and AU-12 controls; Silicon-based Root of Trust and Secured Component Verification for supply-chain integrity attestation aligned with NIST 800-161. For US federal customers we deliver against DISA STIG, FedRAMP-aligned baselines, NIST 800-53 moderate and high baselines, and CMMC level 2 and 3 requirements. For EU customers the platform supports NIS2 directive technical measures, GDPR Article 32 requirements, the EU AI Act infrastructure attestation requirements, and the ENISA cloud security baseline. For UK customers, Cyber Essentials Plus alignment is delivered. Industry-specific compliance includes PCI DSS, HIPAA HITECH, SOX IT general controls, FDA 21 CFR Part 11 and the ISO 27001 / ISO 27017 / ISO 27018 control families. Sora Yazilim delivers the compliance baseline documentation, configuration evidence packs and audit-readiness reports required by both internal audit teams and external assessors.

Sustainability and lifecycle. R660 delivers significant power-efficiency improvements over the previous-generation R650 — approximately 20-30 percent higher performance per watt on typical virtualization and HCI workloads, Titanium-class power supplies at 96 percent efficiency at 50 percent load, and a redesigned airflow path supporting inlet temperatures up to 35 degrees Celsius for higher data-center setpoints and reduced cooling overhead. For organizations subject to EU CSRD, California SB 253 climate disclosure or UK SECR reporting, the efficiency improvements translate directly to Scope 2 emissions reductions. Dell publishes per-SKU product carbon footprint disclosures under the EPEAT registry to support customer Scope 3 reporting. For end-of-life, Dell Asset Recovery Services provides certified data sanitization (NIST 800-88 compliant), responsible reuse and recycling, and the documentation trail required for IT asset disposition (ITAD) audits. Sora Yazilim coordinates the end-of-life workflow alongside warranty renewals to maintain a complete asset lifecycle record.

Software and hypervisor support. R660 is certified across the major hypervisor and operating-system platforms: VMware ESXi 8.0 Update 3 and later, Microsoft Windows Server 2022 and 2025 with Hyper-V, Microsoft Azure Stack HCI / Azure Local, Proxmox VE 8, Nutanix AOS 6.7 and later, Red Hat Enterprise Linux 9.4 and later, SUSE Linux Enterprise Server 15 SP6 and later, Ubuntu LTS 24.04 and Oracle Linux 9. Container-platform support includes Red Hat OpenShift 4.16+, Rancher Kubernetes Engine 2, SUSE Rancher and CNCF-conformant upstream Kubernetes via Cluster API providers. For composable-infrastructure workflows, OpenManage templates and the Dell PowerEdge Ansible collection provide infrastructure-as-code workflows compatible with Ansible, Terraform and Pulumi. The full hypervisor compatibility matrix is published in Dell's PowerEdge Operating System Support Matrix and is reviewed quarterly by Sora Yazilim as part of our standard customer-success cadence.