Dell PowerEdge R960 Specs and Features
Four Sockets. Four Rack Units. Maximum Business-Critical Density
Form Factor — 4U rack server, 16th Generation PowerEdge (Gen 16); air-cooled design engineered for maximum compute and storage density in deployments where raw capacity takes priority over rack space
Four-Socket Architecture — Supports up to four 4th Gen Intel® Xeon® Scalable processors (Sapphire Rapids); unique stacked PEM board positions CPUs 3 and 4 directly above CPUs 1 and 2, consolidating four sockets into a single 4U chassis without the cost of a full-width 8U platform
Maximum Core Count — Up to 60 cores per socket; 240 total cores across all four sockets for the most demanding parallel computing, in-memory analytics, and AI/ML inference workloads
Memory Capacity — 64 DDR5 RDIMM slots delivering up to 16 TB of registered ECC memory at speeds up to 4800 MT/s — the highest in-chassis memory capacity in the Gen 16 family
PCIe Gen 5 Expansion — Up to 12 PCIe Gen 5 slots (10 Full Height + 2 Low Profile) — 50% more slots than comparable 2U four-socket platforms, enabling GPU, InfiniBand, and high-throughput NVMe coexistence
GPU Acceleration — Supports up to 4 × 400 W double-width GPU cards for AI inference, HPC simulation, and visualization workloads; dedicated GPU air shroud required
Massive Storage Options — Drive configurations from 8 to 32 × 2.5-inch SFF bays, up to 24 × 2.5-inch NVMe, or up to 16 × EDSFF E3.S Gen5 NVMe; supports SAS4, SATA, and NVMe simultaneously with dual RAID controller options
Dimensions & Weight — 482 mm W × 174.3 mm H × 883.2 mm D (with bezel); fully loaded weight: 60.2 kg (132.71 lbs); up to 4 PSUs with 2+2 redundancy for 24/7 continuous operation
Engineered for the Most Demanding Mission-Critical Applications
In-Memory Databases — 16 TB of DDR5 across 64 DIMM slots powers the largest SAP HANA, Oracle In-Memory, and Microsoft SQL Server in-memory deployments without scale-out complexity; keep entire working datasets in RAM for sub-millisecond query response
AI & Machine Learning Inference — Four 400 W DW GPUs combined with 240 Sapphire Rapids cores and Intel AMX (Advanced Matrix Extensions) on-die acceleration deliver AI inference at scale for healthcare, financial services, and manufacturing analytics
Business-Critical Virtualization — 240 cores across 4 sockets sustain massive VM densities; high vCPU-to-core ratios, hardware-assisted virtualization, and NUMA-aware scheduling for VMware vSphere and Red Hat KVM enterprise deployments
ERP & Line-of-Business Applications — Four-socket compute headroom handles peak transaction loads for SAP ERP, Oracle E-Business Suite, and Microsoft Dynamics without vertical scaling limits or planned downtime windows
HPC & Technical Computing — 12 PCIe Gen 5 slots plus 4 GPU slots and 240 compute cores accelerate simulation, modeling, CFD, and parallel scientific workloads at datacenter scale within a single 4U chassis
Data Analytics & OLAP — DDR5 memory bandwidth across four sockets feeds high-throughput analytics engines processing terabytes of resident in-memory data sets with no I/O bottlenecks
Server Consolidation — Replace racks of aging 2-socket servers with a single R960; consolidate compute, memory, GPU, and storage resources in 4U while preserving capital investment in existing rack infrastructure
4th Gen Intel® Xeon® Scalable — Up to 240 Cores Across Four Sockets
Socket — LGA4677; supports up to four 4th Gen Intel® Xeon® Scalable (Sapphire Rapids) processors; CPUs 1 and 2 install on the main system board; CPUs 3 and 4 install on the stacked PEM board — minimum one CPU required to boot
Maximum Core Count — Up to 60 cores per socket; 240 total cores when all four sockets are populated — the highest parallelism available in a 4U air-cooled rack server without multi-node blade infrastructure
Intel UPI Interconnect — Up to four Intel Ultra Path Interconnect (UPI) links at up to 16 GT/s per link providing high-bandwidth, low-latency coherent multi-socket fabric across all four CPUs
PCIe Gen 5 I/O — Up to 80 PCIe Gen 5 lanes per socket at 32 GT/s per lane; 2× the per-lane bandwidth of PCIe Gen 4 for saturating NVMe storage arrays, GPU interconnects, and 400GbE adapters
Built-In Accelerators — Intel QuickAssist Technology (QAT) for cryptography and data compression; Intel AMX for AI matrix math workloads; Intel IAA and DSA for memory analytics — all on-die, consuming no PCIe expansion slots
Heatsink Guide — Standard (STD) ≤ 185 W; L-type High Performance (HPR) 195–250 W; L-type Very High Performance (VHPR) ≥ 270 W; each of the four sockets requires its own correctly rated heatsink
Optional Intel QuickAssist — On-die QAT available on supported Sapphire Rapids SKUs; offloads cryptographic operations (AES-GCM, SHA-256/384) and LZS/Deflate compression from application cores — no add-in accelerator card required
CPU Mixing — Mixing different processor models across the four sockets is not supported; all installed CPUs must be identical models to ensure proper NUMA topology and UPI link negotiation
64 DDR5 DIMM Slots — Up to 16 TB at 4800 MT/s
DIMM Slots — 64 total DDR5 RDIMM slots (16 per processor socket); slots 1–32 reside on the main system board (CPUs 1 and 2); slots 33–64 reside on the PEM board (CPUs 3 and 4) — all four processors required to use all 64 slots
Maximum Capacity — 16 TB total when all 64 slots are populated with the highest-density DDR5 RDIMMs available; the largest single-chassis in-memory capacity in the Gen 16 PowerEdge line
Operating Speeds — 4800 MT/s at 1 DIMM per channel (1DPC); 4400 MT/s at 2 DIMMs per channel (2DPC); DDR5 delivers approximately 50% greater memory bandwidth per pin compared to DDR4 at equivalent channel counts
Memory Type — Registered ECC DDR5 RDIMMs only; DDR4 is NOT supported in any R960 configuration; verify DIMM compatibility against the Dell EMC PowerEdge R960 Memory Configuration Guide before ordering
Memory Mixing — Memory mixing across DIMM slots is not supported on the R960; all populated DIMM slots within a channel must be the same capacity and speed for correct operation
Voltage — 1.1 V DDR5 operation; improved per-gigabyte power efficiency versus DDR4's 1.2 V baseline — critical when fully populated with 64 DIMMs in a power-sensitive datacenter environment
In-Memory DB Scale — 16 TB resident memory capacity enables SAP HANA scale-up configurations that previously required proprietary appliances or multi-node scale-out clusters; consolidate to a single manageable platform
ECC Protection — All DIMMs provide Error Correcting Code protection; single-bit error correction and multi-bit detection are essential for mission-critical in-memory databases where a single undetected bit flip cannot be tolerated
Eight Drive Configurations — From 8 to 32 Hot-Swap SFF Bays
8 × 2.5" SAS/SATA (Smartflow) — Entry SFF configuration; supports SAS4 and SATA HDDs and SSDs; Smartflow backplane enables universal drive compatibility with hardware RAID via PERC controller
16 × 2.5" SAS/SATA (Smartflow) — Mid-range SFF layout; doubles capacity over the 8-bay config while preserving PCIe slot availability for GPU and network expansion
32 × 2.5" SAS/SATA (Dual Controller) — Maximum SFF capacity; dual PERC controller architecture provides independently managed RAID sets — two separate RAID domains with separate controllers for storage redundancy at the controller level
16 × 2.5" SAS/SATA + 8 × 2.5" NVMe — Hybrid tiered storage; combines SAS4/SATA capacity drives with 8 NVMe SSDs for a two-tier storage model — fast NVMe for hot data, affordable SAS for warm/cold tier
24 × 2.5" NVMe Gen4 Passive — All-flash NVMe configuration; 24 passive NVMe bays over PCIe Gen 4 for latency-sensitive transactional and analytical workloads requiring maximum storage IOPS
Universal Backplane HW RAID Rules — The Universal Backplane supports hardware RAID for SAS/SATA drives with direct-attach NVMe; hardware RAID for NVMe is not supported through the Universal Backplane — NVMe RAID requires an APERC controller
Drive Type Mixing in RAID — Mixing SAS, SATA, and NVMe drive types within a single RAID volume is not permitted; configure separate RAID controllers for each drive type when deploying mixed media environments
Hot-Swap on All Configurations — Every R960 drive bay supports hot-swap replacement without server downtime; swap failed drives while workloads run for maximum business continuity in continuous-operation deployments
Up to 16 × EDSFF E3.S Gen5 NVMe — Next-Generation Flash at Scale
EDSFF E3.S Form Factor — Enterprise and Datacenter Standard SSD Form Factor; engineered to overcome the thermal and areal density limitations of traditional 2.5-inch U.2 NVMe drives in high-density 4U deployments
PCIe Gen 5 Bandwidth — Each EDSFF E3.S drive connects over PCIe Gen 5 (32 GT/s per lane); delivering up to 2× the sequential bandwidth per drive versus PCIe Gen 4 NVMe — maximum flash throughput for I/O-intensive workloads
16-Drive Configuration — Up to 16 × EDSFF E3.S Gen5 NVMe drives; the highest EDSFF drive count available in the Gen 16 rack server lineup — purpose-built for extreme-density, lowest-latency flash environments
16 EDSFF + 8 SAS/SATA Combo — Mixed backplane configuration combines 16 × EDSFF E3.S Gen5 NVMe drives with 8 × 2.5-inch SAS/SATA drives in a single 4U chassis — flash-performance NVMe tier plus SAS capacity storage without a separate JBOD
Higher Density Per Bay — EDSFF longer form factor accommodates higher-capacity NAND configurations per drive compared to equivalent U.2 2.5-inch enclosures — more GB per bay with better sustained write endurance
Ideal for AI/ML Data Pipelines — PCIe Gen 5 NVMe delivers the storage throughput required to feed 240-core CPU compute plus 4 × GPU inference pipelines without storage I/O becoming the bottleneck
Thermal Advantage — E3.S form factor improves airflow routing around NAND packages compared to traditional 2.5-inch drives; enables higher sustained write performance in fully loaded 4U environments without thermal throttling
PERC 11 & PERC 12 RAID — Dual Controller and HBA Options
PERC 12 (H965i & H965e) — Latest-generation RAID controller; 24 Gb/s SAS4 + PCIe Gen 5 NVMe; supports RAID 0/1/5/6/10/50/60; 8 GB NV-backed write cache for high-throughput SAS4 and NVMe RAID configurations in the R960's largest drive bay counts
PERC 11 (H755 & H355) — 12 Gb/s SAS RAID controllers supporting RAID 0/1/5/6/10/50/60; H755 includes 8 GB NV cache for optimal write performance; H355 is the entry RAID controller for cost-sensitive configurations
Dual Controller Architecture — The 32-bay SAS/SATA configuration requires two independent PERC controllers, each managing a separate group of drives; if one controller fails, its drive group is unavailable but the surviving controller continues serving its drives — maximizing availability for the most critical storage pools
Separate RAID Controller Per Drive Type — In mixed-media configurations (SAS + NVMe), each drive type requires its own dedicated RAID controller; mixing drive types within a single RAID volume across one controller is not supported on the R960
HBA Pass-Through Mode — HBA (non-RAID) operation available for software-defined storage, Ceph, or OS-managed RAID configurations where the application layer controls RAID and the controller provides direct device access
APERC (Advanced PERC) — Advanced PERC installation consumes one PCIe riser slot; used for NVMe RAID capability or configurations beyond what the standard PERC slot supports; factor APERC into PCIe slot planning to preserve GPU and network expansion capacity
BOSS Boot Module — Boot Optimized Storage Subsystem (BOSS) provides dedicated NVMe M.2 RAID-1 boot mirroring; OS drives are isolated from data drives; BOSS does not consume a drive bay or PCIe riser slot, preserving all storage and expansion capacity for workload data
Up to 4 × 400 W Double-Width GPUs — AI, HPC, and Visualization
GPU Slot Count — The R960 supports up to 4 × double-width (DW) GPU cards consuming up to 400 W each; 4U chassis height provides the physical clearance and power headroom required for full-length, double-slot GPU accelerators
PCIe Gen 5 GPU Connectivity — GPU slots operate at PCIe Gen 5; 2× the bandwidth of PCIe Gen 4 for GPU-to-host data transfers — critical for AI inference pipelines that stream large batches from system memory to GPU VRAM at high throughput
GPU Air Shroud Required — Dell GPU air shroud must be installed when any GPU card is present; the shroud channels chassis airflow directly across the GPU thermal management zone, maintaining safe GPU junction temperatures at full 400 W TDP
AI Inference at Scale — Combine 4 GPUs with 240 Sapphire Rapids cores and Intel AMX on-die acceleration for hybrid CPU+GPU inference workloads; serve multiple AI models concurrently from a single 4U platform without separate GPU servers
HPC Simulation & Visualization — GPU-accelerated CFD, FEA, materials science, and molecular dynamics simulations benefit from PCIe Gen 5 bandwidth and DDR5 memory bandwidth sourcing large input datasets at high rates
On-Die CPU Acceleration (Intel AMX) — Beyond discrete GPUs, each Sapphire Rapids socket includes Intel Advanced Matrix Extensions (AMX) for INT8/BF16 matrix math; many inference workloads run efficiently on the 240 AMX-equipped cores without any GPU card installed
Power Budget Considerations — A fully populated R960 with four CPUs, 64 DIMMs, and 4 × 400 W GPUs draws substantial power; right-size PSU wattage at configuration time and verify rack PDU capacity before deploying a fully accelerated R960
Up to 12 PCIe Gen 5 Slots — 10 Full Height + 2 Low Profile
12 PCIe Gen 5 Slots — The R960 provides up to 12 PCIe Gen 5 expansion slots (10 Full Height + 2 Low Profile) on the system board — 50% more slots than comparable 2U four-socket platforms and enough capacity to simultaneously deploy GPUs, InfiniBand HCAs, NVMe adapters, and 100+ GbE NICs
PCIe Gen 5 Throughout — All expansion slots operate at PCIe Gen 5 (32 GT/s per lane); 2× the per-lane bandwidth of PCIe Gen 4 for GPU interconnects, NVMe storage expanders, and high-speed network adapters with no bandwidth compromise
Full Height vs Low Profile — 10 Full Height slots support standard dual-slot GPU cards and full-length network adapters; 2 Low Profile slots are ideal for dedicated management cards, compact network adapters, or secondary storage controllers
OCP Slot — Independent of PCIe Count — The R960 OCP 3.0 network slot is isolated from the 12 PCIe riser slots; an OCP network adapter installed here does NOT reduce available PCIe expansion count — preferred approach for adding 25/100 GbE networking without consuming a Full Height slot
APERC Impact Planning — Installing an APERC (Advanced PERC) controller for NVMe RAID consumes one Full Height PCIe slot; plan riser configuration at order time so APERC requirements don't compromise GPU or InfiniBand card placement
GPU + PCIe Coexistence — The 4U chassis provides enough physical and thermal headroom for simultaneous GPU and PCIe coexistence; 4 GPU slots + 8 remaining Full Height PCIe slots enable concurrent GPU, storage, and networking expansion in a single chassis
Slot Width Planning — Confirm physical slot width (x8 vs x16) against adapter interface width before ordering; insertion of a x16-interface card into a x8 physical slot will negotiate to x8 bandwidth, reducing performance for high-throughput adapters
Optional LOM + OCP 3.0 — Flexible Network Configuration
Optional 2 × 1 GbE LOM — An optional 2-port 1 GbE LAN-on-Motherboard card provides baseline network connectivity; the R960 supports either a LOM card, an OCP 3.0 adapter, or both simultaneously depending on configuration requirements
OCP 3.0 Slot — Dedicated OCP (Open Compute Project) 3.0 network slot supports optional OCP network adapter cards up to 100 GbE aggregate bandwidth; OCP slot is independent of the 12 PCIe expansion slots — install an OCP adapter without consuming any PCIe slot capacity
Both LOM and OCP Simultaneously — The R960 allows LOM card and OCP card to be installed at the same time for combined management (LOM) and production (OCP) network interfaces without consuming PCIe riser real estate
25 GbE / 100 GbE OCP Options — OCP 3.0 adapters with 4 × 25 GbE (100 GbE aggregate) support high-throughput storage replication, east-west VM traffic, and database cluster communication within a 4U server footprint
InfiniBand via PCIe — HDR/NDR InfiniBand host channel adapters install in Full Height PCIe Gen 5 slots for MPI-based HPC cluster fabric connectivity and lossless fabric requirements for tightly-coupled parallel workloads
100+ GbE via PCIe — High-performance 2-port 100 GbE or 4-port 25 GbE RDMA-capable adapters (RoCE, iWARP) install in Full Height PCIe Gen 5 slots for storage area network, AI training cluster, and high-frequency trading connectivity
iDRAC Dedicated Management Port — A dedicated 1 GbE iDRAC9 management port (independent from LOM and OCP) provides out-of-band server management on an isolated management network; production traffic is never mixed with iDRAC management traffic
Four PSUs with 2+2 Redundancy — AC and DC Options
Up to 4 PSUs — The R960 supports up to four power supply units, a unique capability among Gen 16 rack servers; two additional PSUs beyond standard 1+1 dual-PSU configurations provide a higher level of redundancy for mission-critical deployments
2+2 Redundancy — With four PSUs installed, the R960 operates in 2+2 redundant mode; the server continues running at full performance if up to two PSUs fail simultaneously — a level of power path resilience not available in servers limited to two PSUs
AC and DC Power Options — Supports both AC power supply units (standard datacenter) and DC power supply units for telecom, edge, and colocation facilities with -48 V DC power distribution — rare flexibility in a rack server
Hot-Swap Replacement — All PSUs are hot-swap capable; a failed unit is pulled and replaced while the server operates at full workload without interruption; the surviving PSUs absorb the load until a replacement is inserted
Platinum / Titanium Efficiency — Available in 80 PLUS Platinum and Titanium efficiency tiers; Titanium-rated PSUs minimize conversion losses at standard operating load ranges, reducing electricity costs across multi-year platform deployments
Power Budget for Full Configuration — A fully populated R960 with four 60-core CPUs, 64 DIMMs, 4 × 400 W GPUs, 32 SFF drives, and 12 PCIe cards draws substantial power; use Dell's Power Calculator to right-size PSU count and wattage at configuration time
iDRAC Power Monitoring — Real-time per-PSU power consumption monitoring via iDRAC9 telemetry; set power budgets and alerts to maintain control over energy consumption in high-density deployments
High-Volume Air Cooling — Up to 12 Hot-Plug Fans for Four Sockets
Air-Cooled Platform — The R960 relies on an optimized high-volume air-cooling architecture; Dell engineering tuned the 4U chassis airflow path to simultaneously cool four CPUs, 64 DIMMs, up to 4 GPUs, and up to 32 drives — no liquid cooling infrastructure required
Up to 6 Dual Fan Module Sets — The R960 accommodates up to 6 sets of dual fan modules (12 individual fans); high-volume fan array provides the airflow capacity required for the sustained full-TDP operation of four Sapphire Rapids processors and multiple 400 W GPUs
Hot-Plug Fan Replacement — Fan modules are hot-plug replaceable; swap a failed fan module without powering down the server; remaining fan modules automatically ramp speed to compensate during replacement
Heatsink Selection by CPU TDP — Standard (STD) heatsink for CPUs ≤ 185 W; L-type High Performance (HPR) for CPUs 195–250 W; L-type Very High Performance (VHPR) for CPUs ≥ 270 W; each of the four CPU sockets requires its own correctly rated heatsink for thermal compliance
PEM Board Thermal Path — The chassis fan subsystem is engineered to direct sufficient airflow across both the main system board (CPUs 1 & 2) and the elevated PEM board (CPUs 3 & 4) simultaneously; the 4U height provides the vertical clearance needed for the stacked thermal design
GPU Air Shroud — When GPU cards are installed, the mandatory GPU air shroud redirects a portion of the high-volume chassis airflow directly across GPU cooling zones; prevents GPU thermal throttling at sustained 400 W TDP under heavy AI/ML or HPC loads
ASHRAE A2 Compliance — Designed for ASHRAE A2 datacenter ambient conditions (up to 35° C) at standard CPU TDP; verify thermal compliance for maximum-TDP CPU configurations in ambient environments above 25° C
Drive Bay Blanks Required — All unused drive bays must have blanks installed to maintain the chassis pressure differential and prevent hot-air recirculation across drive backplanes; missing blanks can cause unexpected thermal events in fully loaded configurations
Rear Ports — USB, VGA, iDRAC Direct, and Data Connectivity
Rear USB Ports — USB 3.0 Type-A ports on the rear panel for external storage devices, bootable USB recovery media, and firmware update drives without requiring front-panel access in rack-dense environments
Front USB Port — USB 3.0 Type-A port on the front panel for local administrator access; enables drive identification, firmware staging, and diagnostic USB tools without reaching to the rear of the chassis
VGA Port — DB-15 VGA connector on the rear panel provides direct console video output for local KMM (keyboard-monitor-mouse) access during initial deployment, BIOS configuration, and emergency recovery
iDRAC Direct (Micro-USB) — Micro-USB port on the rear panel enables direct laptop-to-iDRAC connection via RNDIS for out-of-band management without a network switch; ideal for initial provisioning and on-site emergency access
Dedicated iDRAC9 Management Port — Separate 1 GbE RJ45 iDRAC9 dedicated management NIC; management traffic is completely isolated on a separate network segment from production data traffic; meets PCI-DSS and HIPAA network segmentation requirements for management channels
LOM and OCP Data Ports — Optional LOM card and OCP 3.0 adapter provide production network interfaces; LOM provides 2 × 1 GbE RJ45; OCP 3.0 adapters provide 10/25/100 GbE ports in SFP+, QSFP28, or RJ45 depending on adapter model selected
SD Card Slot — Internal MicroSD card slot supports hypervisor boot and OS media, providing a dedicated boot device that keeps primary drive bays free for workload storage
PCIe Rear Panel Access — Up to 12 PCIe Gen 5 riser slots exit through the rear panel; Full Height and Low Profile bracket options accommodate standard-length and compact expansion cards depending on riser configuration
Cyber Resilient Architecture — Root of Trust at the Silicon Level
Silicon Root of Trust — Cryptographic identity baked into the R960's silicon at manufacturing; the server can only boot firmware that has been cryptographically signed and verified — hardware-level supply chain attack mitigation that cannot be disabled by software or a compromised OS
Secured Component Verification (SCV) — SCV certificates digitally verify that all installed server components are authentic Dell-sourced parts; SCV checks run at power-on to detect unauthorized component substitution during shipping, staging, or post-deployment service
System Lockdown Mode — iDRAC9 System Lockdown prevents firmware updates, BIOS changes, and configuration modifications without explicit administrator authorization; deploy the R960 in Lockdown Mode for PCI-DSS, HIPAA, FedRAMP, and SOC 2 compliance environments
BIOS Signed Firmware — All BIOS updates are cryptographically signed by Dell; downgrade attack prevention ensures the server cannot be forced to run an older vulnerable BIOS version even if an attacker has administrative access to iDRAC
TPM 2.0 — Trusted Platform Module 2.0 provides hardware-based cryptographic key storage and measured boot capability; required for BitLocker full-disk encryption, vTPM virtualization, and zero-trust platform attestation workflows
UEFI Secure Boot — Verifies OS bootloader cryptographic signatures before transferring execution control; prevents unauthorized or malware-infected operating systems from loading during the boot sequence
Self-Encrypting Drive (SED) Support — SED drives provide instant cryptographic erase at end-of-life or drive replacement; drive data is permanently unrecoverable without the encryption key — eliminates physical drive destruction or expensive sanitization services
iDRAC Enterprise Security Features — Certificate-based authentication, two-factor authentication, RSA SecurID integration, LDAP/Active Directory authentication, and role-based access control; granular privilege assignments ensure the principle of least privilege for all iDRAC management users
iDRAC9 Management — Full Lifecycle Control from Any Location
iDRAC9 (Integrated Dell Remote Access Controller 9) — Out-of-band management controller operates completely independently of the server OS; manage, monitor, update, and remediate the R960 from any location even when the OS is hung, crashing, or powered off
iDRAC9 License Tiers — Express (base), Enterprise, and Datacenter licenses; Datacenter license unlocks OpenManage Integration with VMware vCenter, advanced multi-server dashboards, telemetry streaming, and automated proactive support via SupportAssist
Redfish API — Full DMTF Redfish 1.x RESTful API support; enables infrastructure-as-code management via Ansible, Terraform, and PowerShell DSC — configure and manage the R960 programmatically without installing any in-OS agents
OpenManage Enterprise — Dell's centralized multi-server management console; monitor firmware compliance, orchestrate multi-component firmware update campaigns, manage configuration drift, and respond to alerts across entire R960 deployments from a single interface
Virtual Console & Virtual Media — Full remote KVM console access with virtual CD/DVD and USB media mounting; re-image operating systems, recover from failed updates, and reconfigure BIOS settings without requiring physical datacenter access
iDRAC Direct (Laptop Interface) — Micro-USB port enables direct RNDIS-based laptop-to-iDRAC management without a network switch; critical for initial provisioning in disconnected environments or emergency on-site access when management networks are unavailable
Automated Firmware Orchestration — iDRAC9 integrates with Dell Repository Manager to execute multi-component firmware update campaigns across BIOS, PERC, NICs, PSUs, and fan controllers in a single coordinated operation with rollback capability
Telemetry Streaming — Real-time CPU, memory, PCIe, GPU, and power telemetry streamed via Redfish Server-Sent Events (SSE) or Kafka to Splunk, Grafana, custom SIEM, and AIOps platforms for predictive analytics and capacity planning
Power Enable Module — Four Sockets in a Single 4U Chassis
What Is the PEM Board? — The Power Enable Module (PEM) is a secondary planar board installed directly above the first two processors on the main system board; it physically holds CPUs 3 and 4, plus their 32 DDR5 RDIMM slots — enabling true 4-socket compute density in a single 4U enclosure
Stacked CPU Design — By mounting the PEM board over CPUs 1 and 2, Dell eliminates the need for an extended-width chassis or a second 4U server; the vertical stacking approach is what makes the R960 uniquely dense compared to traditional 4-socket platforms that required dual-chassis or large custom form factors
Required for 3rd and 4th CPU — The PEM board must be physically installed before CPUs 3 and 4 can be seated; a 2-socket R960 without a PEM board operates normally with 32 DDR5 DIMM slots and 8 TB maximum memory — still a significant in-memory computing platform
32 Additional DDR5 DIMM Slots — The PEM board contributes 32 DDR5 RDIMM slots (16 per CPU 3/4 socket), bringing the total chassis DIMM count to 64 when all four CPUs and the PEM board are installed, enabling the full 16 TB memory configuration
Thermal Integration — The 4U chassis fan subsystem is engineered to deliver adequate airflow across both the main system board and the elevated PEM board simultaneously; 4U chassis height provides the vertical airflow clearance not available in 2U designs
Installation Sequence — Install CPUs 1 and 2 into the main system board first; seat the PEM board above them; install CPUs 3 and 4 into the PEM board sockets last; populate DIMMs on both boards before first boot for maximum memory availability
Business Advantage — Organizations needing 4-socket compute density without procuring a specialized proprietary superserver gain significant TCO advantages; the R960 uses standard PowerEdge tooling, firmware, and management — operate it exactly like any other PowerEdge in the fleet
ReadyRails II Sliding Rails — Tool-Less 4-Post Rack Mounting
ReadyRails II — Dell's ReadyRails II sliding rail kit supports tool-less installation in 4-post EIA-310-compliant racks with square, round, or threaded hole patterns; no rack-specific adapter hardware required for most standard rack enclosures including Dell, HPE, Emerson, and APC racks
Full-Extension Sliding Rails — Rails extend fully to allow complete server removal from the rack without disconnecting cables; simplifies drive installation, PCIe card swaps, GPU replacements, PSU changes, and memory upgrades in production deployed racks
CMA (Cable Management Arm) — Optional Cable Management Arm attaches to the rail kit and routes power and data cables neatly alongside the server chassis; CMA flexes with the sliding rails so the server can be extended without disconnecting cables or disturbing adjacent 1U servers
4U Standard Depth — 883.2 mm depth with bezel (34.77 inches); 869.2 mm without bezel (34.22 inches); compatible with all standard 1000 mm and 1100 mm deep datacenter rack enclosures without depth extension kits
Weight Capacity Planning — Fully loaded weight of 60.2 kg (132.71 lbs); verify rail kit and rack enclosure static and dynamic weight ratings before deploying a fully populated 4-socket, 64-DIMM, 4-GPU, 32-drive R960 configuration
4U Density Advantage — The R960 delivers 4-socket compute density — historically requiring 8U or larger proprietary form factors — in a standard 4U chassis; every 4U slot in a 42U rack can now house a complete 240-core, 16 TB, GPU-capable compute node
Rack Thermal Planning — A fully loaded R960's power draw directly impacts rack power and cooling budgets; validate rack PDU circuit capacity, hot-aisle/cold-aisle containment airflow, and CRAC unit capacity before deploying multiple fully populated R960 nodes in the same rack
R960 Gen 16 vs R940 Gen 14 — The Case for Upgrading
PCIe Generation Leap — R960 delivers PCIe Gen 5 (32 GT/s per lane) vs PCIe Gen 3 (8 GT/s per lane) on the R940; 4× the per-lane bandwidth for NVMe storage, GPU interconnects, and 100 GbE networking without bandwidth compromise
Memory Technology — DDR5 at 4800 MT/s vs DDR4 at 3200 MT/s on the R940; approximately 50% more memory bandwidth per pin feeds Sapphire Rapids' larger core counts and GPU pipelines at speeds DDR4 could not achieve
Maximum Memory Capacity — R960 supports up to 16 TB (64 × DDR5 RDIMMs) vs R940's 6 TB maximum (48 × DDR4 LRDIMMs); nearly 3× more in-chassis memory capacity enables SAP HANA and Oracle In-Memory scale-up deployments previously requiring multiple nodes
Core Count Increase — Up to 240 cores (4 × 60c Sapphire Rapids) vs up to 112 cores (4 × 28c Cascade Lake) on the R940; more than double the core parallelism plus Intel AMX on-die AI acceleration not available on any prior Xeon generation
GPU Expansion — R960 supports up to 4 × 400 W DW GPU accelerators via PCIe Gen 5; the R940 had significantly more limited GPU support and PCIe Gen 3 bandwidth constraints; the R960 is purpose-built for GPU-accelerated workloads the R940 could not efficiently run
EDSFF Storage — R960 supports up to 16 × EDSFF E3.S Gen5 NVMe drives; the R940 had no EDSFF support — all NVMe was limited to 2.5-inch U.2 form factor with PCIe Gen 3 bandwidth
RAID Controller Generation — PERC 12 (24 Gb SAS4 + PCIe Gen 5 NVMe) vs PERC 10 (12 Gb SAS + PCIe Gen 3 NVMe) on the R940; 2× SAS bus bandwidth and Gen 5 NVMe for all-flash storage deployments
Form Factor Unchanged — Both platforms are 4U; upgrading from R940 to R960 requires no rack reconfiguration — swap in the same 4U slot with no new rail kits or rack space changes needed
| Feature | R960 (Gen 16) | R940 (Gen 14) |
|---|---|---|
| CPU Generation | 4th Gen Xeon Scalable (Sapphire Rapids) | 2nd Gen Xeon Scalable (Cascade Lake) |
| Max Cores | 240 (4 × 60 cores) | 112 (4 × 28 cores) |
| PCIe Generation | PCIe Gen 5 (32 GT/s) | PCIe Gen 3 (8 GT/s) |
| Memory Type | DDR5 at 4800 MT/s | DDR4 at 3200 MT/s |
| Max Memory | 16 TB (64 DIMM slots) | 6 TB (48 DIMM slots) |
| PCIe Slots | Up to 12 (10 FH + 2 LP) | Up to 8 |
| GPU Support | Up to 4 × 400 W DW GPUs | Limited GPU options |
| EDSFF Storage | Up to 16 × E3.S Gen5 NVMe | Not supported |
| PERC Controller | PERC 12 (24 Gb SAS4) | PERC 10 (12 Gb SAS) |
| On-Die Accelerators | QAT, AMX, IAA, DSA (Sapphire Rapids) | None (software-only) |
| Security | Silicon Root of Trust + SCV | Cyber Resilient Architecture (no SCV) |
| Management | iDRAC9 Rev. 7+ with Redfish 2.x | iDRAC9 Rev. 4.x |
| PSU Redundancy | Up to 4 PSUs (2+2) | Up to 4 PSUs (2+2) |
| Form Factor | 4U (883.2 mm depth) | 4U (841.5 mm depth) |
Frequently Asked Questions — Dell PowerEdge R960
The Dell PowerEdge R960 supports up to 16 TB of DDR5 ECC memory in a quad-socket 4U configuration — providing unmatched in-memory headroom for the world’s largest SAP HANA instances, Oracle Database, and real-time mission-critical analytics requiring terabyte-scale working sets.
The R960 supports up to 32 × 2.5-inch SFF hot-plug drives or 16 × EDSFF E3.S Gen 5 solid-state drives — delivering extreme IO throughput from NVMe and high capacity from SAS/SATA arrays to feed four Xeon processors simultaneously.
The R960 supports 4th Gen Intel Xeon Scalable processors (Sapphire Rapids, LGA4677) in a quad-socket configuration with up to 60 cores per CPU — up to 240 total cores and 480 threads for the most demanding Tier-1 enterprise workloads, consolidation projects, and scale-up database platforms.
Yes. Express Computer Systems stocks professionally reconditioned Dell PowerEdge R960 rack servers tested and configured for mission-critical enterprise workloads. Shop refurbished Dell R960 servers at ECS.
The R960 (Gen 16) replaces the R940 (Gen 14) — a two-generation jump. Highlights: 4th Gen Xeon Scalable (Sapphire Rapids) vs 2nd Gen (Cascade Lake-SP), DDR5 vs DDR4 with 16 TB capacity, PCIe Gen 5.0 vs Gen 3.0, Intel in-die AMX/DSA/QAT accelerators, 32 drive bays vs 24, EDSFF E3.S Gen 5 support, and iDRAC10 automation tooling.
Ready to Deploy the Dell PowerEdge R960?
Express Computer Systems specializes in professionally refurbished Dell PowerEdge servers — tested, reconditioned, and ready for your most demanding mission-critical workloads. Configure your R960 with the exact processor count, memory capacity, GPU accelerators, drive configuration, and PCIe riser layout for your environment.
Start building your custom server today