• Lenovo ThinkSystem SR860 V3 – Scalable 4U Rack Server with Up to 4× Intel® Xeon® Scalable CPUs

    Keywords

    [Lenovo ThinkSystem SR860 V3](#Lenovo ThinkSystem SR860 V3), [4th Gen Intel Xeon Scalable](#4th Gen Intel Xeon Scalable), [4-socket server](#4-socket server), [DDR5 memory](#DDR5 memory), [NVMe storage](#NVMe storage), GPU-dense, [PCIe 5.0](#PCIe 5.0), [Lenovo Neptune cooling](#Lenovo Neptune cooling), [XClarity management](#XClarity management), [4U rack server](#4U rack server)

    Description

    The [Lenovo ThinkSystem SR860 V3](#Lenovo ThinkSystem SR860 V3) is a 4U, 4-socket rack server designed for mission-critical workloads—from SAP HANA in-memory computing to AI training and large-scale virtualization Lenovo Press. Powered by up to four 4th Generation Intel® Xeon® Scalable processors (up to 60 cores each, 3.7 GHz) in a mesh topology with three UPI links, it delivers massive parallel compute with up to 480 threads Lenovo Press.

    With 64 DDR5 DIMM slots supporting speeds up to 4800 MT/s and 16 TB total capacity, the SR860 V3 provides the highest memory bandwidth available for data-intensive analytics and in-memory databases Lenovo Press. Its flexible storage architecture offers up to 48× 2.5″ drive bays (24 AnyBay NVMe/SAS/SATA), plus rear 7 mm or M.2 boot drives, eliminating I/O bottlenecks and enabling sub-millisecond response times Lenovo Press.

    For GPU-accelerated workloads, the SR860 V3 supports up to four double-width 350 W GPUs (e.g., NVIDIA A100/H100) or eight single-width 75 W GPUs, with optional Lenovo Neptune™ liquid-to-air hybrid cooling maintaining optimal thermals under full load.

    Enterprise-grade management is delivered via Lenovo XClarity Controller 2 and the XClarity Administrator suite—automating deployment, firmware updates, and health monitoring to reduce provisioning time by up to 95 % Lenovo Press.

    Key Features

    • Up to 4× 4th Gen Intel® Xeon® Scalable CPUs: Mesh topology, 3 UPI links, 240 cores/480 threads max Lenovo Press

    • 16 TB DDR5 Memory: 64 DIMM slots, 4800 MT/s for highest-speed in-memory workloads Lenovo Press

    • Massive GPU Support: Four 350 W double-width or eight 75 W single-width GPUs for AI/HPC acceleration

    • 48× 2.5″ Drive Bays & NVMe AnyBay: Up to 24 direct-attach NVMe drives, eliminating storage bottlenecks

    • Lenovo Neptune™ Cooling: Hybrid liquid-to-air heat exchanger for GPU modules, reducing noise and power

    • PCIe 5.0 & OCP 3.0: Up to 16 Gen5/Gen4 PCIe slots plus two OCP 3.0 for 1–100 GbE networking

    • Lenovo XClarity Management: Embedded XClarity Controller 2, Administrator, Integrator plugins, Redfish API Lenovo Press

    • Enterprise-grade Reliability: Predictive Failure Analysis, light-path diagnostics, RoT/PFR security Lenovo Press

    Configuration

    ComponentSpecification
    Form Factor4U rack, 42.8 × 482.4 × 807 mm
    Processors2–4× 4th Gen Intel® Xeon® Scalable (up to 350 W, 60 cores/CPU, mesh UPI) Lenovo Press
    Memory64× DDR5 slots (64–256 GB TruDDR5 UDIMMs, up to 4800 MT/s, 16 TB max) Lenovo Press
    Drive BaysFront: 48× 2.5″ AnyBay (NVMe/SAS/SATA) or direct NVMe; Rear: 2× 7 mm or 2× M.2 boot SSDs Lenovo Press
    GPU OptionsUp to 4× double-width 350 W GPUs or 8× single-width 75 W GPUs; Lenovo Neptune cooling for HGX modules 
    I/O ExpansionUp to 16× PCIe slots (12× Gen5 + 4× Gen4) or 18× Gen4; 2× OCP 3.0 slots Lenovo Press
    NetworkingOCP 3.0 slots support 1/10/25/100 GbE; optional InfiniBand adapters
    Power Supplies2–4× Titanium/Platinum hot-swap PSUs (N+N redundancy)
    ManagementXClarity Controller 2, XClarity Administrator, Redfish API; LiCO orchestration optional

    Compatibility

    Supports all major OS and hypervisors:

    Usage Scenarios

    • In-Memory Databases (SAP HANA): Mesh-connected CPUs and 16 TB DDR5 accelerate real-time analytics with sub-millisecond latency Lenovo Press.

    • AI Training & Inference: Up to four 350 W GPUs with NVLink and Neptune cooling deliver scalable multi-instance GPU (MIG) performance for deep learning workloads.

    • High-Performance Computing (HPC): 64 DIMMs at 4800 MT/s and PCIe 5.0 I/O maximize throughput for simulations, genomic sequencing, and weather modeling Lenovo Press.

    • Virtualization & VDI: Up to 480 threads and 48 NVMe drives enable high VM density with low boot-storm impact and rapid cloning Lenovo Press.

    Frequently Asked Questions

    Q1: Can the SR860 V3 mix 350 W and 75 W GPUs?
    A1: Yes—you may install up to four double-width 350 W GPUs or up to eight single-width 75 W GPUs, but mixing both widths in the same chassis is not supported.

    Q2: How does the “pay-as-you-grow” CPU upgrade work?
    A2: You can start with two 4th Gen Intel Xeon Scalable CPUs and add a customer-installable mezzanine tray to upgrade to four CPUs and 64 DIMMs without downtime Lenovo Press.

    Q3: What cooling options exist for GPU modules?
    A3: Lenovo Neptune™ liquid-to-air hybrid cooling is available for HGX GPU modules, offering up to 30 % lower acoustic noise and 20 % energy savings over air-only cooling.

    Q4: Which RAID options are supported?
    A4: The SR860 V3 supports ThinkSystem PCIe RAID/HBA cards as well as CPU-based VROC for RAID 0/1 boot volumes; hardware RAID controllers are required for data drives Lenovo Documentation.

    Q5: What network fabrics can be deployed?
    A5: Two OCP 3.0 slots can host adapters for 1/10/25/100 GbE, InfiniBand EDR/HDR, or iWARP, enabling flexible cluster interconnects and low-latency storage fabrics Lenovo Press.

PRODUITS LIÉS À CET ARTICLE
Lenovo SR860V3 Rack Server 4U Hôte GPU à double direction - Serveur de profondeur à haute performance pour les applications OA - Nouveau système Barebone Recommander
Lenovo ThinkSystem SR860 Server - Double Intel Xeon Gold 5118 CPU, plusieurs configurations pour l'ERP, la virtualisation et les applications de base de données Recommander
Libérez les performances inégalées avec le serveur ThinkSystem 4U SR860 V2 - conçu pour les charges de travail exigeantes Recommander