2 Compute Nodes per 1U
Lenovo High-Density HPC Water Cooled
Lenovo ThinkSystem SD650 V3 Server
Direct Water Cooled · 2-Node per 1U
Lenovo ThinkSystem SD650 V3 Server leverages direct water cooling to pack 2 full compute nodes into a single 1U slot — delivering unprecedented performance per rack unit. Designed for high-performance computing environments, scientific research, and simulation workloads, its direct liquid cooling eliminates thermal constraints that limit traditional air-cooled systems. This means higher core counts, higher sustained clock speeds, and dramatically lower energy costs. For HPC clusters where every watt and every rack unit counts, the SD650 V3 is the unmatched choice.
Ideal For HPC / Research Centers
2 Compute Nodes per 1U Direct Liquid Cooling (DLC)
Click image to enlarge · Full resolution available
Key Specifications
Direct Liquid Cooling (DLC)
Ultra-High Core Density
HPC Optimized Architecture
Ultra-Dense Blade Server Lenovo ThinkSystem SD650 V3 with Neptune Liquid Cooling
Lenovo ThinkSystem SD650 V3 is an ultra-dense, 2-socket high-performance blade server (blade node) occupying half-width in the specialized ThinkSystem D3 Chassis (2U form factor). The platform is purpose-built for supercomputing, high-performance computing (HPC), large-scale cloud rendering, AI inference, and complex scientific simulations. Its key technological feature is integration with the Lenovo Neptune™ Direct Water Cooling (DWC) system. Coolant is delivered directly to key components, eliminating the need for traditional air fans in the node and achieving extreme data center energy efficiency.
A single 2U ThinkSystem D3 chassis accommodates up to four independent SD650 V3 blade nodes, allowing up to eight processors in minimal rack space. This makes the platform the optimal choice for supercomputing centers where compute density and energy efficiency are decisive factors.
Use Cases
The Lenovo ThinkSystem SD650 V3 is the optimal solution for organizations requiring:
- High-performance computing (HPC) and supercomputing clusters
- Complex scientific simulations and modeling (CFD, FEA, molecular dynamics)
- Large-scale cloud rendering and 3D visualization
- AI inference in high-density compute environments
- Ultra-low-latency clusters based on InfiniBand
- Workloads requiring maximum energy efficiency and reduced data center PUE
Energy Efficiency and Lenovo Neptune™ Technology
Thanks to water cooling, the server delivers unique infrastructure metrics:
- Heat removal. Direct liquid cooling removes up to 98% of the heat generated by processors, memory, voltage regulator modules (VRM), network cards, and local NVMe drives.
- PUE reduction. The absence of internal fans and the ability to use inlet water at temperatures up to +45 °C minimize chiller costs. This reduces total data center electricity costs by up to 40%.
- Heat reuse. Hot water from the cooling loop (reaching +50 °C…+55 °C at the outlet) can be redirected to heat buildings or for facility domestic needs.
Compute Node Technical Specifications
Compute Subsystem. Each node supports two 4th Gen Intel Xeon Scalable (Sapphire Rapids) or 5th Gen Intel Xeon Scalable (Emerald Rapids) processors with TDP up to 350 W per socket. The architecture supports maximum Turbo frequency operation without risk of overheating or throttling — critical for HPC workloads sensitive to sustained performance.
Memory. The node provides 16 DIMM slots (8 per processor) for high-speed DDR5 ECC RDIMM memory. Maximum memory capacity per node is 2 TB at speeds up to 5600 MT/s. All memory modules are covered by specialized copper-loop water blocks, ensuring stable operation under sustained load.
Storage Subsystem. In the base configuration, the node accommodates up to 2 × 2.5″ (SFF) NVMe PCIe Gen5 hot-swap drives (front-panel access), or up to 4 × E1.S NVMe drives. Additionally, 2 × M.2 NVMe slots are soldered on the board for mirrored OS boot.
Network Interfaces and Interconnect. The board design supports up to 2 high-speed PCIe Gen5 x16 adapters (OCP 3.0 and custom DWC form factors), optimized for InfiniBand ND 400 Gbps or Ethernet 200/400 GbE cards. This is critical for building ultra-low-latency compute clusters.
Management and Fault Tolerance
- Centralized Power. Node power is supplied centrally through the Power Bus Bar of the ThinkSystem D3 chassis. The chassis accommodates up to 3 redundant power supplies rated at 2400 W or 3000 W with Titanium (96%+) certification.
- Monitoring. Each node is equipped with the Lenovo XClarity Controller 2 (XCC2) service processor. Cooling parameters (inlet/outlet water temperature, flow rate, leak detection) are fully integrated with Lenovo XClarity Energy Manager software.
- Security. The ThinkShield hardware suite (TPM 2.0 crypto chip, Platform Firmware Resiliency code verification compliant with NIST SP 800-193) protects each compute node from BIOS firmware compromise.
Why GHI SERVERS
- Direct supply from China. We work with CNY and RUB. Turnkey logistics of server equipment from China, including customs clearance and delivery to our Moscow warehouse.
- Custom configuration. We help select blade node configurations, ThinkSystem D3 chassis, power supplies, InfiniBand/Ethernet network adapters, and cooling systems for your task and budget.
- Moscow warehouse. Prompt replacement of defective units and warranty service without long downtime.
- 20+ years of expertise. Authorized Lenovo partner since 2005. Pre-sales engineering, configuration assistance, and post-sales support.
Service and Support
- 24/7 technical support
- Pre-sales engineering and configuration assistance
- OEM warranty and full provenance documentation
- Installation, setup, and deployment of server equipment
- Component replacement and warranty service
Interested in Lenovo ThinkSystem SD650 V3 Server?
Our certified engineers will configure the right solution for your workloads and budget.