The Lenovo ThinkSystem SR670 V2 is a versatile GPU-rich 3U rack server that supports eight double-wide GPUs including the new NVIDIA A100 and A40 Tensor Core GPUs, or the NVIDIA HGX A100 4-GPU offering with NVLink and Lenovo Neptune hybrid liquid-to-air cooling. The server is based on the new third-generation Intel Xeon Scalable processor family (formerly codenamed “Ice Lake”) and the new Intel Optane Persistent Memory 200 Series.

The server delivers optimal performance for Artificial Intelligence (AI), High Performance Computing (HPC) and graphical workloads across an array of industries. Retail, manufacturing, financial services and healthcare industries can leverage the processing power of the GPUs in the SR670 V2 to extract greater insights and drive innovation utilizing machine learning (ML) and deep learning (DL).

Key features

The SR670 V2 features a modular design for ultimate flexibility. Multiple configurations are supported, including:

• Up to eight double-wide GPUs with NVLink bridges
• NVIDIA HGX A100 4-GPU with NVLink and Lenovo Neptune hybrid liquid cooling
• Choice of front or rear high-speed networking
• Choice of local high speed NVMe storage

There are three different base configurations of the SR670 V2 as shown in the following figure. The configurations determine the type and quantity of GPUs supported as well as the supported drive bays.

Figure 2. Three base configurations of the ThinkSystem SR670 V2

The SR670 V2 is built on two third-generation Intel Xeon Scalable processors and is designed to support the latest GPUs in the NVIDIA Ampere datacenter portfolio. The SR670 V2 delivers performance optimized for your workload, be it visualization, rendering or computationally intensive HPC and AI.

Scalability and performance

The SR670 V2 offers numerous features to boost performance, improve scalability and reduce costs: 

• Supports up to eight high-performance PCIe double-wide GPUs including the new NVIDIA A100 and A40 Tensor Core GPUs.
• Supports the NVIDIA HGX A100 4-GPU complex with NVLink and Lenovo Neptune hybrid liquid cooling.

• Supports NVIDIA NVLink, which offers a GPU-to-GPU direct connection of up to 600 GB/s bandwidth and supported in both SXM and double-wide PCIe GPU configurations. NVLink also allows for a larger combined memory footprint for bigger batch sizes or the processing of larger images.
• Supports two third-generation Intel Xeon Processor Scalable processors
  • Up to 40 cores and 80 threads
  • Core speeds of up to 3.6 GHz
  • TDP ratings of up to 270W
• Supports up to 32 TruDDR4 memory DIMMs operating at up to 3200 MHz means you have the fastest available memory subsystem.
• Supports configurations of 2 DIMMs per channel to operate at the 3200 MHz rated speed of the memory DIMMs.
• Using 64GB RDIMMs, the server supports up to 2TB of system memory.
• Supports the new Intel Optane Persistent Memory 200 Series for advanced in-memory database applications, dense-virtualization; up to 16 PMem Modules can be installed in conjunction with regular system memory.
• A variety of slot configurations available, depending on the GPU and NVMe storage configuration selected:
  • Two front PCIe 4.0 x16 slots
  • Four rear PCIe 4.0 x16 slots
  • One rear OCP 3.0 slot, PCIe 4.0 x8 or x16
• Supports a variety of internal storage configurations:
  • 8x 2.5-inch hot-swap SSDs or HDDs, with SAS, SATA or NVMe interfaces
  • 6x EDSFF E1.S hot-swap NVMe SSDs
  • 4x 3.5-inch hot-swap SATA SSDs

• Supports NVMe drives without oversubscription of PCIe lanes (1:1 connectivity). The use of NVMe drives maximizes drive I/O performance, in terms of throughput, bandwidth, and latency.
• Supports SATA or NVMe drives using the onboard controller, enabling an internal storage solution that doesn’t require a PCIe adapter.

• Supports high-speed RAID controllers from Broadcom providing 12 Gb SAS connectivity to the drive backplanes. A variety of PCIe 3.0 and PCIe 4.0 RAID adapters are available.
• Supports M.2 drives for convenient operating system boot functions. Available M.2 adapters support either one M.2 drive or two M.2 drives in a RAID 1 configuration for performance and reliability.
• The server has an industry-standard OCP 3.0 small form factor (SFF) slot, with a PCIe 4.0 interface, up to x16, supporting a variety of Ethernet network adapters. A simple-swap mechanism with a thumbscrew and pull-tab enables tool-less installation and removal of the adapter. The adapter supports shared BMC network sideband connectivity to enable out-of-band systems management.
• The server offers PCI Express 4.0 I/O expansion capabilities that doubles the theoretical maximum bandwidth of PCIe 3.0 (16GT/s in each direction for PCIe 4.0, compared to 8 GT/s with PCIe 3.0). A PCIe 4.0 x16 slot provides 64 GB/s bandwidth, enough to support a 200GbE network connection.
• The server offers either a choice of PCIe 4.0 x16 full-height slots, depending on the GPU and NVMe connections selected. Available slots are two front slots and four rear slots, plus the slot dedicated to the OCP adapter. The flexibility of configuration ensures you can get the PCIe resources you need for a high-performance HPC/AI implementation.

Availability and serviceability

The SR670 V2 provides many features to simplify serviceability and increase system uptime:

• The server offers Single Device Data Correction (SDDC, also known as Chipkill), Adaptive Double-Device Data Correction (ADDDC, also known as Redundant Bit Steering or RBS) and memory mirroring for redundancy in the event of a non-correctable memory failure.
• The server offers hot-swap drives, supporting RAID redundancy for data protection and greater system uptime.
• Available M.2 RAID Boot Adapters support RAID-1 which can enable two SATA or two NVMe M.2 drives to be configured as a redundant pair.
• The server has four hot-swap redundant power supplies and five simple-swap redundant fans to provide availability for business-critical applications.
• The Liquid Assisted Cooling Module on the configuration with SXM GPUs employs four redundant low-pressure pumps to circulate the liquid to cool the GPUs.
• The light path diagnostics feature uses LEDs to lead the technician to failed (or failing) components, which simplifies servicing, speeds up problem resolution, and helps improve system availability.
• Solid-state drives (SSDs) offer more reliability and performance than traditional mechanical HDDs for greater uptime.
• Proactive Platform Alerts (including PFA and SMART alerts): Processors, voltage regulators, memory, internal storage (SAS/SATA HDDs and SSDs, NVMe SSDs, M.2 storage, flash storage adapters), fans, power supplies, RAID controllers, server ambient and subcomponent temperatures. Alerts can be surfaced through the XClarity Controller to managers such as Lenovo XClarity Administrator, VMware vCenter, and Microsoft System Center. These proactive alerts let you take appropriate actions in advance of possible failure, thereby increasing server uptime and application availability.
• The built-in XClarity Controller continuously monitors system parameters, triggers alerts, and performs recovery actions in case of failures to minimize downtime.
• Built-in diagnostics in UEFI, using Lenovo XClarity Provisioning Manager, speed up troubleshooting tasks to reduce service time.
• Lenovo XClarity Provisioning Manager supports diagnostics and can save service data to a USB key drive or remote CIFS share folder for troubleshooting and reduce service time.
• Auto restart in the event of a momentary loss of AC power (based on power policy setting in the XClarity Controller service processor)
• Offers a diagnostics port on the front of the server to allow you to attach an external diagnostics handset for enhanced systems management capabilities.
• Support for the XClarity Administrator Mobile app running on a supported smartphone or tablet and connected to the server through the service-enabled USB port, enables additional local systems management functions.
• Three-year or one-year customer-replaceable unit and onsite limited warranty (varies by geography), 9 x 5 next business day. Optional service upgrades are available.

Manageability and security

Systems management features simplify local and remote management of the SR670 V2:

• The server includes an XClarity Controller (XCC) to monitor server availability. Optional upgrade to XCC Advanced to provide remote control (keyboard video mouse) functions. Optional upgrade to XCC Enterprise enables the additional support for the mounting of remote media files (ISO and IMG image files), boot capture, and power capping.
• Lenovo XClarity Administrator offers comprehensive hardware management tools that help to increase uptime, reduce costs and improve productivity through advanced server management capabilities.
• UEFI-based Lenovo XClarity Provisioning Manager, accessible from F1 during boot, provides system inventory information, graphical UEFI Setup, platform update function, RAID Setup wizard, operating system installation function, and diagnostic functions.
• Support for Lenovo XClarity Energy Manager, which captures real-time power and temperature data from the server and provides automated controls to lower energy costs.
• Supports Lenovo intelligent Computing Orchestration (LiCO), a powerful platform that manages cluster resources for HPC and AI applications. LiCO supports multiple AI frameworks, including TensorFlow, Caffe, Neon, and MXNet, allowing you to leverage a single cluster for diverse workload requirements.
• An integrated industry-standard Unified Extensible Firmware Interface (UEFI) enables improved setup, configuration, and updates, and simplifies error handling.
• Support for industry standard management protocols, IPMI 2.0, SNMP 3.0, Redfish REST API, serial console via IPMI.
• An integrated hardware Trusted Platform Module (TPM) supporting TPM 2.0 enables advanced cryptographic functionality, such as digital signatures and remote attestation.
• Administrator and power-on passwords help protect from unauthorized access to the server.
• Supports Secure Boot to ensure only a digitally signed operating system can be used. Supported with HDDs and SSDs, as well as 7mm and M.2 drives.
• Industry-standard Advanced Encryption Standard (AES) NI support for faster, stronger encryption.
• Intel Execute Disable Bit functionality can prevent certain classes of malicious buffer overflow attacks when combined with a supported operating system.
• Intel Trusted Execution Technology provides enhanced security through hardware-based resistance to malicious software attacks, allowing an application to run in its own isolated space, protected from all other software running on a system.
• Available chassis intrusion switch as an additional physical security feature.

Energy efficiency

The SR670 V2 offers the following energy-efficiency features to save energy, reduce operational costs, and increase energy availability:

• Energy-efficient system board components help lower operational costs.
• High-efficiency power supplies with 80 PLUS Platinum certification
• Solid-state drives (SSDs) consume as much as 80% less power than traditional spinning 2.5-inch HDDs.
• Optional Lenovo XClarity Energy Manager provides advanced data center power notification, analysis, and policy-based management to help achieve lower heat output and reduced cooling needs.

Components and connectors

The following figure shows the main components on the front of the configuration with 4x SXM GPUs and 8x 2.5-inch hot-swap drives.

Key features

The SR670 V2 features a modular design for ultimate flexibility. Multiple configurations are supported, including:

• Up to eight double-wide GPUs with NVLink bridges
• NVIDIA HGX A100 4-GPU with NVLink and Lenovo Neptune hybrid liquid cooling
• Choice of front or rear high-speed networking
• Choice of local high speed NVMe storage

There are three different base configurations of the SR670 V2 as shown in the following figure. The configurations determine the type and quantity of GPUs supported as well as the supported drive bays.

Figure 2. Three base configurations of the ThinkSystem SR670 V2

The SR670 V2 is built on two third-generation Intel Xeon Scalable processors and is designed to support the latest GPUs in the NVIDIA Ampere datacenter portfolio. The SR670 V2 delivers performance optimized for your workload, be it visualization, rendering or computationally intensive HPC and AI.

Scalability and performance

The SR670 V2 offers numerous features to boost performance, improve scalability and reduce costs: 

• Supports up to eight high-performance PCIe double-wide GPUs including the new NVIDIA A100 and A40 Tensor Core GPUs.
• Supports the NVIDIA HGX A100 4-GPU complex with NVLink and Lenovo Neptune hybrid liquid cooling.

• Supports NVIDIA NVLink, which offers a GPU-to-GPU direct connection of up to 600 GB/s bandwidth and supported in both SXM and double-wide PCIe GPU configurations. NVLink also allows for a larger combined memory footprint for bigger batch sizes or the processing of larger images.
• Supports two third-generation Intel Xeon Processor Scalable processors
  • Up to 40 cores and 80 threads
  • Core speeds of up to 3.6 GHz
  • TDP ratings of up to 270W
• Supports up to 32 TruDDR4 memory DIMMs operating at up to 3200 MHz means you have the fastest available memory subsystem.
• Supports configurations of 2 DIMMs per channel to operate at the 3200 MHz rated speed of the memory DIMMs.
• Using 64GB RDIMMs, the server supports up to 2TB of system memory.
• Supports the new Intel Optane Persistent Memory 200 Series for advanced in-memory database applications, dense-virtualization; up to 16 PMem Modules can be installed in conjunction with regular system memory.
• A variety of slot configurations available, depending on the GPU and NVMe storage configuration selected:
  • Two front PCIe 4.0 x16 slots
  • Four rear PCIe 4.0 x16 slots
  • One rear OCP 3.0 slot, PCIe 4.0 x8 or x16
• Supports a variety of internal storage configurations:
  • 8x 2.5-inch hot-swap SSDs or HDDs, with SAS, SATA or NVMe interfaces
  • 6x EDSFF E1.S hot-swap NVMe SSDs
  • 4x 3.5-inch hot-swap SATA SSDs

• Supports NVMe drives without oversubscription of PCIe lanes (1:1 connectivity). The use of NVMe drives maximizes drive I/O performance, in terms of throughput, bandwidth, and latency.
• Supports SATA or NVMe drives using the onboard controller, enabling an internal storage solution that doesn’t require a PCIe adapter.

• Supports high-speed RAID controllers from Broadcom providing 12 Gb SAS connectivity to the drive backplanes. A variety of PCIe 3.0 and PCIe 4.0 RAID adapters are available.
• Supports M.2 drives for convenient operating system boot functions. Available M.2 adapters support either one M.2 drive or two M.2 drives in a RAID 1 configuration for performance and reliability.
• The server has an industry-standard OCP 3.0 small form factor (SFF) slot, with a PCIe 4.0 interface, up to x16, supporting a variety of Ethernet network adapters. A simple-swap mechanism with a thumbscrew and pull-tab enables tool-less installation and removal of the adapter. The adapter supports shared BMC network sideband connectivity to enable out-of-band systems management.
• The server offers PCI Express 4.0 I/O expansion capabilities that doubles the theoretical maximum bandwidth of PCIe 3.0 (16GT/s in each direction for PCIe 4.0, compared to 8 GT/s with PCIe 3.0). A PCIe 4.0 x16 slot provides 64 GB/s bandwidth, enough to support a 200GbE network connection.
• The server offers either a choice of PCIe 4.0 x16 full-height slots, depending on the GPU and NVMe connections selected. Available slots are two front slots and four rear slots, plus the slot dedicated to the OCP adapter. The flexibility of configuration ensures you can get the PCIe resources you need for a high-performance HPC/AI implementation.

Availability and serviceability

The SR670 V2 provides many features to simplify serviceability and increase system uptime:

• The server offers Single Device Data Correction (SDDC, also known as Chipkill), Adaptive Double-Device Data Correction (ADDDC, also known as Redundant Bit Steering or RBS) and memory mirroring for redundancy in the event of a non-correctable memory failure.
• The server offers hot-swap drives, supporting RAID redundancy for data protection and greater system uptime.
• Available M.2 RAID Boot Adapters support RAID-1 which can enable two SATA or two NVMe M.2 drives to be configured as a redundant pair.
• The server has four hot-swap redundant power supplies and five simple-swap redundant fans to provide availability for business-critical applications.
• The Liquid Assisted Cooling Module on the configuration with SXM GPUs employs four redundant low-pressure pumps to circulate the liquid to cool the GPUs.
• The light path diagnostics feature uses LEDs to lead the technician to failed (or failing) components, which simplifies servicing, speeds up problem resolution, and helps improve system availability.
• Solid-state drives (SSDs) offer more reliability and performance than traditional mechanical HDDs for greater uptime.
• Proactive Platform Alerts (including PFA and SMART alerts): Processors, voltage regulators, memory, internal storage (SAS/SATA HDDs and SSDs, NVMe SSDs, M.2 storage, flash storage adapters), fans, power supplies, RAID controllers, server ambient and subcomponent temperatures. Alerts can be surfaced through the XClarity Controller to managers such as Lenovo XClarity Administrator, VMware vCenter, and Microsoft System Center. These proactive alerts let you take appropriate actions in advance of possible failure, thereby increasing server uptime and application availability.
• The built-in XClarity Controller continuously monitors system parameters, triggers alerts, and performs recovery actions in case of failures to minimize downtime.
• Built-in diagnostics in UEFI, using Lenovo XClarity Provisioning Manager, speed up troubleshooting tasks to reduce service time.
• Lenovo XClarity Provisioning Manager supports diagnostics and can save service data to a USB key drive or remote CIFS share folder for troubleshooting and reduce service time.
• Auto restart in the event of a momentary loss of AC power (based on power policy setting in the XClarity Controller service processor)
• Offers a diagnostics port on the front of the server to allow you to attach an external diagnostics handset for enhanced systems management capabilities.
• Support for the XClarity Administrator Mobile app running on a supported smartphone or tablet and connected to the server through the service-enabled USB port, enables additional local systems management functions.
• Three-year or one-year customer-replaceable unit and onsite limited warranty (varies by geography), 9 x 5 next business day. Optional service upgrades are available.

Manageability and security

Systems management features simplify local and remote management of the SR670 V2:

• The server includes an XClarity Controller (XCC) to monitor server availability. Optional upgrade to XCC Advanced to provide remote control (keyboard video mouse) functions. Optional upgrade to XCC Enterprise enables the additional support for the mounting of remote media files (ISO and IMG image files), boot capture, and power capping.
• Lenovo XClarity Administrator offers comprehensive hardware management tools that help to increase uptime, reduce costs and improve productivity through advanced server management capabilities.
• UEFI-based Lenovo XClarity Provisioning Manager, accessible from F1 during boot, provides system inventory information, graphical UEFI Setup, platform update function, RAID Setup wizard, operating system installation function, and diagnostic functions.
• Support for Lenovo XClarity Energy Manager, which captures real-time power and temperature data from the server and provides automated controls to lower energy costs.
• Supports Lenovo intelligent Computing Orchestration (LiCO), a powerful platform that manages cluster resources for HPC and AI applications. LiCO supports multiple AI frameworks, including TensorFlow, Caffe, Neon, and MXNet, allowing you to leverage a single cluster for diverse workload requirements.
• An integrated industry-standard Unified Extensible Firmware Interface (UEFI) enables improved setup, configuration, and updates, and simplifies error handling.
• Support for industry standard management protocols, IPMI 2.0, SNMP 3.0, Redfish REST API, serial console via IPMI.
• An integrated hardware Trusted Platform Module (TPM) supporting TPM 2.0 enables advanced cryptographic functionality, such as digital signatures and remote attestation.
• Administrator and power-on passwords help protect from unauthorized access to the server.
• Supports Secure Boot to ensure only a digitally signed operating system can be used. Supported with HDDs and SSDs, as well as 7mm and M.2 drives.
• Industry-standard Advanced Encryption Standard (AES) NI support for faster, stronger encryption.
• Intel Execute Disable Bit functionality can prevent certain classes of malicious buffer overflow attacks when combined with a supported operating system.
• Intel Trusted Execution Technology provides enhanced security through hardware-based resistance to malicious software attacks, allowing an application to run in its own isolated space, protected from all other software running on a system.
• Available chassis intrusion switch as an additional physical security feature.

Energy efficiency

The SR670 V2 offers the following energy-efficiency features to save energy, reduce operational costs, and increase energy availability:

• Energy-efficient system board components help lower operational costs.
• High-efficiency power supplies with 80 PLUS Platinum certification
• Solid-state drives (SSDs) consume as much as 80% less power than traditional spinning 2.5-inch HDDs.
• Optional Lenovo XClarity Energy Manager provides advanced data center power notification, analysis, and policy-based management to help achieve lower heat output and reduced cooling needs.

Standard specifications

The following table lists the standard specifications.