Lenovo introduced a new suite of artificial intelligence (AI)-centric infrastructure systems designed to accelerate AI innovation from the cloud to the edge.

Built in collaboration with AMD, Lenovo’s new ThinkSystem SR685a V3, ThinkSystem SD535 V3 and ThinkAgile MX455 V3 servers all leverage Lenovo TruScale. This service “uniquely equips businesses with immediate access to edge AI deployments and Lenovo’s turnkey AI solutions, which address a wide range of applications spanning numerous industries,” Lenovo VP and GM of Server, Storage and Software-defined Infrastructure Kamran Amini told SDxCentral.

TruScale enables the delivery of end-to-end Lenovo-managed solutions with “a predictable and manageable monthly payment” based on exact usage. In particular, “our new ThinkSystem and ThinkAgile solutions using Lenovo TruScale offer financial services and health care organizations advanced generative AI and [large language model] LLM workloads.”

The Lenovo ThinkSystem SR685a V3 server includes eight AMD Instinct MI300X GPUs, 1.5TB of high bandwidth memory capacity and up to 1TB/s of peak aggregate theoretical GPU I/O bandwidth performance – suitable for training and deploying AI models.

To that point, the ThinkSystem SR685a V3 is an ideal server for enterprise private on-premises AI and for public AI cloud service providers.

“We understand that enterprise customers are requiring increased acceleration as they go further on their AI journey. Up until recently, this level of compute had only been required by the largest enterprises and cloud service providers,” Amani said. “We developed the new ThinkSystem SR685a V3 server that comes with 8 GPUs for increased acceleration and advanced air-cooling thermals to sustain maximum performance without hitting any thermal limits – a key requirement for operation in an enterprise data center environment.”

With increased thermal headroom from air-cooling, the server provides all eight GPUs with the compute necessary for cloud service providers to support their customers’ AI requirements. The ThinkSystem SR685a V3 also includes system management software optimized for both on-premises AI and public cloud AI service providers.

“We have a proven track record of bringing amazing solutions, jointly, to the market and are excited to add AMD Instinct MI300X to Lenovo’s portfolio, helping enterprises to rapidly adopt AI,” EVP and GM of AMD’s Data Center Solutions Group Forrest Norrod said.

Lenovo’s ThinkSystem SD535 V3 is a multi-node, thermally efficient server built for maximum performance per rack with intensive transaction processing. This one-socket (1S) and one-rack unit (1U) server node is powered by a sole fourth-generation AMD EPYC processor and is ideal for use cases like high performance computing (HPC) (HPC), virtualization at scale and big data analytics.

The ThinkSystem SD535 V3 supports both Intel and AMD CPUs in a 2U four-node configuration, with the ability to scale from two to four nodes within the ThinkSystem D3 chassis for deployment.

In addition, the server’s unified power and cooling implementation reduces power consumption in the data center by up to 30% compared to a standard 1U rack server. The ThinkSystem SD535 V3 “has 4 simple-swap 40mm dual-rotor fans which are used to cool all components. All fans are standard. In addition, each power supply has its own integrated fan,” Amini said.

Lenovo also introduced the new ThinkAgile MX455 V3 Edge Premier Solution, an AI-optimized platform that improves compute and storage performance at the edge. In particular, this platform promises significant total cost of ownership (TCO) reduction and touts “the best power efficiency of any Azure Stack HCI Solution,” according to Lenovo.

Both Microsoft and Lenovo continuously tested the AI-optimized platform during development, and found the solution provides twice the CPU performance with half the power consumption compared to the closest competing offerings. These results were validated by both vendors.

In terms of reducing TCO, the system can be deployed with Lenovo Open Cloud Automation, which accelerates distributed infrastructure rollouts with less resources required. “This simplifies the logistics chain and reduces carbon emissions of the deployment. Faster maintenance of the solutions results in improved uptime, which reduces TCO,” Amini said. “Organizations also see total cost improvements with Lenovo through the improved security of each solution, which ensures staying updated with the latest bug fixes and software updates is made easier.”

This AI-optimized platform delivers AI, compute and storage performance at the edge Highlights of the ThinkAgile MX455 V3 Edge Premier Solution include advanced metadata handling, near-zero-touch provisioning and template-based infrastructure deployment orchestration.