Lenovo markets the ThinkSystem SR685a V3 8GPU server, which the company developed together with chip manufacturer AMD. According to Lenovo, the server is particularly suitable for performance that requires a lot of computing power, such as – how could it be otherwise – GenAI and Large Language Models (LLM). It also comes with other AMD-equipped products and offers a complete suite of tools to manage the released AI computing power.
Choosing AMD as a partner ensures versatile applications. After all, AMD is good at designing both CPUs and GPUs, where the competition focuses on CPUs (Intel) or GPUs (Nvidia). This new infrastructure underlines that: it runs on a range of AMD CPUs and GPUs. The Infinity Fabric connection architecture is also from AMD.
One server contains eight AMD Instinct MI300X GPUs for tasks such as machine learning and deep learning and is powered by a fourth-generation AMD EPYC processor. The aforementioned Infinity Fabric provides the interconnect between the GPUs. This provides 1.5 TB of HBM3 memory capacity and up to 1 TB/s (peak) theoretical GPU I/O bandwidth.
This makes the product suitable for processing large data sets needed in sectors such as financial services, healthcare, energy, (climate) science and transport. It is intended for use both within enterprises (on-premises) and for AI cloud service providers.
Thanks to significant thermal capacity via air cooling, this server can maintain maximum performance for both powerful CPUs and GPUs, according to Lenovo. According to the company, it is suitable for future generations of AMD CPUs, but also offers support for Nvidia’s HGX GPUs.
Efficient 1U rack server
It is not the only collaboration between Lenovo and AMD. The duo also comes with the Lenovo ThinkSystem SD535 V3, an efficient 1U rack server powered by a single fourth-generation AMD EPYC processor. Use cases include cloud computing, virtualization, big data analytics and e-commerce transactions. According to the company, this server offers a high degree of flexibility by supporting both AMD and Intel CPU architectures within the same enclosure.
It offers a degree of scalability with configuration options for two to four nodes in a single chassis, to integrate separate server units within a single physical enclosure. In addition, this server is said to be more energy efficient: it has an integrated power and cooling implementation that should ensure up to thirty percent less energy consumption compared to standard 1U rack servers. It is managed via Lenovo XClarity software.
Definitely not an edge case
For edge networking, Lenovo introduces the ThinkAgile MX455 V3 Edge Premier Solution, which is also equipped with AMD processors. In this case the EPYC 8004 CPU for AI inferencing and real-time data analysis in edge computing. It integrates with both on-premises and Azure cloud environments. Key features include automation, near zero-touch provisioning and Lenovo Open Cloud Automation (LOC-A) for rapid deployment and management from anywhere. It also offers automated software updates validated by both Microsoft and Lenovo.
Lingering all these new products is Lenovo AI Advisory and Professional Services. This includes a range of services, solutions and platforms designed to help companies brave the AI landscape. The set of services must implement the right AI solution quickly, cost-effectively and at scale, from initial exploration and advice to implementation, management and scaling.
Read more: AMD claims leadership in the battle for the fastest AI PC
