March 18th, 2019 by Adam Armstrong
Lenovo Makes Several AI Announcements At GTC 2019
Today at the 10th annual GPU Technology Conference, Lenovo made several announcements centered on AI. Lenovo announced the launch of AI Workshops, AI innovation centers aimed at helping customers to best leverage AI. The company also announced an enhancement to its ThinkSystem SR670 server that is ideal for AI workloads.
AI will be perhaps the most disruptive change over the next few years. AI comes with promises of making business more efficient to increasing treatment in healthcare with better diagnosis to making us all safer with autonomous vehicles. However, AI is not as easy as flipping a switch. Intelligent applications typically need massive amounts of data, which can be difficult to move around for compliance and security reasons. Part of this issue can be addressed by moving compute power closer to the data, but it can’t be solved through hardware alone.
On this end Lenovo is launching global AI Innovation Centers that can “help customers identify use cases that can deliver business value and execute proof-of-concepts by providing AI expertise and optimized infrastructure to reduce business risk and prove technical viability.” The company is also helping customer procure hardware and software tools to support AI while keeping an eye on TCO. AI development isn’t managed in one spot it takes advantage of several open source software frameworks such as TensorFlow, Caffe, MxNet, and others as well as packaged software applications. To get everything under one umbrella, users can leverage Lenovo intelligent Computing Orchestration (LiCO). According to Lenovo, LiCO simplifies AI development by efficiently managing cluster resources, open source frameworks and typical AI workflows.
The hardware side of things can get complicated quickly as it typically needs lots of performance but can still run other workloads if need be. Lenovo’s ThinkSystem SR670 fits this bill nicely. Designed for scale-out AI workloads, the SR670 also can handle HPC, VDI, video processing and other tasks. The SR670 has two models for different AI workloads, one supporting NVIDIA T4 the other NVIDIA V100 Tensor Core GPUs. Today Lenovo announced that the SR670 can support up to eight T4 GPU’s in its 2U footprint. The current version supports four NVIDIA V100 GPUs, and delivering 448 deep learning teraflops or 28 double-precision teraflops.
Lenovo states that users looking for balance in performance and TCO can do pretty good leveraging the ThinkSystem SR670 and LiCO.