AMD unveils first MLPerf 5.1 Artificial Intelligence training results on Instinct MI350 series

AMD submitted the first MLPerf 5.1 training benchmarks using its Instinct MI350 Series GPUs, including the MI355X and MI350X, marking the first public Artificial Intelligence training results for the new accelerator family.

AMD announced the first MLPerf 5.1 Training submission that uses the Instinct MI350 Series GPUs. The submission covers both the MI355X and MI350X models and represents the first public benchmark of the Instinct MI350 Series for Artificial Intelligence training workloads. The release follows AMD’s earlier MLPerf 5.1 Inference results and extends the company’s public performance data into training scenarios.

The MLPerf 5.1 Training results highlight generational performance gains for the Instinct MI350 Series. According to the announcement, the benchmarks demonstrate clear progress in scalability, efficiency, and compute performance across several of today’s most demanding Artificial Intelligence training workloads. AMD also emphasized broad ecosystem participation in the submission, indicating that multiple software and system partners were involved in validating the training runs on the new hardware.

By publicizing these MLPerf 5.1 Training results, AMD positions the Instinct MI350 Series as a platform aimed at accelerating the development of next-generation generative Artificial Intelligence models. The company framed the results as evidence that the MI355X and MI350X deliver the compute and efficiency characteristics needed for large-scale training. Overall, the submission serves both as a milestone for the Instinct MI350 Series and as a continuation of AMD’s benchmarking efforts across inference and training workloads.

55

Impact Score

Rdma for s3-compatible storage accelerates Artificial Intelligence workloads

Rdma for S3-compatible storage uses remote direct memory access to speed S3-API object storage access for Artificial Intelligence workloads, reducing latency, lowering CPU use and improving throughput. Nvidia and multiple storage vendors are integrating client and server libraries to enable faster, portable data access across on premises and cloud environments.

technologies that could help end animal testing

The uk has set timelines to phase out many forms of animal testing while regulators and researchers explore alternatives. The strategy highlights organs on chips, organoids, digital twins and Artificial Intelligence as tools that could reduce or replace animal use.

Nvidia to sell fully integrated Artificial Intelligence servers

A report picked up on Tom’s Hardware and discussed on Hacker News says Nvidia is preparing to sell fully built rack and tray assemblies that include Vera CPUs, Rubin GPUs and integrated cooling, moving beyond supplying only GPUs and components for Artificial Intelligence workloads.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.