Particle.news
Download on the App Store
14 ARTICLES
·
6d ago

AMD Unveils MI350X and MI355X AI GPUs to Challenge Nvidia

Industry leaders’ adoption of the MI350X signals momentum for AMD’s open-standards AI platform ahead of Helios server shipments planned for 2026.

Image
Image
Lisa Su, CEO of AMD, attends the Artificial Intelligence (AI) Action Summit at the Grand Palais in Paris, France, February 10, 2025. REUTERS/Benoit Tessier/File Photo
A smartphone with a displayed AMD logo is placed on a computer motherboard in this illustration taken March 6, 2023. REUTERS/Dado Ruvic/Illustration/File Photo

Overview

  • AMD introduced its Instinct MI350X and MI355X GPUs at its Advancing AI event, citing up to 3× gains in training performance and a 35× jump in inferencing versus the MI300X.
  • The liquid-cooled MI355X features 288 GB of HBM3E memory, 8 TB/s bandwidth and a 1.4 kW board power envelope, matching or exceeding Nvidia Blackwell chips on key FP4 and training benchmarks.
  • The ROCm7 software stack now delivers native support for major AI frameworks, expanded hardware compatibility and new development tools to boost performance on AMD accelerators.
  • OpenAI, Meta, Oracle and Microsoft have begun deploying MI350-series GPUs in their AI clusters, with OpenAI’s Sam Altman confirming plans to use the chips for ChatGPT workloads.
  • AMD plans to launch Helios rack-scale servers in 2026, integrating 72 MI400-series GPUs and open UALink networking standards to rival Nvidia’s proprietary rack designs.