AMD’s Series 350 GPUs

It compares well against Nvidia’s B200.

June 16, 2025

Jon Peddie

At its Advancing AI event, AMD launched the MI350 series GPUs, built on the new CDNA 4 architecture. These accelerators challenge Nvidia’s B200 on cost-performance for small-to-medium AI model inference but are not a rack-scale competitor to the GB200 NVL72 for large-scale training. The architecture uses eight compute chiplets and 288GB of HBM3E memory. Its key innovation is support for new, reduced-precision data types like FP4, dramatically increasing AI throughput and efficiency. The series includes an air-cooled MI350X and a more powerful, liquid-cooled MI355X, connecting eight GPUs per node via Infinity Fabric. AMD Instinct MI350 chiplet construction. (Source: AMD) At

...

Enjoy full access with a TechWatch subscription!

TechWatch is the front line of JPR information gathering service, comprising current stories of interest to the graphics industry spanning the core areas of graphics hardware and software, workstations, gaming, and design.

A subscription to TechWatch includes 4 hours of consulting time to be used over the course of the subscription.

Already a subscriber? Login below

AMD’s Series 350 GPUs

Enjoy full access with a TechWatch subscription!

This content is restricted

Login