.png)
The AMD Instinct MI350X and MI355X GPUs are AMD’s most powerful accelerators yet, purpose-built for the demands of AI training, inference at scale, and high-performance computing.
Engineered on the latest CDNA™ 4 architecture and advanced 3nm process, these GPUs deliver industry-leading efficiency and performance for next-gen workloads.
- Massive Onboard Memory: Up to 288GB HBM3e and over 2.6 TB/s bandwidth—perfect for large language models and memory-hungry AI workloads.
​
- Breakthrough Performance: Up to 5× faster AI inference than previous generations, enabling faster training and real-time inference for transformers, LLMs, and agentic AI systems.
​
- Multi-GPU Scaling: Support for 8-GPU configurations via Infinity Fabric, providing ultra-low latency and high throughput between accelerators.
​
- Superior Efficiency: Leading performance-per-watt, lowering data center energy costs while maximizing compute density.
​
- Ready for AI at Scale: Fully compatible with the AMD ROCm ecosystem, ensuring seamless integration with frameworks like PyTorch, TensorFlow, and Hugging Face.
.png)