Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

AMD Launches Instinct MI300X AI GPU Accelerator, Up To 60% Faster Than NVIDIA H100

Dec 07, 2023 - wccftech.com
AMD has officially launched its flagship AI GPU accelerator, the MI300X, which reportedly offers up to 60% better performance than NVIDIA's H100. The MI300X, part of the AMD Instinct MI300 class of AI accelerators, uses advanced packaging technologies from TSMC and boasts 2.4X higher memory capacity, 1.6X higher memory bandwidth, and 1.3X FP8 and FP16 TFLOPs compared to the NVIDIA H100. The MI300X also outperforms the H100 in various 1v1 and 8v8 server comparisons.

The MI300X is powered by ROCm 6.0 software stack, which supports various AI workloads and the latest compute formats. The GPU accelerator is designed on the CDNA 3 architecture and hosts a mix of 5nm and 6nm IPs, delivering up to 153 billion transistors. It also offers a significant upgrade in memory, boasting 50% more HBM3 capacity than its predecessor, the MI250X. Despite its powerful performance, AMD's MI300X is rated at 750W, only 50W more than the NVIDIA H200.

Key takeaways:

  • AMD has launched its flagship AI GPU accelerator, the MI300X, which offers up to 60% better performance than NVIDIA's H100.
  • The MI300X is equipped with advanced packaging technologies from TSMC and offers higher memory capacity, higher memory bandwidth, and faster performance in various comparisons with NVIDIA's H100.
  • The MI300X is powered by the ROCm 6.0 software stack, which supports various AI workloads and offers significant speedups in various operations.
  • Despite the competition from NVIDIA and Intel, AMD's MI300X is poised to be a leader in the AI segment, with support from companies like Oracle, Dell, META, and OpenAI.
View Full Article

Comments (0)

Be the first to comment!