Microsoft adds AMD chips to its pool of AI-focused processors

by Kareem


Kareem is a journalist from the bay area, now living in Florida. His passion for technology and content creation drives are unmatched, driving him to create well-researched articles… read more

AMD Instinct MI300X AI chip

While Microsoft partners with NVIDIA for processors in its flagship Surface Laptop Studio PCs, the company is opting to use AMD accelerator Instinct M1300X chips in its datacenters.

During AMD’s Advancing AI Event, the company took the stage to off a bevy of news regarding its plan to combat the recent influx of NVIDIA-led processor competition surrounding AI focused GPUs and demand.

On stage, AMD CEO Lisa Su unveiled the Instinct MI300X GPU and hybrid CPU-GPU MI300A APU which the company claims have a 1.6X lead ahead of NVIDIA’s own competing GPU performance.

In part, AMD was able to achieve its claimed supremacy with its new MI300 architecture that includes a design that involves 13 chiplets built in a 3D-stack that incorporates twenty-four Zen 4 CPU cores meshed with CDNA 3 graphics engine and eight stacks of high-bandwidth memory (HBM3) with an overall size of 146 billion transistors.

The Instinct M1300X comes with 192GB of HBM3 that equates to a 2x higher limit than NVIDIA’s H100 SXM GPU 80GB HMB3 ceiling. Not satisfied to rest on its laurels, AMD’s Instinct M1300 was also built to house a higher 141GBs of HBM3e which should also outperform NVIDIA’s newly announced H200 chip.

The M1300X marks the largest chip that AMD has produced to date.

Moving down a notch on the performance ladder, the M1300A copies the same 86-based Zen 4 core architecture as the companies EPYC chips that host a combined GPU and CPU core with 128GB of HBM3 memory and shoul produce up to a 60% higher capacity over NVIDIA’s H100 GPU as well.

As for the performance gains over the current industry leader NVIDIA’s H100, AMD claims its M1300 AI chip can reach 122.6 teraflops of FP64 matrix operations which would put it ahead of the H100 by almost 80%.

AMD has already taken orders for its new GPUs from Dell, with other companies such as HP, Lenovo, and Supermicro looking to take advantage of competition in the market.

While Microsoft is working on its own OpenAI partnered AI server chips, it was among the largest cloud service providers to back AMD’s new release as it plans to use the M1300X to power virtual machines running on Azure.

More Reading

Post navigation

Leave a Comment

Leave a Reply

Your email address will not be published. Required fields are marked *