AMD’s CDNA 3-based MI300X accelerator takes AI to a new level
AMD's Mi300X is changing the AI game.
Share this article...
WePC is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Prices subject to change. Learn more
AMD announced a plethora of new technology at the recent AMD Data Center & AI Technology Premiere. Within the impressive myriad of tech announced came the MI300X, a variant of the AI-accelerated MI300A APU. This is how AMD’s CDNA 3-based Mi300X Accelerator takes AI to a new level.
AMD announces MI300X AI Accelerator
AMD has shared new information about their AMD Instinct MI300 Series accelerator family. They have introduced the AMD Instinct MI300X accelerator, which is considered the most advanced accelerator for generative AI.
If you think this is great, you should take a look at the 128-core Genoa-X CPU AMD just announced.
This accelerator is based on the next-generation AMD CDNA 3 architecture. It and can support up to 192 GB of HBM3 memory. The MI300X is designed to provide efficient computing and memory capabilities for the training and inference of large language models in generative AI tasks.
With the AMD Instinct MI300X’s ample memory, customers can now run large language models, such as Falcon-40, which has 40 billion parameters, on a single MI300X accelerator. Not only that, but the MI300X can support running an 80 billion parameter AI model Using FP16 inferencing.
AMD has also unveiled the AMD Instinct Platform, which brings together eight MI300X accelerators in a standard design. This platform offers a comprehensive solution for AI inference and training.
The MI300X accelerator is currently being sampled for key customers, with availability expected in the third quarter.
Additionally, AMD has announced the sampling of the AMD Instinct MI300A, the world’s first APU Accelerator for high-performance computing (HPC) and AI workloads, to customers.
AMD Instinct MI300X performance
AMD CEO, DR. Lisa, Su wasn’t afraid to talk numbers on the upcoming MI300X AI Accelerator.
When compared to Nvidias H100 AI GPU, the MI300X has up to 2.4X the HBM density and 1.6X the HBM bandwidth. This could be a chance for AMD to chip into some of Nvidias sweet AI market share. This is because, currently, Nvidia is seen as leading the charge on AI-enabled hardware.
Whilst we do not have any more performance metrics for you just yet, it’s exciting to see AMD dive head-first into AI acceleration. And show just how much they can achieve in such a short space of time. This was how AMD’s CDNA 3-based Mi300X Accelerator takes AI to a new level.
If you missed the AMD data center and AI technology premiere, you can watch it on Youtube.
