The Instinct MI300X accelerator, an advanced accelerator for generative AI, has been introduced by chipmaker AMD, along with additional information about its Instinct MI300 Series accelerator family processors. The launch of our Instinct MI300 accelerators later this year and the expanding ecosystem of enterprise-ready AI software optimised for our hardware are driving our laser-focused effort to hasten the deployment of AMD AI platforms at scale in the data centre, according to AMD Chair and CEO Dr. Lisa Su.
According to the manufacturer, the MI300X is built on the AMD CDNA 3 accelerator architecture of the next generation and supports up to 192 GB of HBM3 memory to deliver the compute and memory efficiency required for big language model training and inference for generative AI workloads.
With the AMD Instinct MI300X processor's large memory, customers can now fit big language models like Falcon-40, which has 40B parameters, on a single MI300X accelerator. The chip maker also released the AMD Instinct Platform, which combines eight MI300X accelerators into an industry-standard design for the best option for AI inference and training.
According to AMD, major customers will begin receiving samples of the MI300X processor in the third quarter. The AMD Instinct MI300A, the first APU accelerator for HPC and AI workloads, was also revealed by the firm as being available for sampling to clients.
Image Source: Twitter
© Copyright 2023. All Rights Reserved Powered by Vygr Media.
