AMD launched a new Instinct MI300X AI chip that is specifically designed to handle generative AI workloads. The chip will be available later this year.
The MI300X utilizes the advanced AMD CDNA 3 accelerator architecture and has a capacity of up to 192 GB of HBM3 memory. This enables efficient computation and memory utilization, making it ideal for the training and inference of large language models (LLMs) in generative AI tasks. LLMs such as Falcon-40 can now fit on a single MI300X chip.
In addition, AMD also introduced the AMD Instinct Platform, which is powered by eight MI3000X chips. It is a self-contained platform for running generative AI inference and training workloads and is an alternative to Nvidia’s DGX supercomputer platform.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.