Designed to deliver leadership performance for Generative AI workloads and HPC applications
This powerful GPU is enabling leadership performance for the data center, at any scale. These GPUs are uniquely well suited to power even the most demanding AI and HPC workloads, offering exceptional compute performance, large memory density, high-bandwidth memory and support for specialized data formats.
192 GB of HBM3 memory provides cost-effective generative AI performance for more or larger AI models at scale, so fewer GPUs are needed.
The GPU is optimized for matrix and tensor operations with FP8, FP16, BF16 and INT8 precision, balancing performance and accuracy.
AMD-ROCm open software includes a broad set of programming models, tools, compilers, libraries and runtimes. By supporting APIs deployed by industry leaders, developers can easily port development code.
The instance on IBM Cloud comes with the following specifications
This offering is currently under select availability. Please create a support case if you are interested in purchasing and using AMD MI300X on IBM Cloud.