AMD has expanded its ROCm 5.7 platform to include the Radeon RX 7900 XT GPU, providing more options for AI developers and researchers.
The GPU, which features the RDNA 3 GPU architecture, 20GB of on-board memory, and 168 AI accelerators, is designed for machine learning (ML) development workflows using PyTorch.
Leveraging the parallel computing power of desktop GPUs, PyTorch users now have additional local options, reducing reliance on cloud-based solutions.
AMD ROCm 5.7 is an open software platform offering flexibility and performance, providing access to open compute languages, compilers, libraries, and tools for accelerating code development in high-performance computing and AI applications.
The latest version enhances functionality while incorporating features from previous releases. Notable components include HIPTensor, a C++ library for accelerating tensor primitives on AMD Instinct GPUs, and MIGraphX, which optimizes machine learning inference on AMD hardware through graph-level transformations.
ROCm 5.7 has optimized discrete distributions that work with recent HIP compiler optimizations, resulting in improved rocRAND RNG performance.
Erik Hultgren, Software Product Manager at AMD, expressed excitement about the addition to the portfolio, making AI more accessible from both software and hardware perspectives.
