Advanced Micro Devices (AMD) has shared more details about its upcoming code-named “Vega” family of GPUs, as well as information regarding the progress the company has made with its open-source ROCm software stack for HPC and Machine Learning (ML). Investors and AI practitioners alike would like to see AMD enter this market as a second source alternative. NVIDIA has enjoyed envious growth in selling GPUs to accelerate Machine Learning for Artificial Intelligence applications, having grown its Datacenter Business by roughly 200% YOY in the last 2 quarters and approaching a $1.6B in run-rate revenues. While we still await detailed product specs and application-level benchmarks from AMD, and benchmarks run by third parties, these new Vega GPUs look like they have sufficient raw performance (as measured by FLOPS) to be competitive with NVIDIA’s family of P100 PASCAL GPUs. Taken together with the optimized ML software AMD has developed, they form a solid first step for market entry.
You can download the paper here.
Table of Contents
- The New Vega GPU’s
- The Machine Learning Software
- Figure 1: AMD Software Optimization for Machine Learning