Why Latency-Tolerant Architectures Matter in the Age of AI Supercomputing
High Bandwidth Memory (HBM) has become the defining enabler of modern AI accelerators. From NVIDIA’s GB200 Ultra to AMD’s MI400, every new AI chip boasts faster and larger stacks of HBM, pushing memory bandwidth into the terabytes-per-second range. … Read More
The Impact of the AI Revolution on the Server Manufacturers
While some will designate my research as market research, I view it differently. Having done and bought plenty of market sizing research, I have not seen it lead to any change in behaviour or strategy. It has been used to confirm a strategy already decided and the “great” … Read More
