Go away, CPU, you’re not needed here … mostly
Nvidia, IBM, and university collaborators have a developed an architecture they say will provide fast fine-grain access to large amounts of data storage for GPU-accelerated applications, such as analytics and machine-learning training.…
