Qualcomm to Integrate HBC Architecture into Snapdragon Chips for Enhanced AI
Qualcomm announced it will integrate its High Bandwidth Compute (HBC) architecture into Snapdragon chips. This new near-memory computing approach aims to boost AI performance and energy efficiency on mobile devices.
Qualcomm has announced plans to integrate its High Bandwidth Compute (HBC) architecture into upcoming Snapdragon chipsets. This new approach to near-memory computing aims to enhance the speed and efficiency of artificial intelligence (AI) operations on mobile devices.
The HBC architecture, initially revealed for data center products, positions the compute chip below stacked LPDDR DRAM memory. By utilizing 3D integration and Through-Silicon Vias (TSV), data transfer distances are shortened. The goal is to address the 'memory wall' bottleneck in AI inference.
For end-users, the integration of HBC into Snapdragon chips means mobile phones will be able to run larger AI models locally, faster, and with reduced power consumption. More AI functionalities can be executed directly on the device.
Qualcomm views the HBC architecture as a method to decrease the amount of data moved between the processor and memory, thereby lowering latency and reducing memory power consumption. This is expected to significantly improve energy efficiency and enable the execution of more complex AI models locally.