高带宽内存（High Bandwidth Memory，HBM）是一种基于三维堆叠技术的高性能内存标准，通过将多层DRAM芯片垂直堆叠并以硅通孔（TSV）互联，实现极高的数据传输带宽和能效比。HBM主要用于对内存带宽要求极高的场景，包括人工智能训练、高性能计算及图形处理等领域，是现代GPU和AI加速芯片的重要组成部分。
HBM通过TSV（硅通孔）垂直互连多层DRAM，HBM3为8-12层，TSV是直径约5-10微米的铜柱
Traditional GPU memory relies primarily on HBM, which achieves high bandwidth through 3D stacking and ultra-wide buses but has limited capacity, high expense, and significant power consumption
Storage-class memory exhibits notable disadvantages in access latency and bandwidth compared to HBM's multi-terabyte-per-second bandwidth
SRAM访问延迟约1ns，HBM约100ns，SRAM延迟比HBM低约10-50倍
HBM芯片目前主要由SK海力士和三星供应，产能扩张速度远跟不上需求增长
Traditional GPU memory relies primarily on HBM, which achieves high bandwidth through 3D stacking and ultra-wide buses but has limited capacity, high expense, and significant power consumption
90%UnverifiedStorage-class memory exhibits notable disadvantages in access latency and bandwidth compared to HBM's multi-terabyte-per-second bandwidth
85%UnverifiedHBM芯片目前主要由SK海力士和三星供应，产能扩张速度远跟不上需求增长
85%UnverifiedHBM通过TSV（硅通孔）垂直互连多层DRAM，HBM3为8-12层，TSV是直径约5-10微米的铜柱
50%UnverifiedSRAM访问延迟约1ns，HBM约100ns，SRAM延迟比HBM低约10-50倍
50%