NVIDIA NVLink Fusion Expands With NVHBM Custom High-Bandwidth Memory
Article image or reusable cover for NVIDIA
NVIDIA has expanded its NVLink Fusion platform with NVHBM, a new high-speed memory technology.
Instead of placing the memory controller on the AI chip itself (XPU), NVHBM places it in the memory stack, delivering 30 percent higher memory speed, 15 percent lower power consumption, and freeing up 25 percent more space for computation on the chip. Amazon will be the first to use the technology with its future Trainium4 chips.
By integrating the memory controller into the 3D HBM stack instead of the XPU, NVHBM delivers up to 30% greater memory bandwidth and 15% lower HBM power consumption, and frees up to 25% more area on XPU compute die compared with standard HBM4E.
Read the full story at NVIDIA →
Vibekollen prepared this summary with AI from the original publication. The content belongs to NVIDIA.
