Nvidia Boosts Custom Silicon Memory with NVHBM

Nvidia has expanded its NVLink Fusion strategy to supercharge custom silicon memory. Dubbed NVHBM, the base-die innovation ties in with NVLink Fusion and extends high-speed NVLink backplane technology to partner specialized custom processing units (XPUs).

With NVHBM, memory controllers are integrated directly into the 3D HBM stack instead of directly on the XPU board. This decision brings up to 30% more memory bandwidth compared with standard high-bandwidth memory generation 4 (HBM4e) and as much as 15% lower power consumption.

The move widens the compute area, allowing partner silicon designers to pack more power onto their devices. By removing HBM from the physical layer of the XPU, it provides up to a 30% increase in available main-die silicon for compute or other features.