• TRM$ 3.461,92 +1,37 % +$ 46,82TRM oficial (Superintendencia Financiera), vigente del 21 nov al 23 nov
  • Euro$ 3.804,64 +1,37 % +$ 51,45Referencia: TRM cruzada con la tasa del Banco Central Europeo del 20 nov; no es la cotización de mercado
  • Peso mexicano$ 181,56 +1,37 % +$ 2,45Referencia: TRM cruzada con la tasa del Banco Central Europeo del 20 nov; no es la cotización de mercado
← Volver a la comunidad
Tecnología

Why is backside power delivery the make-or-break battleground between TSMC A16 and Intel 18A for AI chips?

Routing power and signal lines through the same front-side metal layers has become an existential bottleneck for angstrom-era processors. By moving power delivery to the underside of the silicon wafer, Intel's 18A and TSMC's A16 nodes are eliminating voltage droop—but creating a brutal new war over packaging yields and heat dissipation.

In conventional front-side interconnects, power rails compete with signal routing through more than 15 ultra-thin copper layers, causing severe IR voltage drop that wastes up to 15% of an AI accelerator's clock frequency. Intel took the early gamble by simultaneously debuting RibbonFET Gate-All-Around (GAA) nanosheets and its PowerVia Backside Power Delivery Network (BSPDN) on the 18A node, stabilizing wafer yields into the 75% to 80% range for Panther Lake after overcoming early wafer-to-wafer defect variability.

TSMC countered by decoupling the transition—validating nanosheet GAA on its 2nm (N2) node first before introducing its backside Super Power Rail (SPR) architecture on A16 (1.6nm-class) with direct-contact nano-Through-Silicon Vias (nTSVs) that unlock 15% to 20% power savings. Yet flipping the wafer and grinding the substrate down to sub-micron thickness traps heat directly between the backside power metal and the logic die, forcing us to rethink thermal interface materials and 3D die-stacking tolerances.

Will Intel's first-mover advantage with PowerVia on 18A win external foundry customers, or will TSMC's Super Power Rail on A16 dominate high-density AI accelerators?

Which foundry strategy will win the angstrom-era BSPDN race?

0 votos

Comentarios

Sé respetuoso. Los comentarios que reporten varias personas se ocultan hasta que los revise un moderador.

Para leer no necesitas cuenta; solo para comentar. Entras con un código que te llega al correo o con tu cuenta de Google. Mostramos solo el nombre que elijas; nunca tu correo.