← The AI Systems Map
L2 · ChipsStandard / protocol

PCIe / CXL

Host bus and memory-coherent expansion fabric

PCIe (Gen5/Gen6) remains the host-to-accelerator bus, and CXL (Compute Express Link) adds cache-coherent memory pooling and expansion increasingly used to attack the HBM capacity wall in inference. Both are governed by industry consortia and gate how much memory and bandwidth a server node can address around each GPU.

Latest

PCIe 6.0, CXL 3.x

Use

memory pooling/expansion

How it fits the stack

PCIe / CXL with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.

used bydepends onPCIe / CXLChipsRetimers &connectivity silicon(Astera, Credo)Nvidia Data-Center GPU(Blackwell/Rubin)chokepoint
PCIe / CXLDepends on ↑Feeds ↓

PCIe / CXL in the AI stack. PCIe / CXL with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.

Graph data (text) — 3 entities, 2 relationships
  • PCIe / CXLused byRetimers & connectivity silicon (Astera, Credo)
  • Nvidia Data-Center GPU (Blackwell/Rubin)depends onPCIe / CXL