DenseChip

GPU Design · All levels

GPU Design - Architecture to Silicon

Practical GPU design mastery covering SIMT programming model, SM microarchitecture, memory hierarchy, warp scheduling, graphics pipeline, interconnect, physical design, and bring-up.

Course promise

This is the execution-focused GPU design course for silicon teams. It teaches how to reason from warp-level behavior to system-level performance, then close design risk with reproducible evidence across software, architecture, and implementation.

  • Translate SIMT and graphics concepts into measurable engineering decisions.

  • Diagnose bottlenecks across SM scheduling, memory traffic, and interconnect.

  • Connect architecture choices to timing, power, thermals, and bring-up risk.

  • Use one repeatable framework: workload -> mechanism -> artifact -> owner -> decision.

Course map

diagram
SIMT programming and occupancy
  -> SM/shader core behavior
  -> memory hierarchy and coalescing
  -> warp scheduling and divergence control
  -> graphics and compute pipeline balance
  -> interconnect and memory-system scaling
  -> physical design, thermal, timing closure
  -> verification, counters, and bring-up

Related topics

Full course index

Every section and lesson in this track — expand folders in the sidebar or jump from here.

Open question bankPractice questions for this track