CPU Performance

CPU Performance Walls

The CPU performance walls are the four hardware barriers that collectively ended the era of “next year’s chip runs your code faster for free.”

These walls motivate basically every modern CPU trick:

Why?

Before ~2005, CPUs got faster every generation — higher clock, same program, more speed. Then the power wall hit and clock scaling died. You’d think you could just go around it, but then we also hit three other walls too, and now we are surrounded.

#WallWhat it isHow we respond
1Power wallPast ~3 GHz, higher clock needs more voltage → more heat → thermal failure. Hit around 2005.Stop cranking the clock; go multicore and wider instead
2ILP wallBranch prediction already ~95% accurate; diminishing returns on further ILPBigger reorder buffers, more functional units, OoO — small gains
3Memory wallDRAM latency grew ~3× slower than CPU speed → cache misses dominate runtimeMulti-level cache, miss shadow, prefetch, overlap misses via OoO
4Speed of lightSignal propagation across the die; fundamental physicsShorten wires, stack dies, accept it

Practical implication: when someone asks “why doesn’t my code just get faster every year anymore?”, the bottleneck is one of these, usually the memory wall.