Why More Processor Cores Can Stop Making a Job Faster

More workers do not create more independent work

Adding processor cores helps when a program has work that can be performed concurrently. It helps less when parts must wait for earlier results, exchange data or pass through a shared bottleneck. The practical question is how the whole application behaves as resources change, not how many cores the machine can advertise.

Lawrence Livermore's parallel-computing tutorial distinguishes strong scaling, where the total problem remains fixed, from weak scaling, where the work per processor remains fixed as the problem grows. The Colorado School of Mines scaling guide makes the same distinction useful for benchmarking. A program that solves a larger problem in the same time has achieved something different from one that solves the original problem faster. Both can matter, but their claims should remain separate.