Nope. Look at nvidia kepler and maxwell GPUs. Same manufacturing process but maxwell spanked kepler due to more efficient architecture. As we go down to smaller nodes, there will be a larger focus on architecture and software stack to squeeze ever bit of performance.
Besides, improvements can always be made to the manufacturing nodes. Aside from making transistors smaller, you can make them leak less energy, improve power delivery circuitry to optimize clocks ect. There are lots of things can be done outside just transistor shrinking to improve performance and power attributes of any chip (GPUs included)