This paper emphasizes a comparative experimental evaluation of three Smith predictor configurations. The three control schemes are tuned using different optimization algorithms; then, they are applied to an Arduino Temperature Control Lab used to test and evaluate their performance. The three deadtime compensators are compared using radial graphs. They were assessed under different conditions using various performance indices: the Integral Square Error, the Total Variations of Control Efforts, the Maximum Overshoot, and the Settling Time for tracking and disturbances changes. The Mejia et al. approach presented a better overall performance than the other two. Finally, the results showed that Smith Predictor is suitable for thermal processes with elevated dead time.