"Temperature and Performance Dataset of a Consumer CPU under Stress Testing"

"Artificial\u2011intelligence inference, cryptographic hashing, and extended gaming keep today\u2019s consumer processors near their thermal limits, allowing latent damage to accumulate long before any drop in benchmark score is noticed. This research captures that \u201cunseen aging\u201d inside a commercial off\u2011the\u2011shelf (COTS) 14\u202fnm mobile CPU (marketed as Intel\u202fCore\u202fi7\u20116600U) by synchronising eight calibrated K\u2011type thermocouples with 25\u202fHz long\u2011wave infrared imaging while the processor executes mixed CPU\u2013GPU workloads representative of gaming, AI, and office use. The fused data reconstruct a three\u2011dimensional, millisecond\u2011resolved temperature field, revealing hotspot migration of \u2248\u202f5\u202fmm within 10\u202fs, lateral gradients up to\u202f9\u202f\u00b0C across the die, and ramp rates of 0.30\u202f\u00b0C\u202fs\u207b\u00b9 during load transitions. A dual modeling approach\u2014comprising Coffin\u2013Manson-based solder fatigue projection and thermo-elastic stress estimation\u2014transforms these spatiotemporal profiles into comparative damage rates across office, gaming, AI, and crypto workloads. Gaming workloads exhibit up to 174\u00d7 accelerated fatigue compared to office use, while constant-load crypto scenarios yield the lowest degradation despite consuming high power. These findings show that thermal cycling, not peak temperature, dominates fatigue-induced aging. Our findings can inform future studies on fatigue-aware workload management and reliability-optimized cooling, by quantifying how different usage patterns contribute to aging."