How much do SoCs change battery life?

01 Tem 06:31

The truth is that the relationship between the manufacturing node (nm) and battery life isn't as linear as many believe. The jump from 5nm to 3nm doesn't just improve raw performance (as CPU/GPU benchmarks show)—it also introduces changes in energy efficiency *by design*. But here's the catch: it's not magic. TSMC's 3nm (for example) prioritizes performance-per-watt over absolute idle power consumption. That means you gain a lot of battery life in intensive tasks (like gaming or rendering), but in light usage (browsing, social media), the savings can be minimal or even nonexistent if the SoC isn't well optimized for that scenario. What explains the "imbalance" you're referring to is the *efficiency core* vs. *performance core*. Qualcomm, MediaTek, or Apple use small/big cores to distribute the load. A 3nm chip may have less raw power but better IPC (instructions per cycle), meaning fewer CPU cycles for the same task. However, the manufacturer must fine-tune the *PMIC* (power management IC) and the software stack to take advantage of it. This is where we see the big difference: a poorly optimized Snapdragon 8 Gen 2 (4nm) on Android can consume more than a well-tuned Exynos 2200 (5nm). The key lies in how the *big core* is managed—if it's constantly active due to bad coding, the new node's efficiency gains vanish. And speaking of Android... here's the big problem! Custom kernels (like those in LineageOS) and governor mods in kernels like *EAS* or *cgroups* can squeeze out an extra 15-20% battery life compared to a poorly optimized stock kernel. For example, a 4nm chip with a stock kernel might last less than a 5nm chip with a modded kernel. The node matters, but the software matters even more. A real-world example? Compare a Pixel 7 Pro (5nm) with the same battery as a OnePlus 11 (4nm): real-world battery life is usually similar, even though the latter has a "more advanced" SoC. In short: yes, 3nm chips win in *proportional* efficiency (more performance per watt), but in practice, it depends on how they're integrated into the device. OEMs prioritize marketing ("3nm! Revolution!") over real optimization, and that's the trick. If you want to see noticeable differences, look at benchmarks like *Geekbench* (for IPC) or *PCMark* (for real-world use), not generic wattmeters like AnTuTu.