Pixel 11's Tensor G6 and Pixel 10's Tensor G5 still fall behind in performance per watt, testing shows

Chinese hardware reviewer Geekerwan has published what appears to be the first power-efficiency test of both Google's Tensor G6 and its predecessor, the Tensor G5, in a Bilibili post. The chips are built on TSMC's 3nm process (N3E for the G5 and N3P for the G6). The test plots performance against power draw in CPU (Geekbench 6) and GPU (Steel Nomad Light) workloads and compares the results with those of several rival chips, including Huawei's Kirin 9030 Pro, which was reportedly built on SMIC's N+3 process, a node roughly comparable to TSMC's N6.

On the CPU curve, the Tensor G6 tracks closely with the Tensor G5, Snapdragon 8 Gen 2, and Kirin 9030 Pro throughout the low-to-mid power range, topping out at roughly 8,000 points near 11W. That leaves it well behind the Snapdragon 8 Elite curve, which reaches about 10,100 points at 15W, and even further behind the Apple A17 Pro and A19 Pro data points, which show higher scores at comparable or lower power draw.

The GPU curve shows a similar gap. The Tensor G6's Steel Nomad Light result peaks at around 1,200 points at roughly 6W, again close to the Tensor G5, Snapdragon 8 Gen 2, and Kirin 9030 Pro curves, while the Snapdragon 8 Elite reaches about 2,550 points near 9.5W and the Apple chips post much higher scores.

Geekerwan summarized the results by placing the Tensor G6's CPU efficiency roughly in line with that of the Snapdragon 8 Gen 3 and its GPU efficiency closer to that of the Snapdragon 8 Gen 2. This puts it a few years behind current flagship silicon from Qualcomm and Apple. The Kirin 9030 Pro's near parity in efficiency with the Tensor G5 is notable, given that it uses an older, less dense process than the 3nm chips against which it is plotted.