Google Tensor G3 is Google’s 3rd generation of custom chip, which is going to be used inside the Google Pixel 8 series. Unlike Apple and Qualcomm, Google fabricates their chip on the same foundry as Exynos.
Google Tensor G3 will reportedly use Samsung’s 4nm process node, the same as Exynos 2300. The next gen Tensor chip also resembles Samsung’s unreleased Exynos 2300 SoC in terms of core structure.
The only difference is that the Tensor G3 has one more mid-core than Exynos 2300. But that doesn’t mean that the overall experience with both chips are going to be similar.
While Tensor G3 has similar core structure as Exynos 2300, the TPU (Tensor Processing Unit) makes it stand out. Most of the Pixel exclusive features like face and photo unblur, magic eraser, speech recognition etc. depends on Tensor chips.
Tensor chips are underrated by the core selection
However, there’s one issue with Tensor chips that makes them relatable to Samsung’s Exynos chips.
Tensor G3 differentiates itself in different ways
DSP (digital signal processor) is designed to perform certain tasks like various kinds of image processing quickly and efficiently. Google calls their custom DSP as “GXP”, which plays a key role behind Pixel’s excellent camera performance even with low-end hardware.
The second generation of GXP is rumored to come out with the Tensor G3, which may further improve the smart processing capabilities. Again, GXP is exclusive to Tensor chips, making it stand out from its Exynos counterpart.
Tensor G3 features the latest version of Google’s custom TPU codenamed as “Rio” which apparently makes Tensor chips even smarter. Regardless of the same foundry and similar cores, Tensor can offer better user experience by the clever use of the hardware it has.
Tensor chips and power efficiency
Camera is one of the resource intensive tasks in a smartphone and here Pixel devices suffers due to the low efficiency of Samsung fabricated Tensor chips.
Tensor chips are fabricated using the same foundry process as Exynos, i.e., Samsung foundry. Chips fabricated in Samsung’s foundry have historically been inefficient and have had heating issues. Being fabricated in Samsung’s foundry, Tensor chips do inherit the traits of Exynos.
Having higher efficiency would enable Tensor chips to sustain heavier workloads for longer, especially in high ambient temperatures. This is where Tensor chips lose to their competition.
In such situations, the phone has to stop processing (for example, stop recording videos) in order to control its temperature.
Realistically, higher power consumption for a particular task and more heat generation are the traits that hold Tensor back from being considered excellent.
