Google’s upcoming Tensor G5 and G6 chips, rumored to debut with the Pixel 10 and 11 respectively, have sparked interest following leaks that reveal key specifications and design shifts in Google’s strategy for its custom processors. According to information shared by Android Authority, Google will abandon Samsung Foundry for TSMC’s advanced 3nm process starting with the Tensor G5, codenamed “Laguna,” which is expected to improve both efficiency and performance. This transition from Samsung’s fabrication process to TSMC is seen as a response to the underwhelming thermal, battery, and connectivity performance of previous Tensor generations. TSMC, known for powering Apple and AMD processors, leads the industry in efficient and powerful nodes, a change that could help Google’s chipsets compete more directly with Apple and Qualcomm.
The Tensor G5’s CPU cluster will also see a new configuration, with one Arm Cortex-X4 as the primary core, five Arm Cortex-A725 performance cores, and two Arm Cortex-A520 efficiency cores. This is a shift from the Tensor G4’s 1+3+4 layout. The decision to retain the Cortex-X4 core rather than adopt the newer Cortex-X925 surprised many, as the X925 promises better performance gains. However, the increase to five Cortex-A725 cores in the mid-cluster should still deliver a notable multi-core performance boost over previous Tensor versions. By comparison, Qualcomm’s Snapdragon 8 Elite, the upcoming A-series chips from Apple, and other competing chipmakers have pushed more aggressively toward newer cores. The Tensor G5’s shift in core configuration may reflect Google’s focus on optimizing multi-core tasks while balancing power consumption.
A significant update for the Tensor G5 is the inclusion of a GPU from Imagination Technologies (IMG) rather than ARM’s Mali GPU, which Google has relied on until now. The new IMG DXT-48-1536 GPU introduces ray tracing and GPU virtualization, marking Google’s first foray into advanced graphics capabilities like those traditionally favored by gamers. Ray tracing will enhance the realism of shadows, reflections, and lighting, giving a notable boost to gaming and graphics-intensive applications. GPU virtualization, on the other hand, allows for accelerated graphics processing in virtual machines, a feature aligned with Google’s focus on virtualized and containerized environments. Clocked at 1.1 GHz with only two cores, the new GPU differs from the seven-core Mali-G715 GPU in the Tensor G4, emphasizing quality over quantity in terms of core design. While the IMG DXT has fewer cores, its architecture suggests a focus on specific tasks, possibly signaling Google’s interest in expanding the capabilities of its devices in gaming and cloud-based applications.
In addition to CPU and GPU upgrades, the Tensor G5’s TPU (Tensor Processing Unit) will reportedly see a 40% increase in TOPS (Trillions of Operations Per Second) compared to the Tensor G4, though Google’s internal benchmarks suggest only a 14% real-world performance gain. The TPU’s design now also incorporates embedded RISC-V cores, which enable new on-device training operations, offering developers opportunities to optimize machine learning tasks directly on the device. This could lead to more advanced on-device AI and machine learning features, such as improved natural language processing and image recognition without reliance on cloud-based processing. Google’s emphasis on expanding the TPU’s capabilities aligns with its strategy to develop custom silicon that can keep up with growing AI demands, even if Tensor chips generally lag behind Snapdragon and Apple A-series processors in raw performance metrics.Looking ahead, the Tensor G6, codenamed “Malibu” and expected to power the Pixel 11, will reportedly adopt TSMC’s N3P process node, marking a further evolution in Google’s chip design. The G6’s CPU configuration could shift again to a 1+6 configuration, possibly indicating the removal of traditional efficiency cores and reliance on power-efficient performance cores, similar to recent trends seen in Qualcomm’s high-end designs. Although the Cortex-X930 core has yet to be officially announced, it is rumored to be featured in the G6 as a flagship core, alongside the unreleased Cortex-A730 performance cores. These changes reflect Google’s intent to refine its design approach incrementally with each new Tensor generation, potentially narrowing the performance gap with leading competitors by focusing on efficient, high-performance cores.
The Pixel 10 and 11, expected to debut with the Tensor G5 and G6 respectively, could redefine user experience through Google’s deep integration of AI and machine learning features within its custom silicon. The shift from Samsung to TSMC for manufacturing Tensor chips, the introduction of more robust GPUs, and improvements in TPU performance all indicate a focused effort to make Pixels more competitive with flagship devices from Apple and other leading brands. However, Google’s choice to reuse the Cortex-X4 core in the G5 rather than adopt newer technology raises questions about the overall performance gains, leaving some wondering if Google’s hardware changes will be enough to keep up with ever-increasing user expectations. Still, with on-device AI enhancements and the shift towards advanced, energy-efficient fabrication nodes, the Pixel 10 and 11 will likely showcase noticeable improvements in battery life, graphics, and processing efficiency.
Ultimately, the leaked details suggest that Google’s ongoing investment in custom chip development may lead to a better balance of hardware and software integration across future Pixel devices. Although the G5’s incremental changes may not bridge the full performance gap with Qualcomm and Apple, Google’s focus on AI, machine learning, and graphics capabilities could appeal to a growing audience of users who prioritize innovative, feature-rich devices over raw benchmark scores. The addition of ray tracing, GPU virtualization, and embedded RISC-V cores in the Tensor G5 also suggests that Google is carefully positioning its hardware to support new use cases, from gaming and virtualization to on-device training for machine learning models. Whether these innovations will be sufficient to shift consumer perception of Pixel devices remains to be seen, but the improvements in the Tensor G5 and anticipated enhancements in the G6 position Google as a significant player in the custom silicon space, aiming for seamless integration across hardware and AI-driven features in future Pixels.
