Every year, Google ships Pixel phones with its in-house Tensor chip and every year, the same conversation follows. Is Tensor finally ready to compete with Qualcomm and Apple? With the Pixel 10 Pro and its Tensor G5 chip manufactured by TSMC for the first time, there was genuine reason to believe this generation might be different. The switch from Samsung Foundry to TSMC's 3nm process was a meaningful change, and Google claimed its biggest performance leap yet.
The benchmarks are now in from multiple independent sources. Here's exactly what they show the improvements, the gaps that remain, and what all of it means in actual use.
What the Tensor G5 Actually Is Architecture and Key Changes
Before getting into the numbers, understanding what Google built matters. The Tensor G5 is an octa-core chip with the following CPU configuration:
1x ARM Cortex-X4 performance core clocked at 3.78GHz
5x ARM Cortex-A725 efficiency-performance cores at 3.05GHz
2x ARM Cortex-A520 efficiency cores at 2.25GHz
The GPU is a PowerVR DXT-48-1536 a D-Series chip from Imagination Technologies, replacing the Samsung Xclipse GPU used in previous Tensor generations. This is the first time Google has used a PowerVR GPU in a Tensor chip.
The Tensor G5 is manufactured by TSMC using 3nm process technology. Google explains that this allows it to pack more transistors into the chip for improved power and efficiency. The company claims the Tensor G5's CPU is 34% faster on average than the Tensor G4.
The chip also features a new image signal processor for improved photo and video quality, a new display controller, and is capable of running Google's latest Gemini Nano model on-device reportedly 2.6 times faster and twice as efficient as the previous generation.
Geekbench 6: CPU Performance Numbers
Geekbench 6 is the most widely used CPU benchmark for mobile devices, testing both single-core and multi-core performance.
The Tensor G5 on the Pixel 10 Pro scored 2,285 in single-core and 6,191 in multi-core on Geekbench 6.
The Google Pixel 9 Pro, running Tensor G4, scored 1,876 in single-core and 4,337 in multi-core. That means the Tensor G5 delivers approximately 22% higher single-core performance and around 43% higher multi-core performance over its predecessor a meaningful generational step.
However, the gap versus Qualcomm's flagship remains substantial. The Samsung Galaxy S25 Ultra with Snapdragon 8 Elite scored 2,867 in single-core and 9,491 in multi-core — putting the Tensor G5 roughly 20% behind in single-core and significantly further behind in multi-core performance.
Single-core performance shows a 30% increase from the Pixel 9 to Pixel 10, though the XL models show a more modest 18% improvement between generations.
AnTuTu: Overall Performance Score
AnTuTu measures overall device performance across CPU, GPU, memory, and UX workloads.
Running AnTuTu on the Pixel 10 Pro, the Tensor G5 achieved an overall score of 1,291,252 points with the CPU scoring 457,073 and the GPU scoring 382,578. This represents approximately a 20% improvement over the Tensor G4's AnTuTu result.
For context, flagship Snapdragon 8 Elite devices and Apple A18 Pro-powered iPhones have crossed the 2-million mark on AnTuTu. The Tensor G5 is not close to those figures, falling well short of the 1.5 million threshold despite the improvement over its predecessor.
Early launch-day AnTuTu results were actually lower before software updates. After installing available OTA updates, a Pixel 10 Pro XL unit delivered an updated AnTuTu total of 1,335,326, with a CPU score of 464,475 and a GPU score of 417,648. The CPU score is comparable to a Snapdragon 8s Gen 4 device, while the GPU is comparable to a Snapdragon 8s Gen 3-class chip.
GPU Performance: The Weakest Link
The GPU is where the Tensor G5 draws the most criticism from independent benchmarking.
The Tensor G5 uses a PowerVR D-Series DXT-48-1536 GPU. In 3DMark's Wild Life Extreme test, GPU performance shows approximately a 27% improvement over the Tensor G4. That sounds reasonable until you compare it to the competition.
Compared to flagship chipsets like the Snapdragon 8 Elite and Apple A18 Pro, the Tensor G5's GPU performance looks mediocre. While other chipsets have crossed the 2-million AnTuTu mark, the Tensor G5 struggles to reach 1.5 million total with GPU being the primary drag on the overall score.
In real-world gaming, demanding titles like Genshin Impact cannot maintain a steady 60fps at the highest settings on the Pixel 10 series. This is a practical limitation that matters for users who use their phone for graphically intensive games.
There is a silver lining, though. Tensor G5 shows strong benchmark stability the difference between its highest and lowest scores in sustained testing is relatively small compared to rivals, suggesting the chip maintains its performance under sustained load more consistently than some competing chips.
Thermal Performance: A Genuine Improvement
One area where the Tensor G5 marks a clear improvement over its predecessors is heat management. Previous Tensor chips, particularly the Tensor G3, were known for running hot during sustained tasks. The switch to TSMC's 3nm process has addressed this.
Early real-world testing indicates the Tensor G5 runs cooler than its predecessors. While GPU benchmarks show temperature increases during intense workloads, day-to-day thermal management is improved compared to earlier Tensor generations.
Software updates, particularly the November 2025 update, delivered measurable performance improvements roughly 19-20% CPU improvement and 5-7% GPU improvement over launch-day figures. The December update, which delivered Android 16 QPR2, showed negligible additional differences beyond what the November update had already addressed.
Real-World Performance: What Benchmarks Don't Tell You
Raw benchmark scores are only part of the picture. Day-to-day usability often tells a different story.
In real-world use, there is no major difference in day-to-day tasks or casual gaming on the Pixel 10 compared to Snapdragon-powered rivals. General app performance, multitasking, and scrolling feel smooth and responsive.
Google's focus with Tensor has never been purely about topping performance charts. The chip is designed to run on-device AI features efficiently including Gemini Nano, which on the Tensor G5 runs 2.6 times faster than on the Tensor G4. For Pixel-specific AI features like Live Translate, Call Screen, and Recorder transcription, the Tensor G5 is fast and capable.
Where the performance gap becomes tangible is in sustained GPU workloads extended gaming sessions and graphically demanding applications. This is where Snapdragon 8 Elite devices pull decisively ahead and where the Tensor G5's PowerVR GPU shows its limitations most clearly.
Tensor G5 vs Tensor G4 Side by Side
Here's a clear comparison of confirmed benchmark figures:
Geekbench 6 Single-Core: Tensor G4 (1,876) → Tensor G5 (2,285) — 22% improvement
Geekbench 6 Multi-Core: Tensor G4 (4,337) → Tensor G5 (6,191) — 43% improvement
AnTuTu Overall: Tensor G4 (~1.07M) → Tensor G5 (~1.29M) 20% improvement
GPU (3DMark Wild Life Extreme): Tensor G5 approximately 27% ahead of Tensor G4
Tensor G5 vs Snapdragon 8 Elite The Flagship Gap
Geekbench 6 Single-Core: Tensor G5 (2,285) vs Snapdragon 8 Elite (2,867) — 20% behind
Geekbench 6 Multi-Core: Tensor G5 (6,191) vs Snapdragon 8 Elite (9,491) — 35% behind
AnTuTu Overall: Tensor G5 (~1.29M) vs Snapdragon 8 Elite (~2.1M+) — significantly behind
GPU: Tensor G5 is well behind the Snapdragon 8 Elite's Adreno GPU across all graphics benchmarks
Who Should Care About These Numbers
The honest answer is that for most Pixel 10 Pro users, the benchmark gap versus Snapdragon doesn't translate into a worse daily experience. Browsing, social media, photography, calls, and general productivity run smoothly on the Tensor G5. Where you'll feel the difference is in:
Extended gaming sessions where Snapdragon devices maintain higher frame rates and lower temperatures. Video editing with professional-grade apps that push GPU processing. Running multiple AI tasks simultaneously, where Snapdragon 8 Elite's larger NPU has advantages in some workloads.
For the core Pixel audience users who prioritize camera quality, guaranteed software updates, clean Android experience, and on-device AI features the Tensor G5's benchmark shortfall versus Snapdragon doesn't undermine the phone's value proposition. For pure performance maximalists, Snapdragon 8 Elite devices remain the benchmark leaders by a clear margin.
You must be logged in to post a comment.