On October 4, Google officially released its latest generation of flagship smartphones Pixel 8 and Pixel 8 Pro, both equipped with Google's latest Tensor G3 processor, and will be equipped with the latest Android 14 system, priced from $699. Now some users have preemptively got the new Pixel 8 series of machines, and exposed the core parameters and running score test results of the self-developed chip Tensor G3 equipped with it.
9-core CPU: performance increased by more than 20% compared to the previous generation
According to the information of the Tensor G3 processor on Google Pixel 8 Pro exposed by the Geekbench database, it is based on a 9-core CPU architecture, including 1 Cortex-X3 super large core with a main frequency of 3.00GHz; 4 Cortex-A715 large cores with a main frequency of 2.45GHz; and 4 Cortex-A510 small cores with a main frequency of 2.15GHz. Although these CPU IPs were launched by Arm in 2022, they have been upgraded relatively large compared to the previous generation of Tensor G2.
From the Geekbench 6 running score, Tensor G3's single-core CPU score is 1760 points, multi-core CPU score is 4442 points, compared with Tensor G2 (Pixel 7 Pro, single-core 1463 points, multi-core 3498 points) single-core performance is about 20%, multi-core performance is increased by about 27%, but still far behind Apple A17 Pro (single-core 2914, multi-core 7119 points) and Qualcomm Snapdragon 8 Gen 2 and other competitors.
10-core Mali-G715 GPU: Low performance stability
The Tensor G3's GPU uses a 10-core Arm Mali-G715 GPU, which can support hardware-level ray tracing acceleration like the Apple A17 Pro. In addition, the previous news showed that in terms of encoding and decoding capabilities, Tensor G3 will use the "BigWave" module to further upgrade the video codec capabilities, while retaining the AV1 decoding supported by Tensor G2, it has added up to 4K@30FPS AV1 encoding. Tensor G3 supports up to 8K@30FPS encoding.
Recently, netizens @Tech_Reve exposed the results of Pixel 8 and Pixel 8 Pro running 3D Mark "Wild Life Stress test".
A regular 3D graphics scene was run according to the "Wild Life Stress test" and looped for 20 minutes, testing the continuous performance of the GPU and its stability. Factors at play here include process nodes, GPUs, CPUs, and cooling systems of devices. With an optimal cycle score of 8,216 and a minimum cycle score of 4,316, the Pixel 8 has extremely low stability of 52.5%; The Pixel 8 Pro has an optimal cycle score of 8,572 points, a minimum cycle score of 5,029 points, and slightly better stability at 58.7%.
It should be pointed out that although the Pixel 8 and Pixel 8 Pro are equipped with Tensor G3, the regular Pixel 8 is not equipped with a vapor chamber. From the Pixel 8 real machine disassembly, it can be seen that Google uses copper and graphite films and thermal grease to help transfer heat, but has little effect in alleviating the overheating problem. In contrast, the Pixel 8 Pro is equipped with a vapor chamber, which also makes its overall performance 11% higher than that of the Pixel 8, although both phones claim to use the same Tensor G3 processor.
△ Google Pixel 8 real machine disassembly
This also explains why the Pixel 8 Pro's GPU performs better. However, the continuous performance of the GPU of the Pixel 8 Pro is still not comparable to the Apple A17 Pro or Snapdragon 8 Gen 2, and even lower than the Tensor G2 equipped with the previous generation of Pixel 7 Pro.
The A17 Pro on the iPhone 15 Pro Max has a stability of 78.9%-89% in the "Wild Life Stress test stability" test, while the Snapdragon 8 Gen 2 equipped with the Xiaomi Mi 13 Ultra has a stability of 89%-100%, and the Tensor G2 equipped with the Pixel 7 Pro has a stability of 67.9%-76%.
In the more demanding "Wild Life Extreme Stress test" test, the stability of the A17 Pro on the iPhone 15 Pro Max dropped to 65.4%-71%, while the stability of the Snapdragon 8 Gen2 on the Galaxy S23 Ultra dropped to 58.7%-64%, and the stability of the Tensor G2 on the Pixel 7 Pro increased to 75.6%-82%.
Samsung 4LPP process
Although Samsung announced at the end of June last year that it had mass-produced the 3nm process, no mobile phone chip manufacturer has adopted it. According to the above related tests, Google Tensor G3 is based on Samsung 4LPP (4nm, Low Power Plus) technology node for production, not Samsung 4LPP+ process.
Previous news shows that Samsung 4LPP+ may be first released for Samsung's own Exynos 2400 chip, and next year's Tensor G4 may be based on Samsung 4LPP+ for production. Compared to Samsung's 4LPP+ process, the Samsung 4LPP process used by Tensor G3 is less energy efficient.
Although Google has always claimed that the Tensor G series processor is self-developed, in fact, this series of SoCs is helped by Samsung LSI to help Google deeply customize, Google provides its self-developed tensor processing unit (TPU) IP, and purchased Arm's CPU and GPU IP, (for the previous generation of Tensor G processors) Samsung provides multi-function codec IP, custom hybrid ISP, 5G baseband, etc. This is the main reason why Google's Tensor G series processors have been handed over to Samsung, although Samsung's process has hindered the performance of Tensor G series processors to some extent. In addition, the price of Samsung foundry is also more affordable than TSMC is also a factor.
Support local running of large AI models
Back in July 2018, Google officially launched its edge TPU for edge computing as a complement to its Cloud TPU, which was only used for inference and was designed to run TensorFlow Lite ML models at the edge.
At present, in the middle and high-end smartphone processors, there are basically integrated dedicated AI cores to use various artificial intelligence calculations. Google's self-developed Tenso processor also integrates Google's self-developed TPU core, but in terms of drivers, Google calls it "edge TPU".
The core purpose of Google's self-developed Tensor G series processor has always been to better optimize and coordinate its core software capabilities (including Android system capabilities and AI capabilities) with hardware, especially under the boom of generative AI, Tensor G3 has become a key fulcrum for Google to introduce generative AI capabilities into mobile phones.
At the Google hardware product conference, Google announced that the Pixel 8 series will get a number of onboard generative AI functions. They will include an improved version of the Magic Eraser photo editing software that comes with the system. The Pixel 8 Pro will also get additional features, including Video Boost, which will help improve the quality of your phone's video through generative AI processing.
For example, Magic Eraser photo editing software allows you to adjust the brightness of the photo, the background and even drag the magic retouching function to change the position of the main person.
△Magic retouching tool effect display
Or a photo function that can help users change expressions with one click. Pixel series mobile phone users should be no stranger to Top Shot's best photo filtering ability, perfect photo on the basis of the best photo function, for faces, group portraits and other shooting scenes, with the power of Google AI to upgrade again, allowing users to directly capture different picture frames captured by the camera, choose the most perfect expression for different characters to replace.
△Perfect fit function effect display
For Google Pixel series models, the most intuitive improvement brought by Google Tensor chips may be video shooting. In the Pixel 8 Pro, Google AI continues its efforts, bringing Video Boost that can process 4K video in real time through the HDR+ pipeline, calculated audio that can recognize different types of sound to facilitate post-targeted editing, and video shooting support in night vision mode.
△iPhone 15 Pro Max and Pixel 8 Pro after turning on the movie enhancer shooting effect comparison
In addition, the Pixel 8 Pro is also the first smartphone equipped with Google's AI base model. Thanks to the AI-oriented Tensor core in the Tensor G3 processor, users can process AI tasks that would otherwise need to be completed through cloud servers locally on their smartphones, with excellent effects and response speeds. For example, Pixel 8 Pro can support offline running magic eraser, which can be more based on image generation than blended into the background to achieve better object removal effects; Gboard can automatically generate more natural and more consistent reply suggestions based on conversation information, allowing users to socialize with one click; The Pixel 8 Pro even has a built-in model specifically for image processing, generating sharper detail for images enlarged in the gallery.
It is worth mentioning that MediaTek has also announced that its new generation of Dimensity flagship mobile chip smartphones support AI applications developed by the Llama 2 model, which can bring users an exciting generative AI application experience.
Brief summary:
Although Google Pixel series smartphones have always been low in existence, compared to other brands' flagship smartphones, Pixel series smartphones are mainly by taking the lead in upgrading the latest version of Android as the biggest selling point. However, with the addition and continuous iteration of the Tensor G series self-developed mobile phone processor, it has also brought more software and hardware collaboration experience upgrades to the Pixel series of smartphones. For example, Google has introduced its latest AI research results into the new Pixel 8 and Pixel 8 Pro phones, while the Tensor G3 deeply combines Google's machine learning and AI algorithms to greatly improve the ability to process photo and video recordings. Of course, there is still a certain gap between the performance of the CPU and GPU of the Tensor G series processor and the flagship SoC, which is mainly due to the constraints of chip design and Samsung process technology.
However, there are rumors that the Tensor G5 equipped with the Pixel 10 series to be launched in 2025 may be completely customized by Google and may be handed over to TSMC. At that time, Tensor G5 may be able to bring performance experience comparable to Qualcomm or MediaTek's flagship mobile processors.
Editor: Xinzhixun - Ronin Sword