Google’s Tensor, a machine learning-centric chip, has been crushed by Apple A15 in Machine Learning


Google Tensor

Google unveiled its Pixel 6 series with some big advancements in terms of hardware, especially for the Pixel 6 Pro. However, the biggest highlight in these devices is associated with the Tensor chip. This is Google’s in-house-made chip that focuses on machine learning and AI tech. It’s behind all the fancy features available on the newer smartphones. The Tensor improves the company’s computational photography and transcription abilities. The search giant is studying on bringing some of the Pixel 6 features to older Pixel models. However, some of them will remain exclusive due to the Tensor chip ML capabilities. Despite the stars on these capabilities, the chip seems to still be behind one of its major competitors.

When it comes to brute force, Google Tensor has been outperformed by the latest Qualcomm Snapdragon 888 and Exynos 2100 smartphones. Apple’s Apple A15 and even the A14 also outperform Google’s chipset. Now, a new test performed by the folks at Notebookcheck, states that the Apple A15 Bionic chip is crushing the Google Tensor on its greatest strength – machine learning. The test was done on the new Geekbench ML app that has been developed specifically to measure machine learning performance. Despite all the highlights on machine learning, the Google Tensor is still crushed by Apple’s A15 capabilities.

Apple A15

Apple A15 crushes the Google Tensor chip in Machine Learning capabilities

The Google Tensor managed to score just 307 in the Geekbench ML TensorFlow Lite CPU test, which is under a third of the Apple A15’s impressive score of 939. The 307 figure isn’t terrible by any means but shows that it’s no match for Apple. Furthermore, the Apple A15 scores 2,727 points in the neural accelerator test while the Tensor scores 1,720 points. There is also a huge gap in the GPU tesh scoring 1,428 while the A15 has a higher score of 2,727.

Read Also:  Russia slams another fine on Google as their fracas continue

All in all, the aggregate score of the Apple A15 stands at 5,934 while Tensor gets 3,455. This means that the Apple A15 is around 71 percent faster than the Google Tensor when it comes to machine learning. Of course, this is when we take benchmark tests into consideration. The situation can be different in real-world performance. However, evaluating and comparing this is almost impossible for common users. Of course, the benchmark serves to tell that there is still a gap even if it’s shorter than we think, and the Tensor isn’t the most cutting edge chip when it comes to machine learning.

Anyway, for a first, the Tensor chip is still a solid offering from Google. This is just the beginning, and the company can tweak even more the performance with upcoming updates. Furthermore, we may see Google improving the chip further with newer generations.

Source/VIA :
Previous POCO M4 Pro 5G passes by Geekbench and the specs are disappointing
Next Apple forgot to adapt macOS to the MacBook Pro notched display