Google Tensor Vs Snapdragon 888; While Google has not yet released its Pixel 6 series handsets, there are many questions about how the dedicated tensor processor of these handsets works. Can this processor compete with Apple? Will the tensor really use the latest available technologies?
Google could have used Qualcomm’s Snapdragon chips or Samsung’s Exynos chips, but the company decided to work with Samsung to develop a new processor that also uses its own proprietary machine learning system. According to new reports, the Google Tensor chip will be very different from other chips on the market. While the benchmark results of this new chip have not been released yet, a lot of information has been leaked from it. In this article, we will compare the processors of Google Tensor Vs Snapdragon 888 and Exynos 2100.
Google Tensor vs Snapdragon 888 vs Exynos 2100
With little time left until the release of Qualcomm and Samsung’s next-generation chips, Google Tensor aims to compete with the current-generation Snapdragon 888 and Exynos 2100 chips. As already known, this processor uses the latest technology of Samsung Exynos chips. According to recent reports, the tensor modem and graphics chip are exactly the same as the Exynos 2100.
If this report is true and the Tensor and Exynos 2100 graphics chips are the same, the Pixel 6 will show good gaming performance. However, the chip will be slightly weaker when playing games than the Snapdragon 888. The tensor processing unit (TPU) of this processor is also expected to offer better machine learning and artificial intelligence capabilities than its competitors.
Using a combination of 4 + 2 + 2 cores in this processor has been a strange decision by Google. The use of the powerful Cortex-A1 dual-core cores makes Google Tensor superior in performing single-core tasks, but the older Cortex-A76 cores of this chip will perform worse than competing processors in performing multi-core tasks. The combination is reminiscent of Samsung’s broken Mongoose cores, and their power consumption has yet to be determined.
Designed by Google Tensor
One of the questions that many people have in mind is why Google is using the Cortex-A76 cores, which were produced in 2018, in its new processor? The answer to this question lies in the amount of energy consumed and the heat produced by these nuclei. Experiments show that the Cortex-A76 core is smaller and requires less energy than the Cortex-A77 and Cortex-A78.
The Cortex-A77 is about 17 percent larger than the A76, and the A78 is about 5 percent smaller than the A77. Also, the difference between the energy consumption of A78 and A77 was only 4%, and A76 needed less energy. However, the performance of the Cortex-A76 was about 20% weaker than that of the A77, and the performance of the A78 was about 27% better. Hence, multitasking tasks will be performed at a slower speed on the Pixel 6 than on the Snapdragon 888.
In general, it can be said that Google had no choice but to use smaller cores using the two Cortex-X1 cores. But why did the company use dual-core X1s instead of single-core, while Qualcomm and Samsung chips with single-core X1 offer great performance?
In addition to better single-core performance compared to the A78, the Cortex-X1 kernel offers developers many machine learning capabilities. Using machine learning has been a big part of Google’s strategy in designing this processor. The capabilities of the Cortex-X1 are more flexible and the bandwidth of the SIMD instructions is twice that of the A78. In fact, Google prefers to offer better single-core tasks along with more machine learning capabilities, rather than the ability to perform multitasking tasks better.
With all these interpretations, it is still unclear how Google will deal with the problem of heat generated by the tensor processor. Experiments have shown that the Cortex-X1 kernel consumes a lot of energy, and many companies try to prevent the kernel from performing many tasks. How Google wants to use the dual-core X1 while keeping the Pixel 6 cool enough at the same time is a question that has been on many people’s minds. Note that the Samsung Mangos kernel also failed due to similar problems.
Google Tensor Processing Unit
One of the many unspecified issues with the tensor processor is its tensor processing unit. We now know that this unit is responsible for performing Google’s machine learning tasks, such as voice recognition, image processing, and video decoding. Qualcomm and Samsung processors also use their own machine learning technology, but the interesting thing about the Snapdragon 888 is the fragmentation of its machine learning units.
Qualcomm’s AI engine is divided into Snapdragon 888 CPU, GPU, ISP, DSP and Sensing Hub. Although this has reduced the consumption of this processor, but it has also greatly reduced its machine learning performance. As a result, Qualcomm’s AI system in this processor has not been of much use.
On the other hand, the Google Pixel 6 processor tensor processing unit is responsible for performing most of its machine learning tasks, and if Google can use all the power of this unit, other companies will probably be forced to use more advanced machine learning systems.
Google has been able to introduce a new blood flow to mobile processors with the introduction of the tensor processor. Although it is not yet possible to comment on the performance of this processor, but according to the published information, this processor can be as attractive and powerful as the Snapdragon 888 and Exynos 2100.
As mentioned earlier, Google does not offer tensor processors with the goal of advancing the mobile chip market, but at the same time, the company has taken a new approach to solving mobile processing problems. Using a powerful dual-core processor and a tensor processing unit, the processor has taken a different approach than its competitors in performing machine learning tasks. However, the 5-year soft support of the Pixel 6 could be the main trump card in the competition with other companies’ flagships.
What do you think of Google’s tensor processor? Was the use of dual-core cores in this chip the right decision by Google? Share your comments with us.
Frequently Asked Questions:
- When will the Google Pixel 6 launch?
According to previous rumors, Google will probably launch the Pixel 6 series in October of this year (October 2021).
- Will the Pixel 6 series support 5G technology?
According to the FCC list, the Pixel 6 series supports 5G connections.
Source: Android Authority