A couple of weeks ago we talked about google tensor G2a new generation SoC that the Mountain View giant is going to use in the Pixel 7 and Pixel 7 Pro, and we emphasized a very important question, raw performance or specialization, muscle or brain?
The first information we had so far suggested that the Mountain View giant was going to repeat the Google Tensor G1 approach, that is, that it was going to bet once again on prioritize artificial intelligence and deep learning, and for the functions and features that both make possible, putting raw performance in the background. Brain first, and then muscle.
Thanks to a new leak we have been able to see that this information was not misguided at all, and that is that the performance of the Google Tensor G2 in Geekbench 5 it is inferior to that of the snapdragon 888, a chip that was top of the range a year ago that is behind the current Snapdragon 8 Gen1. It goes without saying that if these performance data are confirmed in the final version of the Google Tensor G2, this SoC would fit rather within the current mid-range, and would not reach the high-end.
The Google Tensor G2 SoC in the Google Pixel 7 Pro gets a Geekbench 5 score of 1,068 in single thread and 3,149 in multithread. To compare those results we have the performance data of the Snapdragon 888 using in the Lenovo Legion 2 Pro, which achieved 1,115 points in single thread and 3,581 points in multithread. The difference is not very big, but it is more than enough to put the Google Tensor G2 one step behind the Snapdragon 888.
But this is not all, thanks to the data collected by Geekbench 5 we know that the Google Tensor G2 will have an 8-core CPU divided into three blocks, one with two Cortex-X1 cores high performance at 2.85 GHz, another with two Cortex-A76 cores at 2.35 GHz and another with quad core Cortex-A55 low consumption at 1.8 GHz. The GPU has not transpired, but we can expect an ARM Mali.
Seeing the architectures that Google has used, it is clear where the problem is, and why the Google Tensor G2 is not capable of being at the same level as the Snapdragon 8 Gen1. The Qualcomm chip uses a much higher configuration at the CPU level, since it has a Cortex-X2 core at 3GHz, three core Cortex-A710 at 2.5GHz and quad core Cortex-A510 at 1.8GHz
If we put this in perspective and look at the medium term, it is evident that when the launch of the Snapdragon 8 Gen2 takes place, scheduled for the end of the year, the Google Tensor G2 it will position itself even worse and will be framed directly in the mid-range. In any case, the Mountain View giant has already made it clear with the Pixel 6 and Pixel 6 Pro that it can create value beyond raw power, so let’s not be pessimistic and give it a vote of confidence, in the end all this may be be compensated with important improvements at the level of computational photography and other features that depend on artificial intelligence and deep learning.