Barry Tech Review

Best review technology by Barry

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: It’s No Longer About the CPU

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: It's No Longer About the CPU


Google’s first-gen in-house silicon, Google Tensor, was introduced earlier last year with the Pixel 6 series launch. Now, after almost a year, Google has unveiled the Google Tensor G2 chipset with the Pixel 7 series. The second-gen Google silicon is supposed to bring marginal improvements in CPU performance and significant gains in the GPU department. However, how well does the Google Tensor G2 stack up against the Qualcomm Snapdragon 8+ Gen 1 and Apple’s A16 Bionic? To find the answer, go through our comparison between the Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic. We have discussed CPU, GPU, TPU (AI and ML), ISP, 5G modem, benchmark numbers, and more.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: An In-depth Comparison (2022)

In this comparison between the Google Tensor G2, Snapdragon 8+ Gen 1, and A16 Bionic, we have analyzed the CPU architecture, GPU performance, benchmark numbers, and more.

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: Specifications

Here, we have mentioned the detailed specs sheet for the Google Tensor G2, Snapdragon 8+ Gen 1, and A16 Bionic chipset. Glance over it to get a rough idea about all three processors.

  Google Tensor Snapdragon 8+ Gen 1 A16 Bionic
CPU Octa-core CPU Kryo CPU, Octa-core CPU Hexa-core CPU, 16 billion transistors
CPU Cores 2x 2.85GHz (Cortex-X1)
2x 2.35GHz (Cortex A76)
4x 1.8GHz (Cortex A55)
1x 3.2GHz (Cortex-X2)
3x 2.5GHz (Cortex A710)
4x 1.8GHz (Cortex A510)
2x High-performance cores
4x High-efficiency cores
Process Technology Samsung’s 4nm LPE process TSMC’s 4nm process TSMC’s 4nm process
GPU Mali G710 GPU Adreno 730 GPU; Snapdragon Elite Gaming Apple-designed 5-core GPU
Machine Learning and AI Google Custom TPU 7th-gen AI Engine; 3rd Gen Sensing Hub; 27TOPS New 16-core Neural Engine; 17 TOPS
ISP Google Custom ISP 18-Bit ISP; Snapdragon Sight Apple-designed New Image Signal Processor
Camera Capability Cinematic Blur
4K 60FPS on all cameras
10-bit HDR support
Active Stabilization
3.2 Gigapixels per second, 240 12MP photos in one second ProRAW photos at 48MP
Photonic Engine
Action Mode
Video Capability 4K 60FPS recording
Google HDRnet
8K HDR, 18-bit RAW, Dedicated Bokeh Engine 4K HDR Dolby Vision @ 60FPS
Cinematic 4K@24FPS
Action mode
Modem Samsung G5300B modem Qualcomm X65 5G Modem-RF, Up to 10 Gbps Peak Download Qualcomm X65 5G Discrete Modem
WiFi Support Wi-Fi 6E Wi-Fi 6E Wi-Fi 6
Bluetooth Bluetooth 5.2 Bluetooth 5.3, LE Bluetooth 5.3

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: CPU

Beginning with the CPU, the Google Tensor G2 has not seen any significant changes compared to Google’s first in-house chipset, the OG Google Tensor. The CPU architecture is the same as last year’s SoC, which includes two Cortex-X1 cores, two Cortex-A76 cores, and four Cortex-A55 cores. The only difference with Tensor G2 is that the Cortex-X1 core is clocked slightly higher at 2.85GHz instead of 2.80GHz and the Cortex-A76 core is clocked at 2.35GHz instead of 2.25GHz. Moreover, the new Tensor G2 is developed on Samsung’s 4nm LPE process node in place of last year’s 5nm fabrication process.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: CPU
Google Tensor G2

Basically, you’re getting almost an identical CPU on the Google Tensor G2, and the benchmark results (more on this below) also reflect the same trend. You don’t get the newer and improved Cortex-X2, A710, or A510 core, which is a bit disappointing.

If you compare Google Tensor G2’s CPU with Snapdragon 8+ Gen 1 and A16 Bionic, well, it doesn’t hold up at all. The SD 8+ Gen 1 packs the new Cortex-X2 core, which is clocked considerably higher at 3.2GHz. Apart from that, there are three A710 and four A510 cores, which are clocked at 2.5GHz and 1.8GHz respectively.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: CPU
A16 Bionic

Apple’s A16 Bionic, on the other hand, is altogether on another level. Instead of an octa-core architecture, it goes for a hexa-core cluster with two high-performance cores and four high-efficiency cores. Simply put, in terms of CPU performance, the Tensor G2 SoC is nowhere closer to either Qualcomm or Apple’s offering.

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: GPU

Coming to the GPU, Google has indeed packed a newer and more performant Mali G710 GPU on the Google Tensor G2. It’s said to bring a 20% performance improvement over the Mali G78 GPU used in last year’s Tensor chipset. At the same time, it consumes 20% less energy, which is great for thermal efficiency.

Hopefully, with the Tensor G2, we will see less thermal throttling, better heat dissipation, and sustained graphics performance. Besides that, the GPU also offers 35% better performance in machine learning compared to the Mali G78 GPU.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: GPU
Google Tensor G2 architecture

Note that the Mali G710 GPU can be configured starting from 7 to 16 cores. We are not sure how many GPU cores Google packed on the Tensor G2 chipset. If it packs 16 maximum cores, we might get a powerful GPU in our hands. For your information, other manufacturers like Xiaomi and Asus have gone with the traditional Mali-G710 MC10 setup that features only 10 cores.

Comparing the Mali G710 GPU on the Google Tensor G2 against Qualcomm and Apple’s offerings, well, I think in this department, Google might pull a surprise. The 10-core Mali G710 GPU is already more power-efficient than Snapdragon 8 Gen 1’s Adreno 730 GPU and comes close to A16 Bionic. In terms of performance too, the 10-core Mali G710 GPU has scored around 160FPS in the GFXBench test, whereas Snapdragon 8 Gen 1 scored 175FPS and A15 Bionic peaked at 180FPS.

Snapdragon 8+ Gen 1 gaming features
Snapdragon 8+ Gen 1 Gaming details

Adreno 730 GPU has not seen major performance improvement on the Snapdragon 8+ Gen 1 over the 8 Gen 1, except for power efficiency improvements. And the 5-core GPU on A16 Bionic is also the same as A15 Bionic, except for more memory bandwidth. In tandem, if Google chose to go with a higher 16-core GPU, it’s going to beat the GPU on the Snapdragon 8+ Gen 1 and may rival A15 Bionic’s GPU.

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: Geekbench and AnTuTu Benchmark Numbers

Talking about benchmark numbers of Google Tensor G2, Snapdragon 8+ Gen 1, and A16 Bionic, let’s begin with Geekbench single-core and multi-core CPU tests. As you can see here, the CPU on the Google Tensor G2 is marginally better than last year’s Google Tensor (2021). And when compared with Snapdragon 8+ Gen 1 and A16 Bionic, it’s miles behind. In fact, the Google Tensor G2 is close to Snapdragon 888+ in terms of CPU performance.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: Geekbench and AnTuTu Benchmark Numbers
Geekbench Test

If we go by the recently leaked AnTuTu score of Google Tensor G2, Snapdragon 8+ Gen 1, and A16 Bionic, Google again disappoints with its latest silicon. The Tensor G2 chipset scores 801,116 in the AnTuTu test and is way behind SD8+ Gen 1 and A16 Bionic.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: Geekbench and AnTuTu Benchmark Numbers
AnTuTu benchmark result for Tensor G2

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: ISP

While the Google Tensor G2 is behind the Snapdragon 8+ Gen 1 and Apple A16 Bionic in CPU performance and is promising in the GPU department, ISP is where Google shines. And that’s because it controls both the hardware and software verticals. Along with all the photo and video capabilities, the custom Google ISP offers cinematic blur for videos, 2x faster Night Sight photography, and 10-bit HDR support. You also get active stabilization using both hardware and software, along with 4K 60FPS shooting across all the cameras.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: ISP
Google Tensor G2

If we talk about the ISP on the Snapdragon 8+ Gen 1, well, it’s definitely more powerful and offers an 18-bit triple ISP architecture. It can capture 3.2 gigapixels per second and supports 8K HDR recording as well. The ISP on the A16 Bionic is also quite advanced and can perform 4 trillion operations per photo. It powers the new Photonic engine for generating sharper and rich images and offers Cinematic videos and Action mode to stabilize shaky videos.

a16 bionic isp
A16 Bionic

Overall, I would say all three ISPs are plenty powerful, but at the end of the day, it depends on the phone manufacturer on how to leverage these hardware capabilities. And it seems like Google is winning this game with new and exciting camera features that are meaningful to the user.

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: AI and ML

In the AI and ML segment, Google is a leader in providing useful features on Pixel phones. And on the Tensor G2 chipset, Google has added a new TPU (Tensor Processing Unit) that can deliver state-of-the-art AI and personal intelligence features to your phone.

Google Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: AI and ML
Google Tensor G2 AI capabilities

In the keynote address, Google said that machine learning runs up to 60% faster on the Tensor G2 chip with 20% more power efficiency. From speech recognition to translating conversations, voice assistance, Pixel Call Assist, Call Screen, Super Res Zoom, etc., you get all the AI-driven features on the Pixel lineup, thanks to its powerful TPU.

Qualcomm’s 7th-gen AI Engine on the Snapdragon 8+ Gen 1 is also powerful, and it can perform 27 trillion operations per second. That said, it depends on your phone manufacturer on how to provide smart and meaningful experiences by utilizing hardware prowess.

Snapdragon 8+ Gen 1
Snapdragon 8+ Gen 1

On the other hand, Apple’s new 16-core Neural Engine on the A16 Bionic can perform 17 trillion operations per second, which helps in computational photography, voice assistance, pixel-by-pixel analysis, speech recognition, etc. However, no other company comes close to Google in delivering smart and personalized experiences, so the Google Tensor G2 wins this round as well.

Tensor G2 vs Snapdragon 8+ Gen 1 vs A16 Bionic: Wireless Connectivity

The Google Tensor G2 chip features Samsung’s unannounced G5300B 5G modem, which supports both mmWave and sub-6GHz bands. However, there is very little information about the specifics of the modem and its peak throughput. If we go by the Pixel 7 product listing, it supports almost 22 5G bands, covering most of the NR frequency bands. Apart from that, it supports Bluetooth 5.2 and Wi-Fi 6E.

Snapdragon 8+ Gen 1
Snapdragon 8+ Gen 1 connectivity options

Moving to the Snapdragon 8+ Gen 1, it includes the in-house X65 5G modem, which offers a peak theoretical download speed of 10Gbps. In addition, the chipset brings support for Bluetooth 5.3 and LE standards. Finally, the A16 Bionic also features a discrete X65 5G modem from Qualcomm and supports Wi-Fi 6 and Bluetooth 5.3.

Overall, in terms of 5G and wireless connectivity, the Google Tensor G2 lags behind SD 8+ Gen 1 and A16 Bionic. Qualcomm is one of the leaders in the modem industry, and Samsung’s modems have not been able to catch up with the best in the industry. In fact, Samsung has decided that it’s only going to use Qualcomm’s X70 5G modem on the Galaxy S23 series, so you can catch the drift.

What’s Your Verdict on the Google Tensor G2?

So that was our in-depth comparison between the Google Tensor G2, Snapdragon 8+ Gen 1, and A16 Bionic. Except in the CPU and modem department, we believe the Google Tensor G2 is a capable chipset with notable gains in GPU, TPU (AI + ML), and ISP. Now, it’s time to test how well Google has optimized the phone for handling intensive tasks and if there are any thermal throttling or heating issues.

Anyway, that is all from us. But what do you think about the Google Tensor G2 chipset? Do let us know in the comment section below. Also, head to our comparison between the A16 Bionic and Snapdragon 8+ Gen 1 to learn more about the performance difference between the two flagship SoCs.



Leave a Reply

Your email address will not be published. Required fields are marked *