- Apple's A17 Pro decimates the Tensor G4 in the Geekbench CPU test. In multi-core tasks, A17 Pro is nearly 50% faster than Tensor G4.
- The 6-core GPU on the A17 Pro delivers about 36% better graphics performance than the 7-core Mali-G715 GPU on the Tensor G4.
- Apple's Neural Engine beats Google's TPU with an exceptional performance in the Geekbench AI test.
Google is putting a lot of effort into wooing iPhone users. The new Pixel 9 series looks very similar to recently launched iPhones. That said, Pixels and iPhones run distinct operating systems and are powered by different chipsets. So today, we compare their chipsets — Tensor G4 on the Pixel 9 Pro and A17 Pro on the iPhone 15 Pro — across several benchmarks, like AnTuTu, Geekbench, 3DMark, and more, to find out which one is more capable.
Tensor G4 vs A17 Pro: Specs Comparison
Specs | Tensor G4 | A17 Pro |
---|---|---|
CPU | Eight-core CPU (1+3+4) | Six-core CPU (2+4) |
CPU Cores | 1x 3.10GHz (Cortex-X4) 3x 2.60GHz (Cortex-A720) 4x 1.95GHz (Cortex-A520) | 2x 3.78 GHz (High-performance cores) 4x 2.11GHz (High-efficiency cores) |
Process Technology | Samsung 4nm | TSMC 3nm (N3B) |
GPU | Mali-G715 MC7 GPU Up to 940MHz | Apple Pro-class 6-core GPU Hardware-accelerated Ray Tracing |
Machine Learning and AI | Google Custom TPU (Codename Rio) | 16-core Neural Engine 35 TOPS |
ISP | Google Custom DSP (Codename Callisto) | Apple-designed Image Signal Processor |
Connectivity | Wi-Fi 7 Bluetooth 5.3 | Wi-Fi 6E Bluetooth 5.3 |
Modem | Samsung Exynos 5400 5G modem Up to 14.79 Gbps Peak Download Satellite SOS | Snapdragon X70 5G modem Up to 10 Gbps Peak Download Up to 3.5 Gbps Peak Upload |
Others | AV1 Encoder and Decoder Titan M2 security chip Ultra-wideband chip NavIC | AV1 Decoder ProRes codec Ultra-wideband chip NavIC |
Tensor G4 vs A17 Pro: Geekbench 6 CPU
In our first test that evaluates the CPU performance, the Google Tensor G4 doesn’t perform well. It scores 1,897 in the single-core test and 3,721 in the multi-core test. In comparison, the Apple A17 Pro absolutely dominates and achieves a single-core score of 2,897 (34% increase) and a multi-core score of 7,261 (a massive 49% uplift). Mind you, Apple achieves this level of performance with just six cores whereas the Tensor G4 packs eight CPU cores.
Geekbench 6 Score Tensor G4 A17 Pro Single-core 1,897 2,897 Multi-core 3,721 7,261
![tensor g4 vs a17 pro geekbench](https://beebom.com/wp-content/uploads/2024/08/tensor-g4-vs-a17-pro-geekbench.jpg?w=494&quality=75)
Tensor G4 vs A17 Pro: AnTuTu Benchmark
In the AnTuTu benchmark, the Tensor G4 scores 1,253,407 points whereas the Apple A17 Pro achieves 1,628,672 points. Hence, the overall performance difference comes down to 23%. In the GPU department, the gap becomes even wider where A17 Pro’s GPU leads by 26%. Apple’s 6-core GPU easily defeats the 7-core Mali-G715 GPU, but we have done more tests to verify this result
AnTuTu Benchmark Tensor G4 A17 Pro AnTuTu Score 1,253,407 1,628,672 CPU 337,131 384,208 GPU 444,488 603,504 Memory 243,643 300,838 UX 228,145 340,122
Tensor G4 vs A17 Pro: 3DMark Wild Life Extreme Stress Test
Next, in the intensive 3DMark Wild Life Extreme Stress test, the Tensor G4 again underperforms. The older 7-core ARM Mali-G715 GPU manages to get the best loop score of 2,590 points and the lowest loop score of 1,525 points. The stability remained at 58.9%.
The Apple A17 Pro, on the other hand, gets the best loop score of 4,110 and the lowest loop score of 2,820 with a stability of 68.6%. In this test, A17 Pro’s GPU performs 36% better than the Tensor G4 GPU.
3DMark Wild Life Extreme Stress Test | Tensor G4 | A17 Pro |
---|---|---|
Best loop score | 2,590 | 4,110 |
Lowest loop score | 1,525 | 2,820 |
Stability | 58.9% | 68.6% |
Tensor G4 vs A17 Pro: Geekbench AI Benchmark
In the Geekbench AI benchmark, the 16-core Neural Engine on the A17 Pro absolutely crushes the TPU on the Tensor G4. Whether in Single Precision, Half Precision, or a Quantized score, the A17 Pro outperforms the Tensor G4 by 6x to 10x in all these tests.
Even if we consider AI performance on the CPU and GPU, the Apple A17 Pro delivers multi-fold gains over the Tensor G4. Overall, for on-device AI tasks, the Neural Engine on A17 Pro is much more capable than Google’s mobile TPU.
Geekbench AI Tensor G4 CPU A17 Pro CPU Tensor G4 GPU A17 Pro GPU Tensor G4 TPU/NPU A17 Pro Neural Engine/NPU Single Precision Score 1,883 3,969 720 4,928 289 3,878 Half Precision Score 1,837 6,855 825 5,581 4,145 21,324 Quantized Score 2,899 5,976 772 4,384 6,103 26,213
Tensor G4 vs A17 Pro: 5G Connectivity Test
Moving to our 5G speed test, well, the A17 Pro packs Qualcomm’s Snapdragon X70 5G modem, and that results in far better radio connectivity. It achieves a peak download/upload speed of 468 Mbps and 20.1 Mbps, respectively.
The Tensor G4 now comes with the upgraded Samsung Exynos 5400 5G modem, and it manages 174.8 Mbps in peak download speed and 35.1 Mbps in peak upload speed. For mobile connectivity, both iPhones and Snapdragon-powered phones continue to offer better bandwidth than the Pixel 9 series.
Google Tensor G4 vs Apple A17 Pro: The Verdict
As far as raw performance is concerned, the Apple A17 Pro has the most powerful CPU among all the mobile chipsets in the market, not just Tensor G4. Its 6-core GPU is also considerably more capable than Tensor G4’s older Mali GPU. Not to mention, Apple’s Neural Engine is an order of magnitude faster than Google’s TPU in the Tensor G4.
Finally, in mobile connectivity too, the A17 Pro outperforms as it packs Qualcomm’s 5G modem. In all segments, the A17 Pro demonstrates superior performance than the Tensor G4. So, to conclude, there’s simply no competition between the A17 Pro and Tensor G4 on paper or in the real world. If you want a smartphone with a capable chipset, you should get an iPhone over the Pixel.