Apple’s product released on Friday, September 20, 2024iPhone 16is an SoC developed for AI.A18” is equipped with Apple’s unique AI function “Apple IntelligenceThis is an iPhone that supports “. At the time of writing, Apple Intelligence is not yet available in Japan, but I was curious to see how much the A18 has evolved from the previous generation, so I compared the benchmark results with the iPhone 15 and iPhone 15 Pro.
iPhone 16 and iPhone 16 Plus – Apple (Japan)
https://www.apple.com/jp/iphone-16/
That’s why I have the iPhone 15, iPhone 15 Pro, and iPhone 16.Geekbench 6“and”Geekbench AIWe decided to use it to compare the performance of the CPU, GPU, and AI accelerator of the installed SoC.
The specifications of the SoC installed in the iPhone 15, iPhone 15 Pro, and iPhone 16 are as follows. The benchmark test was conducted three times at room temperature of 25°C, and the highest score was the result. The table summarizing the benchmark results of the three models ishereClick to jump and check it out.
iPhone 15 | iPhone 15 Pro | iPhone 16 | ||
On-board SoC | A16 Bionic | A17 Pro | A18 | |
architecture | 64bit ARMv8.6-A | 64bit ARMv8.6-A | 64bit ARMv9.2-A | |
Manufacturing Process | TSMC 4nm (N4P) | TSMC 3nm (N3) | TSMC 3nm (N3E) | |
CPU |
core | 6 cores: 2 high-performance cores + 4 high-efficiency cores | 6 cores: 2 high-performance cores + 4 high-efficiency cores | 6 cores: 2 high-performance cores + 4 high-efficiency cores |
L1 Instruction Cache | 128KB | 128KB | 128KB | |
L1 Data Cache | 64KB | 64KB | 64KB | |
L2 Cache | 16MB (high performance core) 4MB (high efficiency core) |
16MB (high performance core) 4MB (high efficiency core) |
16MB (high performance core) 4MB (high efficiency core) |
|
GPU | Apple A16 GPU 5-core | Apple A17 Pro GPU 6 cores | Apple A18 GPU 5-core | |
AI Accelerator | Apple Neural Engine 16 cores | Apple Neural Engine 16 cores | Apple Neural Engine 16 cores | |
RAM | 6GB | 8GB | 8GB |
◆iPhone 16-A18
First, the CPU benchmark results. The A18’s CPU single-core performance was 3213, and the CPU multi-core performance was 7786.
The single-core performance breakdown is as follows:
The multi-core performance breakdown is as follows:
The GPU’s Metal benchmark score was 27974.
Here’s a breakdown of Metal performance:
This is the score by Geekbench AI. “Single Precision Score” isSingle-precision floating pointThe calculation score is 4238, and the Half Precision Score isHalf-precision floating pointThe calculation score is 32251, and the Quantized Score isQuantizationThe calculation score was 44326.
The accuracy and score of the calculations performed are as follows:
◆iPhone 15-A16 Bionic
The A16 Bionic CPU score for the iPhone 15 was 2601 for single-core performance and 6452 for multi-core performance. The results this time are for “Your Device,” and the “Average” displayed below is the average score for that SoC in Geekbench 6.
The single-core performance breakdown is as follows:
The multi-core performance breakdown is as follows:
The GPU’s Metal benchmark score is 22968.
Here’s a breakdown of Metal performance:
The performance scores from Geekbench AI were Single Precision Score of 2231, Half Precision Score of 12798, and Quantized Score of 14796.
The accuracy and score of the calculations performed are as follows:
◆iPhone 15 Pro – A17 Pro
The A17 Pro equipped with the iPhone 15 Pro had a single-core performance result of 2841 and a multi-core performance result of 7012.
The single-core performance breakdown is as follows:
The multi-core performance breakdown is as follows:
The GPU’s Metal benchmark score is 28412.
Here’s a breakdown of Metal performance:
The performance scores from Geekbench AI were Single Precision Score of 3947, Half Precision Score of 24439, and Quantized Score of 33716.
The accuracy and score of the calculation process are as follows.
Summary of results
The results are summarized in the table below. Looking at the scores alone, it can be said that the A18’s CPU performance and AI processing performance are definitely improved compared to the A16 Bionic, let alone the A17 Pro. In particular, the A18 recorded 1.8 to 3 times the AI processing performance of the A16 Bionic, which is understandable why Apple was promoting it as “developed for AI functions.” GPU performance has improved compared to the A16 Bionic, but there was almost no difference compared to the A17 Pro.
iPhone 15 | iPhone 15 Pro | iPhone 16 | ||||||
SoC | A16 Bionic | A17 Pro | A18 | |||||
CPU | Single Core | 2601 | 2841 | 3213 | ||||
Multicore | 6452 | 7012 | 7786 | |||||
GPU | 22968 | 28412 | 27974 | |||||
Neural Engine | Single Precision Score | 2311 | 3947 | 4238 | ||||
Half Precision Score | 12798 | 24439 | 32251 | |||||
Quantized Score | 14796 | 33716 | 44326 |
Next time, I’ll be reviewing the new features of the iPhone 16 and the parts that caught my attention when using it, so stay tuned.
·to be continued
Copy the title and URL of this post