Apple A18 chip NPU capability surpasses M4! But the AI mobile phone market is surrounded by strong competitors

The latest research by JPMorgan Chase shows that iPhone sales in China continued to grow in May this year, with a 44% increase in sales compared to April this year, thanks to the recovery of the Chinese market. However, in the second half of this year, the global AI smartphone market will see a surge of powerful flagship products, and the prospects for the iPhone 16 series are not entirely clear.
Apple A18 provides strong power for the "AI version" of iOS
At this year's WWDC, Apple Senior Vice President Craig Federighi announced in a speech that Apple Intelligence will collaborate with OpenAI's ChatGPT, and shortly after the speech, hinted that Apple iOS18 plans to introduce Google's Gemini platform this fall to provide users with a richer and smarter service experience. Currently, this news has been mentioned by Apple product whistleblower Mark Gurman on the show.
In addition, Apple is also in discussions with Facebook's parent company Meta to establish a partnership and consider integrating Meta's generative artificial intelligence model into Apple Intelligence. It is also reported that artificial intelligence startups Anthropic and Perplexity are in talks with Apple. As Apple opens up its own software ecosystem, the backwardness of iOS in AI has been completely shattered, and it has vaguely risen to become the leader in AI smartphones. However, this requires stronger support from A18 processors.
Several Apple analysts have stated that the A18 processor may be the biggest update highlight of the iPhone 16 series, as the neural engine on the chip will surpass the computing power level of the Apple M4 chip. On May 7, 2024, Apple officially released the M4 chip to the public. Like the A18 processor, the M4 chip also adopts TSMC N3E technology, with a maximum of 10 cores and a unified memory bandwidth of 120GB/s. On top of the new generation graphics processor architecture based on the M3 series chips, Apple has created a brand new 10 core GPU for the M4 chip, bringing dynamic caching, hardware accelerated ray tracing, and hardware accelerated grid coloring functions to the iPad for the first time.
On the M4 chip, Apple has paired it with the most powerful neural network engine NPU to date, specifically designed to accelerate AI tasks. This NPU module has a maximum computing speed of up to 38 trillion times per second, which is an astonishing 60 times faster than the first generation neural network engine in the A11 biomimetic chip. The M4 chip equipped with a powerful NPU gives Apple confidence in creating industry-leading AI PC products.
Now, Apple plans to further enhance NPU capabilities on the A18 processor, with a high probability of exceeding 40 trillion operations per second. However, even so, the iPhone 16 series cannot be said to stand out in the AI smartphone market, as their competitors have strong competitive strength.
The competition among AI processors in the Android camp is fierce
High performance processors are the core of AI phones, and Apple understands this. The same goes for the Android camp. In order to enhance the AI capabilities of Android flagship phones, several powerful AI processors will also be introduced during the same period of the iPhone 16 series.
Firstly, of course, it is Qualcomm's Snapdragon 8 Gen4. Previously, Qualcomm announced that it would hold the Snapdragon Summit 2024 on Maui Island, Hawaii from October 21 to October 23. At that time, Qualcomm's mobile platform Snapdragon 8 Gen4 will officially debut, with the Xiaomi 15 series winning the global debut. Summarizing the leaked information, the code name for the Snapdragon 8 Gen4 processor may be "SUN" (Sun), symbolizing its powerful performance and endless energy. In terms of chip design, the Snapdragon 8 Gen4 is built on the TSMC N3E process and is equipped with a self-developed CPU large core Oryon core. The CPU architecture is based on the Nuvia Phoenix architecture, and the main frequency of the CPU will reach a terrifying 4.2GHz. In Geekbench testing, the single core benchmark score reached 3000 points. The GPU architecture of Snapdragon 8 Gen4 is reported to be the "Slice GPU architecture", integrating Adreno 830 GPU. This architecture divides the GPU into multiple independent physical units, each with its own pipeline, cache, and memory, thereby achieving more efficient task processing and resource allocation. According to sources, the biggest upgrade of Snapdragon 8 Gen4 may come from NPU. In the Snapdragon 8 Gen3, the performance of its Hexagon NPU has improved by 98% compared to the Snapdragon 8 Gen2, and it is expected that the NPU of the Snapdragon 8 Gen4 will continue to be upgraded exponentially.
Previously, the blogger "Digital Chat Station" revealed that the Qualcomm Snapdragon 8 Gen4 exceeded 10000 points in Geekbench multi-core performance, surpassing Apple's A18 processor and setting a new industry benchmark.
Of course, the performance of Qualcomm Snapdragon 8 Gen4 is strong and the price is not low. According to reports from South Korean media, due to the use of TSMC's N3E technology, coupled with a large number of self-developed architectures and cores, the Qualcomm Snapdragon 8 Gen4's quotation has skyrocketed by 25%. Therefore, Samsung plans to seek cooperation with MediaTek to customize Dimensity chips for the Galaxy S25 and S25+, only using Snapdragon 8 Gen4 on the S25 Ultra.
To compete with Snapdragon 8 Gen4, MediaTek's product is the Dimensity 9400. Not long ago, MediaTek's Dimensity 9400 processor appeared in the IMEI database. This processor platform will use the Arm public Cortex-X925 CPU and Immortalis G925 GPU. At present, MediaTek's flagship processor has adopted a full core architecture. Taking the Dimensity 9300+as an example, this processor has one Cortex-X4 super core with a main frequency of up to 3.4GHz, three Cortex-X4 super cores with a main frequency of 2.85GHz, and four Cortex-A720 super cores with a main frequency of 2.0GHz. It is expected that MediaTek will continue to adopt this architecture on the Dimensity 9400, and the ultra large core will be transformed into a Cortex-X925 CPU. The single core performance of the X925 is 36% higher than that of the X4 (based on Geekbench testing). Thanks to a private L2 cache of up to 3MB, Arm reported a 41% improvement in AI workload performance (measured as time to token) for the X925. Of course, MediaTek not only enhances the performance of CPUs and GPUs, but is also worth looking forward to on its self-developed NPU core. The MediaTek NPU 790 on the Dimensity 9300+has built-in hardware, which is MediaTek's seventh generation NPU core. It is deeply adapted to the Transformer model and provides faster and safer edge AI performance. Therefore, the performance of NPUs on the Dimensity 9400 is definitely not bad.
In addition to mainstream Snapdragon and Dimensity, the strength of self-developed processors by manufacturers is also impressive. It is reported that the Tensor G5 chip expected to be used for Google's flagship smartphone next year will be based on TSMC's 3nm process and has successfully entered the chip stage. Previous generations of Tensor processors can generally be understood as variations of Samsung processors, except for Google's own TPU (neural network engine), which was mostly modified and customized on Samsung Exynos. The Tensor G5 chip is a fully self-developed processor that utilizes fully independent chip design and TSMC's latest 3nm technology. With the support of Tensor G5 chips, Google's latest generation Pixel is also highly anticipated, especially in terms of AI capabilities. Google TPU can ensure that the latest AI technology in Android systems receives the fastest hardware support.
If the project goes smoothly, there will also be Samsung's Exynos 2500 on this competition list, which we have previously reported on in our article. The Samsung Exynos 2500 will be based on Samsung's second-generation SF3 technology, using a 10 core four cluster CPU architecture of 1+2+3+4, and will use the Arm Cortex-X925 ultra large core and Arm Cortex-A725 large core. However, the biggest uncertainty of this processor currently comes from Samsung's second-generation SF3 technology, and the current yield cannot support mass production at all.
epilogue
The Apple A18 is expected to bring more powerful NPU capabilities than the M4 chip, which is necessary to enhance the AI capabilities of the iPhone 16 series. Coupled with Apple's recent actions in AI big models, the iPhone 16 series is highly anticipated. A research report suggests that in the second half of this year, AI smartphones are expected to drive up both the quantity and price of consumer electronics components. The application of AI models in mobile phones is expected to break the lack of innovation in the terminal market, and the iPhone 16 series is also expected to increase its market share through this.
However, in the AI smartphone market, the prospects of the iPhone 16 series are not stable and face fierce market competition. Qualcomm Snapdragon 8 Gen4, MediaTek Dimensity 9400, as well as self-developed chips from Google and Samsung, are expected to significantly enhance their AI processing capabilities, which will have a certain impact on the iPhone 16 series equipped with A18 processors. Of course, consumers are happy to see such a competitive situation.