Apple Has Unveiled A Powerful AI Chip M4 Capable of Performing Up To 38 Trillion Operations Per Second

On May 7, local time, Apple launched the latest generation of self-developed computer chip M4, which is equipped with the new iPad Pro. The M4 is said to have Apple's fastest neural engine ever, capable of performing up to 38 trillion operations per second, which is faster than the neural processing unit of any AI PC today.

 

 

M4 chip -28 billion transistors +3nm+LPDDR5X+ up to 10 core cpus

 

Built from the inside, the M4 is made up of 28 billion transistors, only slightly more than the M3; From the point of view of the process node, the chip is built with the second-generation 3-nanometer technology and is a system-on-chip (SoC), which further improves the efficiency of Apple's chips. It is reported that M4's use of the second generation of 3nm is fully consistent with the N3E process previously launched by TSMC. According to TSMC data, N3E is not as dense as N3B, but it can provide slightly better performance and power characteristics.

 

 

From the core architecture point of view, the new CPU of the M4 chip has up to 10 cores, including 4 performance cores and 6 efficiency cores, which is 2 more efficiency cores than the M3. The new 10-core GPU builds on the next-generation GPU architecture introduced in the M3 and brings dynamic caching, hardware-accelerated ray tracing, and hardware-accelerated grid coloring feature time to the iPad for the first time. The M4 gives a huge boost to professional rendering performance in applications such as Octane and is now 4 times faster than the M2.

 

Compared to the powerful M2 in the previous iPad Pro, the M4's CPU performance has increased by 1.5 times. Whether it's handling complex orchestral files in Logic Pro or adding demanding effects to 4K video in LumaFusion, the M4 improves performance across professional workflows.

 

From the memory point of view, the M4 chip uses the faster LPDDR5X, and the unified memory bandwidth reaches 120GB/s. LPDDR5X is an interim update to the LPDDR5 standard that offers higher memory clock speeds than LPDDR5, up to 6400 MT/ s. The current LPDDR5X speeds up to 8,533 MT/ s, but the M4's memory clock speeds only reach about 7,700 Mt/s.

 

 

According to industry data, the Apple M3 has up to 24GB of memory, but there is no more data to indicate whether Apple will increase the memory expansion issue. Each iPad Pro will come with either 8GB or 16GB DRAM, depending on the model.

 

The new neural network engine referenced by the M4 chip has 16 cores and a speed of 38 trillion calculations per second, which is 60 times faster than the first neural network engine on Apple's A11 Bionic chip.

 

In addition, the M4 chip also features a new display engine designed with pioneering technology to achieve the stunning precision, color accuracy and brightness uniformity of the Ultra Retina XDR display, a state-of-the-art display created by combining the light from two OLED panels.

 

Johny Srouji, Apple's senior vice president of hardware technologies, said the M4's energy-efficient performance and its new display engine made the iPad Pro's thin, thin design and game-changing display possible, while the fundamental improvements in CPU, GPU, neural engine and memory systems make the M4 ideal for understanding the latest applications leveraging artificial intelligence. All in all, the new chip makes the iPad Pro the most powerful device in its class.

 

2024: The first year of AI PC

 

At present, AI artificial intelligence is a giant star in the global industry, in addition to servers and other markets, the consumer market ushered in a new opportunity - AI PC. The PC industry has been in a state of silence since the semiconductor industry entered the downward cycle, and under the current tide of AI, AI PC is like a light illuminating the PC market.

 

Previously, TrendForce Consulting said that in 2024, it will expand to more edge AI applications, or continue the AI server foundation, pushing to AI PC and other terminal devices. Due to the rigorous application of Edge AI, it will return to the terminal AI PC, so as to implement the workload of decentralized AI servers, and expand the possibility of AI use scale.

 

However, at present, the market's definition of AI PC is still unclear. Broadly speaking, AI PC refers to a computer with generative AI functions. At present, generative AI such as ChatGPT relies on the cloud platform to perform operations, and through AI PC, users do not need to rely on the cloud to perform AI operations on the device. If enterprises or individual users want to use AI applications to process private data, they do not have to upload the data, and AI PCS can directly process it, not only reducing latency, but also improving security protection. This also poses more challenges to the chips used in AI PCS.

 

According to Apple, the neural engine in the M4 chip is Apple's most powerful neural engine to date, more powerful than any neural processing unit in any AI PC today. According to Tim Millet, Apple's VP of Platform Architecture, the M4 will provide the same performance as the M2 at half the power consumption. Compared with the new generation of PC chips in thin and light laptops, the M4 can provide the same performance with a quarter of the power consumption.

 

The continuous dynamics from other big factories indicate that the AI PC competition is becoming more and more intense, and at the same time, the industry also has expectations for the prospect of AI PC. Microsoft says 2024 is the first year of the "AI PC." The company estimates that by 2026, half of commercial computers will be AI PCS, based on the time it takes for PC brands to launch products.

 

Intel has stressed that AI PC will become a major turning point in the recovery of the PC industry, and AI PC also plays a key role in the industry highlights in 2024. Intel Pat Gelsinger previously said at the meeting that driven by AI PC demand and the Windows update cycle, customers continue to add processor orders to Intel, and Intel's AI PC CPU shipments in 2024 are expected to exceed the original target of 40 million.