Apple introduced its memory architecture called the Unified Memory Architecture with the release of the Apple M1 system-on-a-chip in June 2020. This hardware component is now an integral part of all Apple-designed chips. These include the Apple M series used in Mac devices and the iPad Air and iPad Pro devices and the Apple A series used in iPhone and entry-level iPad devices. The Apple Unified Memory provides several performance improvements and notable advantages. It also has some limitations and disadvantages due to the fact that it is a hardware component that is integrated into a single chip architecture.
Understanding the Unified Memory Architecture of Apple: Purpose, Mechanism, Advantages, and Disadvantages
Shared memory architecture has been utilized in several computer system designs. Consider parallel computing as an example. Coprocessors such as integrated graphics processors that are integrated in a single chip together with the central processing unit or main processor use the same random access memory or RAM of the entire computer system. This is similar to the Unified Memory Architecture of Apple. However, because the memory is also an integrated hardware component of a single system-on-a-chip, its implementation is technically different from a conventional shared memory architecture.
Purpose and Operational Mechanism
A random access memory or RAM is a hardware component that provides a temporary repository for data used by a computer system at a given instant. This allows faster real-time processing of running software or applications Using a RAM is more efficient than using a storage drive such as hard disk drives or solid-state drives because it has faster read and write processes. Traditional RAMs exist as a separate component that fits into a motherboard slot. The Unified Memory Architecture from Apple has a different approach because the RAM is built within the same chip that houses the main processor and coprocessors.
The purpose of the Unified Memory Architecture is to allow the CPU and GPU or other coprocessors in an Apple Silicon to share and access the same memory. This enables faster and more efficient parallel computing operations. Apple specifically introduced this chip design when it dumped Intel processors with the launch of its ARM-based Apple M1 chip in 2020. The Unified Memory component has become a standard part of all of its chips.
It is important to note that traditional memory systems work with the CPU having its own RAM and a discrete GPU also having its own video RAM or VRAM. However, when running graphics-demanding apps or tasks, the data from a particular application is first moved into the RAM for central processing and it then moves to the VRAM for graphics processing. A shared memory architecture has a simpler process because the CPU and GPU use the same RAM but each has access to different spaces in the memory. Both approaches involve a lot of data movement. This can strain the system and make it less efficient.
A Unified Memory Architecture has a simpler design and process. The RAM is located in close physical proximity to the CPU and GPU and other coprocessors such as the Neural Engine. There are no connection bottlenecks unlike traditional memory architecture and conventional shared memory architecture that connect the processors with the RAM via the motherboard. The processors in a single chip also access the same memory pool.
Key Advantages and Disadvantages
The introduction of the Apple M1 chip, its ARM-based architecture, and the Unified Memory Architecture was met with skepticism and criticism. However, based on results from benchmark tests and real-world applications, the new generations of chips from Apple have proven to deliver mind-blowing performance per watt. The Unified Memory has a critical part in this performance improvement. Below are the specific advantages:
• Increased Performance: The memory architecture from Apple allows the CPU, GPU, and other coprocessors to access the same data without having to go through separate memory pools and memory controllers. This eliminates or minimizes the redundancy of data copied between different memories or memory spaces. The proximity of the RAM to the processors also allows high-bandwidth and low-latency data transfer.
• Improved Efficiency: Another advantage of the Unified Memory Architecture is the reduced overall power consumption of the entire computer system. The memory controller is one of the main sources of power draw of a computer system. Eliminating the need for separate controllers and minimizing redundancies in data copies and storage result in a significant reduction in power consumption.
• Reduced Complexity: This memory architecture also simplifies the design of the chip and the entire computer system. This can help in the mass production of chips and the actual devices while also lowering their manufacturing costs. It also provides better flexibility when it comes to chipmaking for different use cases or applications because it allows Apple to optimize the chips for different workloads.
Traditional memory systems are still popular in most computer systems. Take note that high-performance desktop computers or those built for graphics-intensive PC gaming use CPU and discrete GPU with separate RAM and VRAM components. The performance of these computers is still incomparable compared to Mac computers powered by Apple M chips. Below are the specific disadvantages of Unified Memory:
• Resource Competition: The CPU, integrated GPU, and other coprocessors compete for memory resources. The system cannot allocate more memory to a processor without taking away from the other. This can result in performance overhead during situations that involve the CPU running multiple applications at once while the GPU is rendering high-resolution graphics or running graphics-intensive applications.
• Upgradability Issue: Another notable disadvantage of Unified Memory is that it cannot be upgraded. Owners and users of Mac devices running on M chips cannot replace the Unified Memory to increase its capacity because the component is integrated into the system-on-a-chip. The Secure Enclave processor of Apple also prevents hardware upgrades and replacements under certain circumstances.
• Compatibility Concerns: The memory architecture of Apple is a proprietary technology that exclusively works on Apple devices or systems and authorized applications. This means that some third-party devices, components, and applications are not compatible or optimized for its memory architecture. An example would be a virtualized or emulated Windows or Linux running on M-powered Mac computers.
