Since the very inception of Moore’s law in 1965 by Gordon Moore, co-founder of Intel, the whole semiconductor industry has been about shrinking the size of transistors.
The world saw its very first chipset milestone in 1971, the Intel 4004, based on a 10-micron process. By the 1980s, we had already moved to 1-micron-based Intel 80486. Within the next 10 years, the world saw much smaller 0.8-micron-based Intel Pentium. By the 2000s, Intel’s Core 2 Duo and other processors reached 65 nanometers, marking a new era of power efficiency and performance. The 2010s ushered in 7-nanometer technology, and by 2020, Apple’s A14 Bionic leveraged a 5-nanometer process, setting new benchmarks in mobile computing.
And in 2024, we are using 3 nm-based Apple Silicon, Qualcomm Snapdragon Elite processors, and Mediatek processors, and have started talking about 2 nm-based chipsets.
Source :IEEE Spectrum
Clearly, the evolution has been totally based on shrinking the size of transistors. With the aim that as we move forward, chipsets will become much cheaper and more powerful, and yes, it happened, but still, we are not able to fulfill the promise of smart cities, an automated world, etc.
Today, standing at that brink where the researchers are struggling to further reduce the size of transistors, if we look back and just ask, was our approach right?
We could say yes, about 95%, but we lacked a proactive approach towards how we approached the architecture of the chipsets. We started building unique chips for specific tasks; yes, it helped but also resulted in growing complexity and increased cost.
The Evolution of Computing and the Need for Universal Processors
CPUs were developed when a need for general-purpose processors was felt; GPUs came into existence in the late 1990s when someone thought of doing the processes in parallel instead of serially, and so did NPUs come into existence for AI tasks and accelerating neural computations.
So we kept on designing stuff as we needed.
It’s like setting up a power plant only for the current demand and keeping up with increasing the number of power generation units as the demand increases.
Now we have one of the most beautiful technologies humankind has ever made, artificial intelligence and machine learning, and it is the right time that we start thinking about redesigning the architecture for our processors instead of just thinking about shrinking the size of transistors.
Join Our WhatsApp Community
That is basically the idea of a universal processor, a processor based on a flexible processor architecture. It is what the title says: “The future of processors lies in adaptability, not size.“
Why and What is a Universal Processor?
As the name suggests, a universal processor is a processor based on a flexible processor architecture that aims to simplify computing by incorporating the flexibility to handle any workload, be it general-purpose computing, graphics, or AI tasks, all within a single architecture. These processors are designed to dynamically adjust their processing power depending on the task at hand.
Why Do We Need a Universal Processor?
As discussed above, our current processors lack a few things, like adaptability, a unified architecture, and the most important problem is “the lack of scalability,” and universal processors aim to fix all these problems.
- Adaptability: The processor can shift between different workloads, such as handling a traditional computation task or performing AI inference, without requiring multiple chips or complex reconfiguration.
- Unified Architecture: Instead of having separate cores for CPUs, GPUs, or NPUs, a universal processor has a unified architecture that can be adjusted for different functions based on the software requirements or task load.
- AI-Driven Efficiency: With the help of AI, these processors can intelligently allocate their resources based on the specific demands of the task. For example, the processor could dynamically allocate more resources to AI tasks when needed, without the need for a separate NPU.
- Scalability: Universal processors can scale from small, embedded devices to high-performance computing systems, making them highly versatile across different industries and applications.
Conclusion
The future is not far away when a need for Universal Processor and flexible processor architecture will arise ultimately changing the way we see processors now a days and at that time hope:) this WireUnwired Report will cover the internet “The Future of Processors Lies in Adaptability, Not Size: Universal Processors.”
By the way, whatever you think about it, you must write in the comments and contribute towards making the future Universal.
Discover more from WireUnwired
Subscribe to get the latest posts sent to your email.
