I had been publishing books, printed music, and music recordings for decades when the internet came into existence. As a co-founder of a major publishing company, I was fortunate to have access to the most advanced technology available at the time. Publishing tech evolved, but not at the rate it is currently moving.

When I struck out on my own along with my editor wife, budget concerns required that I construct our own computers using components purchased at the best price I could find. Sometimes I had to apply and become a reseller of computer components with the secret intention of only purchasing at wholesale the components we needed for our publishing company.

So I am familiar with the rapid advancements that have occurred over the years. However, I was blind-sided at first when I started using artificial intelligence (AI) that a component that I watched evolve from one purpose blossomed into a ridiculously large and important component for AI. The graphic display card with a GPU (Graphic Processing Unit) has always been vital to display text, images, video, animation and everything on the computer display. Now, the current versions of the GPU have the double duty of processing data for AI. Why does a graphics display card with a GPU have anything to do with processing data? I thought the computer’s CPU (Central Processing Unit) solely handled data processing.

While the CPU is indeed responsible for general-purpose computing tasks and acts as the brain of the computer, newer GPUs have specialized hardware designed specifically to handle specialized graphics-related computations. However, the architecture and capabilities of GPUs also make them well-suited for certain types of data processing tasks, particularly those involved in artificial intelligence and parallel computing.

Here’s why I discovered GPUs are increasingly used for data processing alongside CPUs:

Parallel Processing: GPUs consist of numerous small processing cores capable of executing multiple tasks simultaneously. This architecture is highly parallel, making GPUs particularly efficient at performing repetitive, data-parallel tasks. While CPUs also have multiple cores, they are typically optimized for sequential processing rather than parallel tasks.

Matrix Operations: Many data processing tasks, especially those involved in machine learning and deep learning, rely heavily on matrix operations like matrix multiplication. GPUs are well-suited for these operations due to their architecture, which can perform matrix computations efficiently across multiple cores simultaneously. This capability significantly accelerates tasks like training neural networks.

Performance Efficiency: For certain types of data processing tasks, such as those involving large datasets or complex computations, GPUs can significantly outperform CPUs in terms of performance and efficiency. This is because GPUs are optimized for handling specific types of computations, such as those found in graphics rendering and AI algorithms.

Specialized Hardware: GPU manufacturers, such as NVIDIA, have developed specialized hardware features like Tensor Cores specifically tailored for accelerating certain types of computations common in artificial intelligence workloads. These hardware accelerators further enhance the performance of data processing tasks on GPUs.

Software Support: The software ecosystem around GPUs has evolved to support a wide range of data processing tasks. GPU-accelerated libraries and frameworks, such as CUDA, cuDNN, and TensorRT, provide developers with the tools they need to leverage the computational power of GPUs for various data processing applications.

While CPUs remain essential for handling general-purpose computing tasks and managing system operations, GPUs offer a complementary processing capability that can significantly accelerate certain types of data processing tasks, especially those involved in AI, scientific computing, and parallel computing. As a result, modern computing systems often utilize both CPUs and GPUs to maximize performance and efficiency across a broad range of applications.


As a footnote, because of the dramatic explosion of AI, the demand for GPUs has also exploded. In fact the most prominent manufacturer of the GPU, NVIDIA has seen a windfall with no end in sight:

NVIDIA revenue for the quarter ending January 31, 2024 was $22.103B, a 265.28% increase year-over-year.
Prior years:
2023 $26,974
2022 $26,914
2021 $16,675
2020 $10,918
Financial data from Macrotrends

Additional Information

Why GPUs Are Great for AI
LLM (Large Language Models)