AI Accelerator Hardware in Personal Computers: Is It Already Here?
As the world increasingly relies on artificial intelligence (AI) for various applications, the availability of AI accelerator hardware in personal computers has become a topic of interest. From mid-range GPUs to the latest processors, the integration of AI acceleration in hardware is becoming a standard feature. In this article, we explore whether AI accelerator hardware has already made its way into personal computers, and if so, which options are most suitable for those looking to harness AI power in their devices.
RGX-RTX GPUs and Tensor Cores
The NVIDIA RTX series of GPUs is a prime example of AI accelerator hardware in personal computers. RTX GPUs come equipped with tensor cores, specifically designed to accelerate AI computations. These tensor cores are widely used for applications like DLSS (Deep Learning Super-Sampling), which is an AI-based image upscaling technology. DLSS is particularly popular in gaming, as it improves frame rates without sacrificing image quality. Additionally, tensor cores are essential for AI training and inference, making them invaluable for software such as Stable Diffusion.
Intel and AMD’s AI Hardware Integration
Intel’s Iris Xe and Arc GPUs, as well as AMD’s RDNA3 GPUs, also offer dedicated AI-acceleration hardware similar to NVIDIA’s tensor cores. Intel’s GPU AI cores can be used for technologies like XeSS (Xavier Super Sampling), which competes with DLSS for AI upscaling. This technology enhances image quality and offers a good balance between performance and visual fidelity. AMD’s FSR (FidelityFX Super Resolution) technology is designed for upscaling video content, although its current versions do not utilize AI, resulting in less demanding but less effective performance across older GPUs.
Integrated NPUs in the Latest Processors
Apple, AMD, and Qualcomm have integrated NPUs (Neural Processing Units) into their latest mobile and laptop processors, designed to accelerate AI applications. These integrated NPUs can be used for tasks such as image enhancement. Although there is no common software that currently utilizes AMD or Intel’s integrated NPUs, the tech giant Microsoft has announced plans to incorporate AI features requiring NPUs into future versions of the Windows operating system, presumably version 12.
Desktop PCs and AI Acceleration
The addition of NPUs in desktop CPUs is still limited. AMD recently introduced the Ryzen 7 8700G and Ryzen 5 8600G, both equipped with integrated NPUs. However, none of AMD’s other desktop processors include this feature. Meanwhile, Intel is expected to launch their first desktop CPUs with integrated NPUs, codenamed “Arrow Lake,” in late 2024. For those requiring or wanting more powerful AI acceleration, purchasing a graphics card remains the recommended approach, as it offers significantly better AI performance compared to integrated NPUs. The cheapest RTX GPUs have approximately double the AI performance of the best integrated NPUs available.
Conclusion
In summary, AI accelerator hardware has indeed made its way into personal computers, with options such as NVIDIA RTX GPUs and integrated NPUs in latest processors. For the best performance and versatility, NVIDIA’s RTX GPUs remain the top choice for most people. However, the landscape is evolving, with potential advancements in the desktop CPU market, making AI acceleration more accessible to a broader audience.