In recent years, discussions about whether the computer and smartphone markets are in decline have been rampant. Recently, memory chip manufacturer Micron lowered its revenue forecast for the upcoming quarters due to disappointing sales of AI computers and smartphones, raising concerns that "AI is fading away." However, in reality, there are no signs of decline for AI, especially evident from Nvidia's performance.
Image Source Note: Image generated by AI, licensed by service provider Midjourney
Currently, many laptops and smartphones claiming to have AI capabilities lack sufficient processing power. Even high-performance gaming PCs struggle to run complex AI applications like ChatGPT locally, as these applications require massive data and computational power that cannot simply be handled by personal computers. While there are some alternative applications available, they fall short in performance and response speed compared to most AI programs running on servers.
In the AI ecosystem, the companies and tools that stand out have largely established themselves. For example, users with Nvidia RTX graphics cards can typically outperform many modern CPUs equipped with NPUs in terms of AI performance. Comparatively, the performance difference between an RTX 4080 and a laptop with an Intel Core Ultra 9185H under AI workloads can reach up to 700% to 800%. This highlights the critical role of servers in providing AI performance.
Google has expanded its AI model Gemini to most Android devices and plans to implement it in Nest speakers. Although these devices have been available for four years, they still demonstrate the broad applicability of AI technology. Looking back, the performance of graphics cards was once thought to need to reach tens of trillions of calculations (PFLOPs) to achieve a true virtual reality experience, and current graphics cards have yet to meet this standard, reflecting the challenges still faced by local AI development.
In the development process of GPU manufacturers, AI programming often relies on parallel computing, where GPUs excel. Therefore, future GPU designs will still require time, and significant AI performance improvements may not be seen until the RTX 60 series is launched. This generation of graphics cards could potentially enable the operation of large local models (LLMs).
Key Points:
🌟 AI technology is not fading; market performance is influenced by misconceptions.
💻 Many devices claiming to be AI rely on server performance, making complex calculations difficult to achieve locally.
🚀 Future advancements in GPU technology may drive the development of local AI models.