The end of Moore's Law is starting to feel uncomfortable.

For over twenty years, the idea that Moore's Law has reached its end has been circulating. This theory, formulated by Gordon Moore, co-founder of Intel, suggested that the number of transistors on a microchip would double approximately every two years. In 2006, Moore himself anticipated that this phenomenon would end in the 2020s. In 2016, MIT professor Charles Leiserson declared the Law dead. In 2022, the CEO of Nvidia made the same claim, although days later the CEO of Intel contradicted this statement. Despite the fact that Moore's Law has driven remarkable innovation in desktop processors, its decline is not a singular event, but rather a gradual process that we are beginning to observe.

Recently, AMD and Intel have launched new generations of processors that have not had a triumphant start. For instance, Intel's Core Ultra 9 285K, while featuring significant design improvements, fails to adequately compete with previous offerings. The Ryzen 9 9950X, although an advancement over its Zen 4 predecessors, does not represent the generational improvement that was once expected. Analyzing the figures reveals that the multicore performance improvement from the Ryzen 9 5950X to the 7950X was 36%, while from the 7950X to the 9950X, the increase was only 15%. In transcoding software like Handbrake, the transcoding speed of the 7950X compared to the 5950X increased by 34%, but with the 9950X, the improvement was only 13%.

This trend is not isolated; in terms of single-core performance improvements between the Core i9-10900K and the i9-12900K, Intel achieved a 54% increase. However, comparing the i9-12900K, which has already seen three generations, with the new Core Ultra 9 285K, the improvement was only 20%. Additionally, the performance of the new Core Ultra in Cinebench is unusually high, but in other applications, regressions are observed compared to previous generations.

In recent years, the rate of performance improvements has notably decreased. Moore's Law focuses on the number of transistors, implying a direct relationship with performance. However, increasing the number of transistors is not as simple as before, and the concept of Dennard scaling also plays a role in this limitation. Although AMD and Intel may not openly discuss it, it is evident that both companies are adapting. Intel, for example, has shifted to a hybrid architecture and introduced a radical redesign with its Arrow Lake CPUs. Meanwhile, AMD has been using 3D V-Cache technology to enhance performance, leveraging the large number of transistors in cache units.

A key aspect to consider when analyzing Moore's Law is the physical space of the chips. While it is possible to build microprocessors with a high number of transistors, questions arise regarding their energy consumption and whether they can operate efficiently in devices such as PCs or servers. Each square millimeter of silicon is costly, and continuing to double the transistors may not be economically viable.

The cost of chips has increased, and while there are arguments against the pricing strategies of companies like Nvidia, it has been reported that this company has faced higher prices from TSMC for its RTX 40 GPUs compared to previous prices with Samsung. Although the RTX 4090 has more than double the transistors of the RTX 3090, both models face challenges regarding energy consumption, size, and production costs.

It is clear that Moore's Law is no longer the only reference point for advancements in hardware. Innovation is not dead; it is shifting toward other paradigms. Artificial intelligence is emerging as a solution that could offer significant performance improvements that do not solely rely on adding more transistors.

For example, DLSS technology allows for enhanced performance without necessarily requiring an increase in hardware specifications. Advances in AI processing, such as Nvidia's Ray Reconstruction, are transforming how performance demands are managed in current hardware.

In conclusion, regardless of the existence of Moore's Law, it is vital for companies like AMD, Intel, and Nvidia to continue innovating to meet the growing performance demands of the market, even if the way they do so may differ from what we have seen so far.