Whether you believe it or not, AMD and Nvidia are joining forces.

AMD and Nvidia, despite being competitors in the graphics card market, have decided to collaborate on the launch of the new 5th generation Epyc server CPUs, known as Turin. This collaboration is presented as a "technical partnership" in which Nvidia offers advice on integrating its GPU clusters, both HGX and MGX, with the newly released AMD CPUs. Although AMD has its own AI accelerators, the alliance arises from a common interest in meeting market demand, where many customers will prefer to combine Nvidia GPUs with AMD CPUs.

The architecture of AMD's new CPUs, Zen 5, which also inspires models such as the Ryzen 9 9950X, promises exceptional performance, reaching variations from eight-core models to massive chips with 192 Zen 5c cores. A significant milestone is that the new 64-core model, the 9575F, can reach frequencies of 5 GHz, a remarkable advancement within AMD's data center lineup.

In terms of performance, AMD claims impressive improvements: the new processor series shows a performance that is 2.7 times higher in benchmark tests compared to Intel Emerald Rapid Xeon CPUs, and up to four times more in enterprise tests involving video transcoding. While Intel has introduced its Granite Rapid Xeon CPUs, AMD has yet to disclose comparative performance figures.

On the other hand, AMD has not fallen behind in the development of GPUs for data centers. Alongside the launch of the Turin CPUs, the company introduced its new AI accelerator, the Instinct MI325X, which it had previewed at Computex. This device includes 256 GB of HBM3E memory and can manage up to 2 TB of HBM3E memory and 48 TB/s of memory bandwidth in a cluster. In terms of performance, the MI325X surpasses Nvidia's H200 in AI inference tasks and competes equivalently during AI training, highlighting the rapid evolution of hardware in this field.

Additionally, Nvidia and AMD are exploring the development of ARM-based processors aimed at PCs, a move that would mark a significant shift in the client computer market with Windows. Both companies are announcing their intention to launch system-on-chip (SoC) products starting in 2025, seeking to leverage the success of the ARM architecture, exemplified by Apple's M1 and M2 chips, to enhance adoption in the Windows space.