Nvidia RTX 50 series will usher in a performance leap

At present, Nvidia is advancing the Ada Lovelace architecture GPU plan in an orderly manner, and will soon usher in a new member of the GeForce RTX 40 series, the RTX 4070 Ti, which will gradually move toward the mainstream market. According to the cycle of NVIDIA technology iterations, the next large-scale architecture update will probably be two years later.

It is understood that Nvidia will modify the existing SM design and adopt a new unit structure; at the same time, the RT optical tracking unit will introduce a denoising accelerator, which can greatly improve the optical tracking performance; the new generation of GPU will switch from a single chip to an MCM package, using a high-speed bus to realize the interconnection between small chips. In addition, it is rumored that the Blackwell architecture GPU will be manufactured using a 3nm process. It may be Nvidia’s first GPU packaged in MCM. Regardless of whether the RTX 50 series GPU code name is Blackwell, it is more likely to be manufactured using the same semiconductor process.

With the support of the new architecture, new process, and new packaging, GeForce RTX 50 series GPUs may bring surprises in terms of performance, in order to achieve Nvidia’s goal of continuing to maintain performance leadership. If AMD does not perform better on the next-generation RDNA 4 architecture, the gap between the two sides may even widen further, which is not good news for players.