As we get closer to the launch of the RTX 50 series of Blackwell-based home graphics cards, NVIDIA is revealing more technical details about the products. Agora a empresa o tamanho dos dies das GPUs da linha GB200, e também suas contagens de transistores.

Evidently, the largest die in NVIDIA’s new family of graphics cards is the GB202, used in the RTX 5090. It measures 750mm², surpassing even previous rumors that estimated 744mm² for its size. Não é o maior die que uma GPU doméstica da NVIDIA já teve, mas é o terceiro maior.

• GB202 (Blackwell): 750mm²
• AD102 (Ada): 608mm²
• GA102 (Ampere): 628mm²
• TU102 (Turing): 754mm²
• GV100 (Volta): 815mm²
• GN200 (Maxwell): 601mm²
• GK110 (Kepler): 551mm²

The GB202 die also had the biggest growth compared to its predecessor model among the RTX 50 series cards. Interestingly, this die growth is not proportionally reflected in the number of transistors. The GB202 has 92.2 billion of them, much more than the 76.3 billion used in the AD102, the RTX 4090’s GPU. However, calculating the density of the transistors, the RTX 5090 has 122.9MTr/mm², enquanto a antecessora alcança 124,9MTr/mm².

In fact, as shown in the table below organized by VideoCardz, the three official models of the NVIDIA RTX 50 series have seen a reduction in their density of transistors per square millimeter of die:

In other words, for its Blackwell generation of domestic video cards, NVIDIA was unable to work with smaller transistors that facilitate an evolution for the architecture. Provavelmente esse é um dos motivos para a densidade dos componentes ter até diminuído em relação às placas antecessoras.

Related news:

If the rumors are correct, we will have reviews of the RTX 5090 being released on January 24th, and there will probably be no shortage of comparisons with the previous generation.