Nvidia’s upcoming GeForce RTX 5090 and RTX 5080 graphics cards have been the subject of recent leaks, revealing potential specifications that suggest a significant leap in performance. According to a reputable hardware leaker, the flagship RTX 5090 is expected to feature the GB202 graphics processor with an impressive 21,760 CUDA cores and 32GB of GDDR7 memory using a 512-bit interface. However, this massive number of cores is projected to consume up to 600W of power, making it one of the most power-hungry consumer GPUs to date.
In contrast, the RTX 5080 is rumored to be powered by the GB203 GPU, featuring 10,752 CUDA cores and 16GB of GDDR7 memory with a 256-bit interface. The RTX 5080 is expected to have a TGP rating of 400W, which is still a significant increase from its predecessor. Despite having half the number of stream processors and memory interface width compared to the RTX 5090, the RTX 5080 should still offer a substantial performance boost over the previous generation, particularly in high-resolution gaming.
The considerable performance gap between the RTX 5090 and RTX 5080 has raised questions about Nvidia’s strategy. Some speculate that the company may be aiming to create a new tier of performance or limit the appeal of certain consumer cards as AI compute alternatives.
Nvidia RTX 5090 and 5080 specs
One technical explanation suggests that the GB202 processor could consist of two GB203 dies, utilizing a multi-chiplet design similar to the GB100/GB200 data center GPUs. However, implementing such a design in a consumer-grade product would be costly. An alternative approach would be to build a monolithic graphics processor with over 21,760 CUDA cores using TSMC’s 4nm-class process technology.
While this is possible, it would result in a large die size of around 650 mm², making it challenging and expensive to produce. Given the current focus on AI, Nvidia might prioritize creating a massive data center part and then adapt it for consumer offerings. If this is the case, pricing for the RTX 5090 could be even higher than the already expensive RTX 4090, with AI variants potentially arriving before the consumer models.
It is important to note that the information regarding the Blackwell-based graphics cards for client PCs is still unofficial, and conflicting rumors have emerged about the release date. Some sources suggest that the RTX 50-series may not arrive until early 2025, giving Nvidia ample time to refine the products before launch. As enthusiasts eagerly await official announcements from Nvidia, the leaked specifications have generated excitement about the potential advancements in graphical power that the RTX 5090 and RTX 5080 could bring to both desktop and laptop gaming.
