Nvidia’s first Blackwell architecture gaming graphics cards include RTX 5090/5080
This year, NVIDIA is set to release its next-generation gaming graphics cards, the GeForce RTX 50 series, based on the Blackwell architecture. The series is expected to debut as early as the fourth quarter of this year, initially focusing on high-end products, similar to the previous GeForce RTX 30/40 series. Recently, there has been significant interest among gamers regarding whether both the RTX 5080 and RTX 5090 will be released within the same year, with a particular focus on the RTX 5080.
Recent online disclosures suggest that NVIDIA is likely to launch both the RTX 5090 and RTX 5080, though their release dates will not coincide, being spaced a few weeks apart. The RTX 5080 is expected to hit the market earlier. This is a slight departure from the GeForce RTX 40 series pattern; although both the RTX 4090 and RTX 4080 were announced on September 20, 2022, the higher-end RTX 4090 was released first on October 12, followed by the RTX 4080 about a month later on November 16.
According to previous information, the Blackwell architecture-based GeForce cards will feature five different chips: GB202, GB203, GB205, GB206, and GB207, with respective memory bus widths of 512/384-bit, 256-bit, 192-bit, 128-bit, and 128-bit. The RTX 5090 will be equipped with the GB202 chip, while the RTX 5080 will use the GB203. These new products are manufactured using TSMC’s 4NP custom process, an enhanced version of the existing 4N process, and will also support the new generation GDDR7 memory.
The difference between GB202 and GB203 chips is significant, with the former being substantially larger and expected to feature 192 SMs (24,576 CUDA cores), while the latter will have half that number, with 96 SMs (12,288 CUDA cores). The RTX 5090 will not utilize all SMs, with plans to activate about 170 to 180 groups. The RTX 5080 will follow a similar strategy, enabling additional SMs in future mid-generation “Ti” or “SUPER” models to enhance performance.
