Reports about Nvidia’s RTX 50-series GPUs are circulating, hinting at a possible release in late 2024 or early 2025. Despite the distant timeline, speculations are rife regarding their specifications, architecture, performance, and power demands.
According to recent leaks from tipster @kopite7kimi, the RTX 50 “Blackwell” GPUs will utilize the GB202 and GB203 architectures. The former is anticipated as the flagship variant, while the latter is expected to offer a pared-down version with fewer cores.
Interestingly, there’s uncertainty surrounding the utilization of a multi-chip package for the GB202. Another rumor suggests that these GPUs might feature a 512-bit memory interface, a departure from the 384-bit interface seen in the current-gen AD102 Ada GPU.
Contradictory to earlier rumors, the upcoming GPUs might employ 28Gbps GDDR7 memory, offering improved bandwidth over previous generations. While Nvidia is likely to stick with 16Gbit densities for the GDDR7 memory chips, details regarding memory bus widths remain unknown.
Should these speculations hold true, the RTX 5090 could see a significant performance leap owing to the additional cores, enhanced memory bandwidth, and wider memory bus.
However, it’s essential to approach these rumors with caution, as details may change before the eventual launch of the RTX 50-series GPUs.
