The specifications for Nvidia's flagship next-gen RTX 50 series graphics cards may have been leaked. Fresh claims say that the RTX 5090 will feature as much as 32GB of GDDR7 memory, as well as a GB202-300 GPU with 21,769 CUDA cores. The RTX 5080 has also been detailed.
Focusing on the flagship RTX 5090, kopite7kimi claims that the upcoming GPU will feature a 512-bit memory bus (up from the rumoured 448-bit) and 32GB of GDDR7 RAM. The flagship card is said to use the GB202-300 GPU with 21,760 FP32 (CUDA) cores. Additionally, the flagship model is rumoured for 600W power draw, but it's unclear whether this refers to TBP or TGP.
In a separate report from Benchlife (via VideoCardz), it's also claimed that the flagship RTX 50 series card will feature a 14-layer PCB, a PCIe 5.0 interface, and UHBR20's DisplayPort 2.1a. The card will be powered by a single 12V-2×6 connector, contrary to some recent claims suggest Nvidia would opt for two of these connectors.
As for the RTX 5080, kopite7kimi claimed the graphics card would be based on the GB203-400 and pack 10,752 CUDA cores, along with 16GB of GDDR7 memory and a 256-bit memory bus. Notably, the RTX 5080 features almost half the cores and memory of the GB202, including the memory bus width, and is expected to require 400W of power, up from the RTX 4080's 320W. Given the gap between the RTX 5080 and the RTX 5090, there is plenty of room for a Ti or Super variant.
It is currently unknown when Nvidia plans to announce its next-generation gaming graphics cards. Previous rumours have both claimed late 2024 and early 2025 release windows, so we'll just have to wait and see.
Discuss on our Facebook page, HERE.
KitGuru says: Nvidia hasn't announced anything official for its next-gen RTX GPUs yet, so for now, all of this falls into the rumour category.