Photos of a PCB allegedly belonging to Nvidia’s next-generation GeForce RTX 5090 have leaked online, offering a glimpse of the company’s future flagship GPU. Shared on Chinese tech forum Chiphell, the images reveal certain aspects of the RTX 5090’s architecture, hinting at a major performance leap for Nvidia’s upcoming Blackwell architecture.
The leaked image showcases an unsoldered PCB with a large BGA (Ball grid array) package for the GB202 GPU, which is expected to power graphics cards including the RTX 5090 and potentially the RTX 5090 Ti / RTX 5080 Ti in the future. Interestingly, the PCB lacks any Nvidia branding, indicating it might belong to a third-party vendor. According to Videocardz, markings on the bottom right of the PCB suggest it’s a PNY model based on Nvidia’s PG145 (GB202) board.
The RTX 5090 is also expected to feature 32GB of next-generation GDDR7 memory. If you look at the VRAM layout in the leaked PCB photo, there are a total of 16 solder pads around the main GPU core. While this confirms the possibility of 32GB VRAM, Nvidia could potentially jump to 48GB or even 64GB, depending on the memory modules used. This significant upgrade in memory capacity and bandwidth positions the GPU as a powerhouse for demanding workloads, such as 4K and 8K gaming, AI applications, and professional content creation. Notably, GDDR7 memory is expected to bring higher data transfer speeds and better energy efficiency compared to its predecessor, GDDR6X.
Another significant detail from the PCB design is the inclusion of a single 12V-2×6 power connector, capable of delivering up to 600W of power. However, overclocked versions with dual connectors are likely to emerge from third-party manufacturers, catering to enthusiasts seeking maximum performance. The 12V-2×6 power connector was introduced as a replacement to the controversial 12VHPWR connector that led to melting RTX 40-series GPUs. Hopefully, Nvidia is better prepared this time to avoid similar fiascos.
Previously leaked specifications indicate that the RTX 5090 will adopt DisplayPort 2.1a, a major upgrade from the aging DP1.4a standard that Nvidia has used since 2016. The new standard supports the full bandwidth UHBR20 spec, enabling nearly triple the data transfer capacity of its predecessor. This paves the way for impressive display capabilities, such as 8K resolution at 240 Hz or even 16K at 60 Hz with DSC.
In addition to the DisplayPort advancements, the RTX 50-series is set to become Nvidia’s first consumer GPU lineup to support PCIe 5.0, doubling external bandwidth to 64GT/s. While this enhancement may yield minimal benefits in gaming, it could prove advantageous for non-gaming tasks, such as AI workloads and multi-GPU setups.
Expected the RTX 5090 to be unveiled officially during CES 2025, likely alongside other members of the RTX 50 series, like the RTX 5080, RTX 5070 Ti, and RTX 5070. Industry insiders suggest that the RTX 5080 will launch first, with the RTX 5090 following shortly afterward.