GeForce RTX 5090 Specs Leak Reveals Monster GDDR7 Upgrade And Potential Triple PCB

rog 4090
NVIDIA may be in the news lately mostly because of its AI prowess, but its GeForce GPUs are still vital to millions of gamers worldwide. The GeForce RTX 5090, the presumed flagship of the upcoming RTX 50 series Blackwell GPU lineup, is looking mighty promising according to some recent rumors. 

According to a post on Chiphell, the GeForce RTX 5090 appears to have an unusually dense memory layout. This sets it apart from the current GeForce RTX 4090, which will likely have a lower memory capacity than the upcoming RTX 5090, that's comprised of fewer memory chips.

Coupled with GDDR7 memory technology, which has upgraded speeds and efficiency, the RTX 5090 will reportedly offer massive VRAM updates. With a potential 512-bit memory bus, the Blackwell GB202 GPU that will likely power the RTX 5090 will support as many as 16 memory modules -- the RTX 4090 has 12. A triple PCB is also a possibility according to the rumor, with separate boards for the GPU, IO, and VRM, but that's unlikely. In fact, the diagram in the post seems to show a riser board, that's probably used for testing purposes.

The GeForce RTX 5090's reported 512-bit memory bus would eclipse the current RTX 4090's 384-bit bus, and offer significantly more bandwidth. While the actual VRAM amount isn't confirmed, it may also be as large as 32GB. With 16 memory modules at 2GB each, however, 32GB of VRAM would be likely. 

The memory layout indicated for the RTX 5090 in the post is 4, 5, 2, 5, compared to 3,4,1,4 of the RTX 4090, with those numbers indicating how many chips reside in the board area around the four sides of the GPU.

The existing NVIDIA Founders Edition design

Gamers, creators, and budding AI engineers will all likely be pleased by the possibility of next-gen GeForces having larger memory capacities. There are obvious benefits for gamers that play at high resolutions, with the highest image quality settings and ray tracing. Content creators and 3D graphics artists working with ultra high-resolution models would also benefit. Machine learning and AI use cases would also leverage as much VRAM as is available. 

Other potential specs such as its power draw are not yet known, but rumors indicate that the GeForce RTX 5090's consumption may be in the 600 watt range. 

With such a dense VRAM layout to support massive amounts of memory, the GeForce RTX 5090 is shaping up to be a powerful successor to the RTX 4090. Its release should likely occur sometime before the end of 2024, or early in 2025, if current indicators are accurate. Pricing is also unknown at this time, but a $1599 MSRP similar to the GeForce RTX 4090 is likely.