NVIDIA RTX 5090 Founder’s Edition To Feature 16 GDDR7 Memory Modules

The memory configuration for the forthcoming RTX 5090 graphics card will be different from that of the RTX 4090. The new top-tier GPU, equipped with the Blackwell GB202 chip, is speculated to have a 512-bit memory bus. This change is expected to boost memory capacity, necessitating additional modules around the GPU.

The RTX 5090 is anticipated to support GDDR7 memory, arranged in a denser configuration compared to the current RTX 4090. The layout will change from 3-4-1-4 to 4-5-2-5 (clockwise), enabling the board to accommodate up to 16 memory modules. However, this doesn’t mean that all modules will be utilized or specifies their capacities. If the RTX 5090 uses the same 2GB modules, it could potentially provide up to 32GB of memory.


The rumor also suggests that the design will include three PCBs. The exact meaning of this is unclear, but it is noteworthy that NVIDIA has previously experimented with unconventional PCB designs, such as for the unreleased RTX TITAN ADA GPU. In that instance, the PCB was intended to be positioned parallel to the PCI slot, with the PCIe interface located on a separate PCB.

Although the actual board was never photographed, the prototype cooler designed for this configuration has been widely leaked. According to these leaks, a quad-slot, triple-fan cooler would be used with this PCB setup.


NVIDIA is anticipated to reveal its RTX 50 series this year, likely in the fourth quarter. Usually, leaks become more frequent between June and August prior to a new series launch. This increase occurs as NVIDIA collaborates with more partners during this time, some of whom are already developing prototype coolers. These prototype coolers are often the focus of leaks.

Source: Chiphell
Prototype Image:Hayaka