Nvidia has recently unveiled its GeForce RTX 5090 Founders Edition graphics card, which boasts the title of being the highest-performing consumer GPU available. However, in conjunction with its remarkable performance, it also represents one of the most power-demanding graphics cards to date. To accommodate this, Nvidia introduced a completely new three-piece printed circuit board (PCB) design, which is unlike any previous iteration.
The new design includes a main PCB integrated with the GB202 chip, an impressive 92 billion transistors, alongside 32 GB of GDDR7 memory connected via a robust 512-bit interface. During Nvidia’s CES 2025 presentation, this ‘high-density’ PCB was showcased, highlighting its small size in comparison to the larger GeForce RTX 4090 card. Despite having a copious amount of processing power, the main PCB is notably compact—a fact underscored by the size of its memory chips.
In terms of power requirements, the card employs a single 12V-2×6 power connector capable of delivering 55A of power, catering to a total board power (TBP) of 575W. This design also includes a formidable 30-phase voltage regulating module (VRM), which is an upgrade from the previous 23-phase system used in the GeForce RTX 4090. A well-designed power delivery system ensures stable performance and efficiency, especially under demanding conditions, though it raises questions about potential overclocking capabilities.
While the main PCB lacks display outputs, it may rely on two additional boards for power connectors and PCIe grids that have yet to be displayed. This disaggregation could serve multiple purposes, such as enhancing signal routing and optimizing space while also minimizing electrical noise that might disrupt the GPU’s operation.
Nvidia’s innovative approach is reminiscent of its earlier designs, such as the GeForce 8800 GTX/Ultra, allowing for a more tailored GPU architecture. This strategic design choice may also enable Nvidia to offer varied configurations for different use cases, particularly for OEM needs.
Notably, while Nvidia highlights the unique three-piece PCB, its partners are still expected to utilize traditional single-board designs for the GeForce RTX 5090 graphics cards. This distinction signifies Nvidia’s intention to explore diverse technological designs while accommodating existing manufacturing standards.
This leap in GPU design and performance is a testament to Nvidia’s ongoing commitment to pushing the boundaries of gaming technology. As the capabilities of graphics processing continue to evolve, consumers can look forward to exceptional gaming experiences and innovative features in the coming years.