Card Exteriors and Cooler Design

The NVIDIA GeForce RTX 5090 Founders Edition looks like a piece of jewellery—thanks to its full-metal cover on all sides and the smooth curves. The card measures 30.4 cm x 13.7 cm x 4 cm (LxHxW), and is no more than two slots thick. The card is just as long and tall as the RTX 4090 FE, but 50% thinner. It meets NVIDIA's SFF-ready standard of compact graphics cards that was introduced last year. If you've been through our Architecture Technical Deep Dive article, you'll know what a remarkable piece of engineering this card is.


There are three printed decals on the card, the first one is a debossed NVIDIA logo, the second one is a die-cut RTX 5090 model marker, and the third one is an LED-illuminated GeForce RTX logo on the top. The top and bottom sides have small cutouts for some additional air to flow through, to guide the overall flow of air (we confirmed this with NVIDIA).


The GeForce RTX 5090 Founders Edition debuts NVIDIA second generation of flow-through cooling solutions, this is called simply the NVIDIA Double Flow Through. The company chronicled the design evolution over the past generations in the slide above. In this new generation, the PCB of the card (the actual card) is located in the center, with separate PCBs connecting it to the display I/O and the PCIe interface. This centralization and compacting of the PCB allows the design to be the way it is—make room for so much airflow.


If you think about it, the cooling solution is functionally similar to a dual fin-stack CPU cooler, except that both fin-stacks are arranged along the plane of the PCB instead of being perpendicular to it. At the center of the cooler is a large unified vapor chamber plate that makes contact with the massive NVIDIA GB202 GPU, the sixteen GDDR7 memory chips that surround it, and the DrMOS components of the card's 27-phase VRM solution. This vapor chamber plate has soldered contact with five 8 mm-thick copper heat pipes, which skewer the two fin-stacks.

In the picture above you can see that the heatsink fins above the fans are slightly curved, this is done to improve pressure drop and optimize airflow.


Since the main PCB is toward the center of the card, it relies on two breakaway components, the first one is a PCIe interface riser, which has the card's PCI-Express 5.0 x16 gold fingers. If you notice the lack of any SMDs above these gold fingers, it's because this breakaway PCB does nothing except turn this PCIe connection to a ribbon cable that goes to the "main" card. This ribbon cable is routed along the edges of the cooler's frame, so it doesn't get in the way of the airflow.


The second breakaway component deals with the display I/O, it is arranged along the rear I/O shield, and has the card's four display outputs. You'll notice here that the connectors are upside down compared to other cards. The connectors provided are three DisplayPort 2.1b, each capable of UHBR20, and one HDMI 2.1b.


The card's 12V2x6 power connector is on the main PCB. It doesn't point directly up, but rather at an angle, and is recessed into the card. This design helps reduce the bending of the cable when connected, and the resulting mechanical strain on the connector and its contacts. It's especially useful when you use high-gauge native 12V2x6 ATX 3.1 cables from some of the newer PSUs.


Compare this design to the previous generation RTX 4090 FE, and you'll see where NVIDIA is coming from—on the older card, the connector is pushed completely to the top, and points straight up, so your power cable has to make a steep bend toward it, which can be problematic on some of the narrower mid-tower cases.


This comparison highlights NVIDIA's success in significantly compacting their 50-series flagship over the RTX 4090 FE.