Abstract:
NVIDIA GeForce RTX 5090, as a new generation flagship graphics card, has extremely high performance and power consumption, and its high power consumption is accompanied by the 12VHPWR and subsequent 12V-2x6 power supply interface that have been controversial in recent years. Nowadays, a group of modders have modified the graphics card power supply design, allowing the RTX 5090 to successfully get rid of the original 16-pin power supply interface and switch to the traditional 8-pin PCIe power connector.
RTX 5090 is designed to consume up to 575W, which requires the graphics card to obtain sufficient power through a specially designed high-power power supply interface. The 12VHPWR interface launched by NVIDIA in the RTX 40 series era was originally prepared for this kind of high-power graphics card, but in actual use, there have been problems with the interface and cables overheating, melting, and even burning.
Although NVIDIA subsequently made improvements through the 12V-2x6 interface, including redesigning the contacts and detection mechanism, and improving reliability to a certain extent, the 16-pin high-power interface still requires users to pay extra attention to whether the plug is fully inserted, whether the cable is excessively bent, and whether the connection is secure. For some enthusiasts, although the traditional 8-pin PCIe interface can provide limited power, it has a more mature history of use. Therefore, they have been hoping to find a way to reuse the 8-pin interface for high-power graphics cards.
This modification is centered around this idea.
Modding players first need to completely modify the power supply circuit of the RTX 5090, because the original power input design of the graphics card is built around the 16-pin interface and high-power power supply path. Simply making a "16-pin to 8-pin" adapter cable won't solve the problem, because if you still allow a lot of power to be transmitted through a single traditional 8-pin interface, it will seriously exceed its design capacity.
Therefore, the modification solution is actually to redistribute the input power required by the graphics card, spreading the power supply demand that was originally concentrated on a 16-pin interface to multiple 8-pin PCIe connectors. Doing so can utilize multiple independent connectors and corresponding cables to share the huge power consumption of the RTX 5090.
From an electrical point of view, this method does not suddenly give the 8-pin interface a power supply capacity of 575W, but reduces the current that each connector and cable can withstand by increasing the number of interfaces. This is theoretically possible as long as the power paths, connectors, and copper and power components on the PCB can withstand the load.
After the modification, the RTX 5090 no longer relies on the original single 16-pin high-power interface, but can obtain power through multiple traditional 8-pin interfaces. This design looks more "bulky", but for power users who already have multiple traditional PCIe power supply cables, it has a very obvious advantage, that is, there is no need to rely on the controversial 16-pin high-power connector.
Of course, this modification does not mean that the traditional 8-pin interface itself is more suitable for carrying the full power consumption of the RTX 5090 than the 16-pin interface. What really matters is how the entire power supply system distributes current. If only one or even two 8-pin interfaces are used to force the full power of the RTX 5090, it may also cause overheating of connectors, cables and even power terminals.
The traditional 8-pin PCIe interface is usually designed to provide a rated graphics card power supply capacity of about 150W, and the graphics card can also obtain additional power from the PCIe slot. Therefore, for an RTX 5090 with a power consumption of 575W, if you rely entirely on traditional interfaces, you need to use multiple independent 8-pin connectors.
This is why this modification did not use a simple adapter, but directly modified the power supply structure of the graphics card. Only by truly connecting multiple power supply inputs to the power supply system of the graphics card and reasonably allocating the current can the power supply needs of the flagship GPU be met without overloading a single interface.
This solution also exposes a real problem faced by modern high-end graphics card power supply design: as GPU power consumption continues to increase, traditional PC power connectors are gradually approaching the boundaries of their capabilities. The RTX 5090's 575W power consumption is already far beyond the level of mainstream graphics cards in the past, and higher-end overclocked models may even exceed this number further.
The core reason why NVIDIA uses a 16-pin interface is to provide higher power density within a limited connector volume. Compared with using four or more 8-pin power cables, a 16-pin connector can make the graphics card look cleaner and reduce the number of cables inside the case.
The problem is that the higher the power density, the higher the requirements for connector contact quality and installation accuracy. Once the plug is not fully inserted, or there is abnormal resistance at the contact point, a large amount of heat may be generated under high current. Melting cases that have appeared for a long time have also made many high-end graphics card users have concerns about this interface.
This modification proves in a sense that the RTX 5090 does not have to rely on a single 16-pin interface to work properly. As long as the power supply path is redesigned and a sufficient number of traditional PCIe power supply connectors are provided, the power needs of the GPU can be met.
However, this solution is obviously not suitable for ordinary users to try on their own. The power supply system of the RTX 5090 involves the GPU core, power management chip, MOSFET, inductor and multi-layer power supply network on the PCB. Improper handling of any link may cause the graphics card to fail to start, or even cause permanent damage.
In addition, the modified graphics card also needs to face a very real problem, that is, power supply capacity. Even if multiple 8-pin interfaces are used on the graphics card side, the entire system still requires a high-quality power supply that can stably provide hundreds of watts of GPU power consumption. Multiple 8-pin connectors only change the way power enters the graphics card, and will not reduce the actual power consumption of the RTX 5090 itself.
If further overclocking is performed, the situation will become more complicated. The RTX 5090 can consume much higher power than the default 575W after the power consumption limit is lifted, which means that the power supply interface, cables, PCB and the power supply itself need to bear a higher load. Therefore, the modified 8-pin solution cannot be understood as a method that can infinitely increase the power supply capability.
For ordinary RTX 5090 users, the most realistic option is still to use a compliant power supply, original or high-quality 12V-2x6 cables, and ensure that the connector is fully inserted into the graphics card. Modern high-power interfaces will still work under normal design and usage conditions as long as the power supply connections are correct and reliable.
But from the perspective of hardware modification, this experiment is very interesting. It proves that even in the face of 575W-level flagship GPUs, as long as you are willing to sacrifice the number of connectors and increase cables and PCB power supply paths, you can re-adopt the traditional 8-pin interface system.
For enthusiasts who have always had doubts about the reliability of the 16-pin high-power interface, this undoubtedly provides an alternative idea: instead of concentrating more and more power on a small connector, it is better to use more mature connectors to share the load.
Of course, the price is that more power cords will appear behind the graphics card, the space occupied inside the chassis will also increase significantly, and the overall aesthetics and wiring convenience will decrease. In contrast, the biggest advantage of the 16-pin interface is precisely its ability to transmit extremely high power in a very compact connector.
As such, this modification is more of a technology demonstration than an actual replacement for NVIDIA's existing power supply design. What it proves is not that "the 8-pin interface is more advanced than the 16-pin interface", but that there are multiple solutions to the power supply problem of high-power GPUs. For a 575W-class flagship graphics card like the RTX 5090, what is really important is always to distribute the current reliably, evenly and safely, and whether to use a 16-pin interface or multiple 8-pin interfaces is just a different engineering solution to achieve this goal.
Comments