A new power connection problem has emerged with NVIDIA GeForce RTX 5090 graphics card models. According to user reports, the MSI GeForce RTX 5090 Gaming Trio OC model overheated during DLSS 5 tests, causing the cables to melt.

This technical malfunction has reignited discussions about the power management of high-performance graphics cards. This noteworthy event in the hardware world particularly highlights the burden that new generation AI-based processing technologies place on hardware.

High Power Consumption During DLSS 5 Tests

Tests conducted on the MSI GeForce RTX 5090 Gaming Trio OC model showed that the card exceeded its factory specifications. When DLSS 5 Neural Rendering technology was enabled in NBA 2K27, the card’s power consumption reached 613.5 Watts.

This value is considerably above the 575 Watt factory limit set by MSI. Consumption, expected to be around 450 Watts in standard use, increased significantly with DLSS 5.

In the incident, which resulted in the system suddenly shutting down, the user reported smelling a sharp burning odor. Physical examination revealed that the plastic casing of the 16-pin 12V-2×6 power connector had melted and fused with the socket terminals.

The user stated that they had great difficulty separating the melted cable from the socket. This situation proves that the connector is constantly pushing its limits in terms of high power flow.

Technical Analyzes and Future Expectations

Previous technical analyzes predicted that the GeForce RTX 5090 power connectors might limit DLSS 5 performance. Dual-connector models are known to offer much higher power stability compared to single-connector designs.

Independent tests conducted on the HardForum platform also support these results. In games like Hitman 3, it was observed that the power drawn by the card increased from 435 Watts to 520 Watts when the DLSS 5 code was activated.

The intensive neural process calculations brought about by the DLSS 5 architecture constantly put a load on the tensor cores of the graphics processor. This causes the memory subsystem to demand more power than usual.

Manufacturers are expected to develop new strategies to meet these high power demands. Finalizing power profiles or redesigning cooling solutions via software is critical to preventing these types of issues.

We are eagerly awaiting to see what kind of update hardware manufacturers will release to address this high power consumption problem. Do you think more reliable new standards should be developed for power connections in high-performance graphics cards?