NVIDIA’s flagship GeForce RTX 5090 has become the latest high-end GPU to expose a troubling flaw in its power delivery system—not through excessive wattage, but through persistent connector failures even under conservative settings.

A user report from the Mobile01 forum details how the top row of pins on a Gigabyte Aorus Master Ice RTX 5090’s 12V-2×6 power connector melted despite the GPU being throttled to just 500 W and undervolted to 0.9 V—a 100 W and 0.205 V reduction from stock. The failure occurred on the GPU side of the connection, not the PSU, and was first noticed through repeated crashes during gaming sessions. Both the GPU and PSU connectors showed signs of damage, though the extent of harm might have been worse had the power limit remained higher.

This incident adds to growing evidence that the 12V-2×6 connector standard itself may be inherently unstable. Independent testing suggests the design struggles with inconsistent pin contact, leading to uneven current distribution and elevated temperatures. Unofficial adapter solutions appear to exacerbate these issues, further complicating an already fragile setup.

Not Just a Power Problem

The RTX 5090’s power connector failure contradicts the assumption that excessive wattage is the sole culprit behind these meltdowns. Even when artificially limited to 500 W—the same as many mid-range GPUs—the connector still degraded. This raises concerns about whether the 12V-2×6 standard can reliably handle even moderate workloads without mechanical stress.

While NVIDIA’s RTX 50-series GPUs remain in high demand, particularly for AI workloads, the persistent connector issues could force manufacturers to reconsider power delivery designs. Early reports already hint at potential adjustments, including a rumored $5,000 RTX 5090 aimed at professional markets, but physical reliability remains a critical hurdle.

Broader Implications for High-End GPUs

The problem isn’t isolated to NVIDIA. Some AMD GPUs using the same connector standard have also experienced similar failures, though NVIDIA’s high-end models—particularly the RTX 5090—have been the most visible cases. The RTX 5070 Ti and RTX 5060 Ti, while less powerful, may still face similar challenges if the underlying connector design flaws persist.

Industry speculation suggests NVIDIA could unveil RTX 50-series SUPER models at CES 2026, potentially addressing some of these issues. However, until then, users relying on high-end GPUs may need to monitor their power connectors closely—or prepare for unexpected downtime.

Key Specs

  • RTX 5090: Up to 500 W TDP (with extreme undervolting in this case).
  • RTX 5070 Ti: Standard 12V-2×6 connector, no confirmed power limits.
  • RTX 5060 Ti: 16 GB variant still in production.
  • Power Delivery: 12V-2×6 connector standard, prone to pin contact instability.
  • Reported Issue: Connector melting at 500 W (undervolted from stock).
  • Rumored Pricing: RTX 5090 potentially reaching $5,000 due to AI demand.
  • Event: CES 2026 as a possible launch window for SUPER variants.

For now, the RTX 5090’s connector failure serves as a stark reminder that even cutting-edge hardware isn’t immune to fundamental design limitations. Whether through official fixes or workarounds, the industry will need to address these issues before high-end GPUs become a reliability risk.