RTX 4090 Melting Connector Triggers Cat-Alerted Rescue

In Taiwan a PC owner posted that their Nvidia GeForce RTX 4090 began producing smoke and the smell of burning plastic while the owner was away briefly; their cat’s persistent meowing alerted them. The owner said the system used an FSP HPT2-1000M power supply and that the GPU cable was “directly connected without an adapter.” The owner could not shut the machine down normally and unplugged power; afterward the GPU’s power connector was effectively fused to the card.

Modern flagship GPUs like the RTX 4090 draw very high currents through the graphics-card power interface. The industry moved toward 16-pin (12VHPWR) power connectors to consolidate power delivery, but these pins and the cables must be sized, assembled, and seated correctly to avoid elevated contact resistance. Elevated resistance at the connector interface creates local heating, which can melt insulation, deform plastic housings, and in extreme cases produce smoke or fire.

Details from sources: The primary account (forum post reported by PCGamesN) names the PSU model (FSP HPT2-1000M) and describes the owner’s inability to remove the cable after melting. Several outlets referenced the 16-pin-style connector and noted that hardware monitoring and the power button did not prevent damage in similar incidents. The story is part of a growing set of reports implicating high-end GPUs (including RTX 4090 and successor-series cards, and isolated AMD reports) in connector-melt events when connected with certain cables, adapters, or under heavy sustained load.

This is a systems-integration and safety problem, not just a headline. ML engineers, workstation builders, datacenter operators, and hobbyists need to validate PSU compatibility, use verified cables and adapters from reputable vendors, ensure connectors seat cleanly, and consider continuous current profiles rather than peak draw when specifying cabling. Monitoring for connector temperature, implementing remote-shutdown automation, and better cable-management to avoid micro-bends or partial seating are practical mitigations.

Manufacturer advisories from Nvidia, PSU vendors, and motherboard makers; service bulletins on approved cable/adapter lists; and independent thermal/PEM tests of 12VHPWR/16-pin assemblies. If you manage workstations or racks with high-power GPUs, prioritize inspections and inventory of cable types and make model-specific deployment decisions.