A report published on September 13, 2026 describes a 12V-2x6 cable melting into the PSU-side modular socket of an ASRock Taichi TC-1650T while powering an NVIDIA RTX PRO 6000 Blackwell. Both ends were reportedly fully seated, and the GPU reportedly survived. The 1,650 W power supply includes ASRock TempGuard, a protection system designed to monitor connector heat and shut the system down before temperatures become dangerous.
The failure is a reported incident, not a confirmed product-wide defect. The technical explanation is also still a hypothesis: the proposed issue concerns where an external temperature sensor can detect heat, not proof that TempGuard caused the connector to melt.
What happened at the 12V-2x6 connection
The reported damage occurred at the power-supply end of the cable, inside the modular socket. That is distinct from the GPU-side receptacle. The cable used the 12V-2x6 implementation of the familiar 16-pin high-power connector format, and both ends were reportedly fully inserted before the failure.
The reported GPU was an NVIDIA RTX PRO 6000 Blackwell workstation card. Its reported total board-power figure is up to 600 W, delivered through a single PCIe CEM5 16-pin connection in the configuration described. The GPU itself was reported to have remained functional after the PSU-side failure.
That combination—high power, a compact connector and a damaged socket—is why the incident has drawn attention. A connector can develop a hot spot at one contact even when the surrounding cable does not appear uniformly hot. The local problem is resistance at the contact, not simply the average temperature of the cable jacket.
What TempGuard was meant to do
ASRock TempGuard uses an NTC temperature sensor. An NTC, or negative-temperature-coefficient thermistor, changes its electrical resistance as its temperature changes. The protection system is intended to detect unsafe connector heat and trigger a shutdown.
ASRock identifies the sensor at the VGA connector in the TC-1650T specification. The proposed explanation for this incident is that heat generated at an individual contact inside the connector may not reach an external sensor quickly or reliably enough. The PSU-side location of the reported melt makes that possibility especially relevant, but it does not prove that sensor placement caused this failure.
The distinction is important. A sensor can measure the temperature where it sits; it does not automatically measure the hottest metal contact elsewhere in the connection. Heat can spread through the connector body, cable insulation and surrounding air before reaching the probe. If one pin develops excessive resistance, the local temperature can rise without producing the same temperature rise at the sensor.
The technical alternative discussed for this type of problem is per-pin current monitoring. Measuring current through each contact can reveal uneven load sharing more directly than a single external temperature reading. That approach may offer a more direct protection strategy for this failure mode, but it is not a diagnosis of the reported unit.
The GPU and PSU in context
The two headline power figures describe different things:
- NVIDIA RTX PRO 6000 Blackwell: reported total board power of up to 600 W.
- ASRock Taichi TC-1650T: rated PSU capacity of 1,650 W.
- ASRock Taichi TC-1650T: described as supporting ATX 3.1 and PCIe 5.1, with two native 12V-2x6 cables.
- ASRock TempGuard: NTC-based connector-temperature protection intended to trigger a shutdown.
The 1,650 W rating is the power supply's overall capacity; it is not the amount consumed by the GPU. Likewise, the reported 600 W figure describes the graphics card's total board power, not the temperature or electrical load of one individual contact.
A separate earlier reported case
A different incident reported in March 2026 involved the same NVIDIA RTX PRO 6000 and ASRock Taichi TC-1650T combination. In that case, the cable was reportedly scorched at the GPU-side connection. PNY reportedly found the card functional, and ASRock reportedly replaced the cable.
That earlier GPU-side case and the September 13 PSU-side report should not be treated as one event. Their different failure locations also do not establish a common cause or a population-wide failure pattern.
What system builders should inspect
If you are building or maintaining a system with a high-power GPU, keep the basic checks boring and precise—boring is good when the alternative is melted plastic:
- Use the PSU manufacturer's compatible native cable. Do not substitute a modular cable from another power supply family.
- Seat both ends fully. The connector should be completely inserted at the GPU and PSU sides, without visible gaps.
- Avoid strain and tight bends near the connector. Give the cable enough clearance to prevent side-loading the plug or socket, and follow the specific cable manufacturer's bend guidance.
- Inspect the connection regularly. Discoloration, damaged plastic, scorching or deformation are reasons to power down and stop using the affected cable and hardware.
- Do not treat monitoring as a guarantee. A temperature sensor or current monitor can reduce risk, but neither turns a high-power connection into a zero-risk system.
For builders, the useful takeaway is straightforward: compatible cabling, complete insertion, sensible clearance and immediate attention to discoloration remain the first line of defense. A protection feature can help, but it cannot make a damaged connection safe.
