ZOTAC RTX 5090 SOLID Catches Fire Near the PCIe Slot, Not the Cable

A ZOTAC GeForce RTX 5090 SOLID caught fire five minutes into a new Assassin’s Creed session this month, and for once the wreckage wasn’t at the power connector. The card popped, crackled and filled the case with smoke, but the scorch marks sat near the PCIe slot at the bottom of the board, not the 16-pin cable that has dominated RTX 5090 failure reports for a year and a half.

That distinction matters more than it looks. It’s the second time in under a year that a ZOTAC RTX 5090 has failed this exact way, away from the cable everyone’s been told to watch.

Smoke Rises From the PCIe Slot, Not the Cable

The owner had been playing Assassin’s Creed Black Flag Resynced for about five minutes when the card let out a loud pop, followed by crackling and a thick cloud of smoke. The system shut down immediately.

The graphics card had run without issue for roughly a year before the failure. Photos posted afterward showed dark burn residue on the card itself and around the motherboard’s main PCIe slot, paired with an MSI X870E Tomahawk board. The 12V-2×6 power connector, the 16-pin plug that succeeded the widely criticized 12VHPWR standard, came through untouched.

The owner had recently installed NVIDIA’s GeForce 610.74 driver, released July 7 with day-one support for the new Assassin’s Creed title. There’s no evidence tying the driver to the failure, and the card’s retailer has an RMA underway while the cause gets investigated.

• Confirmed: the card ran about a year before failing, damage clustered near the PCIe slot, and the power connector was undamaged.
• Confirmed: the system paired an MSI X870E Tomahawk motherboard with driver 610.74, installed days before the failure.
• Unconfirmed: whether a short circuit, a cracked trace from board flex, or something else in the power delivery path triggered the fire.
• Unconfirmed: whether ZOTAC or NVIDIA will issue any statement; neither company had commented publicly as of this week.

One working theory floating around the report ties the damage to physical stress on the board itself, a cracked trace or short caused by the card sagging under its own weight rather than anything electrical in origin.

Why Didn’t the Power Connector Take the Blame This Time?

Because for a year and a half, the connector has been the story. NVIDIA’s RTX 5090 draws up to 575 watts, a jump of 125 watts over the RTX 4090’s 450, and nearly every dramatic failure since launch has traced back to the same 16-pin plug overheating, melting or scorching under that load.

NVIDIA said the problem was handled before the RTX 5090 even shipped. At the company’s RTX AI Day 2025 event in South Korea, an executive addressed the RTX 4090’s connector meltdowns directly.

We don’t expect that to happen with the RTX 50 series.

Nvidia Korea senior vice president Sunwook Kim said that during a Q&A session at the event, adding that changes made to the connector two years earlier had resolved the issue. Reports since then say otherwise. Club386 its own RTX 5090 Founders Edition melted during testing, using a fully seated, single native cable on a 1,000-watt power supply, and the outlet called the connector “not fit for purpose.”

This year’s Computex GeForce showcase leaned on DLSS 4.5 rather than new silicon, which means no redesigned power stage is coming soon for cards already in the field. Against that backdrop, a fire that skips the connector entirely and hits the board instead is a different problem than the one NVIDIA said it fixed.

A Nearly Identical ZOTAC Fire Ten Months Earlier

This isn’t the first time a ZOTAC RTX 5090 has burned somewhere other than the cable. In August 2025, a ZOTAC GeForce RTX 5090 Amp Extreme Infinity caught fire mid-session during the Battlefield 6 open beta. The user, posting under the name york4517 on the Gamer-TW forum, said the screen froze partway through the game, then flames appeared from the side of the desk and the room smelled burnt.

Hardware analyst Uniko traced that fire to the MSVDD rail, the memory subsystem’s power line, suspecting a MOSFET short or phase failure in a section of the board with no dedicated cooling. The 12V connector wasn’t implicated. The system was a prebuilt unit, and the shop was left coordinating with the distributor and ZOTAC’s local agent to inspect it.

Hardware analyst Igor Wallossek dug into the board layout afterward and found something specific to ZOTAC’s design. Overlaying ZOTAC’s PCB against NVIDIA’s reference layout and a rival MSI board, he measured just 3.84mm of clearance near a DrMOS component on the ZOTAC card, compared with 5.87mm on the MSI equivalent. NVIDIA’s reference design sets the electrical and thermal spec, not the mechanical layout, leaving each board partner to work out clearance issues on its own.

Two ZOTAC RTX 5090 cards, two different games, eleven months apart, and the same basic shape: a loud failure that has nothing to do with the plug the entire industry has spent a year fortifying.

AIBs Kept Fortifying the Cable

ZOTAC, which used this year’s Computex to talk up pushing enterprise AI hardware alongside its 20th anniversary, still sells the SOLID as a mainstream consumer flagship. Across the RTX 50 lineup, board partners spent the last year and a half adding safeguards, all aimed at the same connector.

• ZOTAC added a Safety Light feature that blocks the GPU from powering on until the cable is fully and securely seated.
• MSI introduced a yellow-tipped 12V-2×6 adapter, so any visible yellow after installation signals an incomplete connection.
• ASRock built a TempGuard cable designed to cut power if the connector overheats, though a recent case saw it fail to trigger before a connector melted anyway.
• ASUS added sensors to Blackwell-generation cards meant to flag trouble on individual pins.

Every one of those fixes targets the cable. None of them would have done anything for a fire that starts near a MOSFET or a PCIe slot instead. A separate community project illustrates why pin-level monitoring is so hard to retrofit: reverse-engineering the connector’s internals, one hobbyist found the design ties all six 12V pins to a single shared shunt resistor, meaning a single pin could carry 42 amps in a worst-case scenario, more than four times its rated limit, with no way for the GPU to know which pin is overloaded.

ASRock has already floated an alternative. The company showed a prototype board hinting at a connector shift away from the current layout, though nothing like it is shipping yet.

What RTX 5090 Owners Can Actually Check

ZOTAC recommends at least a 1,000-watt power supply for every RTX 5090 card it sells, and independent guides suggest 1,250 watts or higher for real headroom. Owners running below that spec are already outside the vendor’s own guidance before anything else goes wrong.

The connector checklist is well worn by now: seat the cable until it clicks, avoid tight bends right at the plug, skip cheap unbranded adapters, and recheck after the first few gaming sessions. One guide to the ongoing problem notes that CableMod recalled 25,300 angled adapters through the Consumer Product Safety Commission after 272 damage reports, a reminder that the cable itself is still a live risk even as this new incident points elsewhere.

What this SOLID fire adds is a second thing to check: the board itself, not just the plug. A heavy triple-fan cooler hanging off a vertical PCIe slot for a year can flex a board enough to stress solder joints and traces near the connector edge, which is why board partners increasingly ship support brackets with their flagship cards. Using one, and periodically glancing at the PCIe slot area rather than only the cable, costs nothing and takes seconds.

Frequently Asked Questions

Is This the Same 12V-2×6 Connector Problem That’s Plagued the RTX 5090?

No. The connector on this ZOTAC RTX 5090 SOLID stayed undamaged; the burn marks sit near the PCIe slot instead. That’s a different failure signature from ZOTAC’s own November 2025 case, where a SOLID card’s 12V-2×6 connector melted completely after being paired with an 850-watt power supply, below the 1,000-watt minimum ZOTAC recommends. Both are ZOTAC RTX 5090 failures, but they point to different weak spots on the same card.

What Caused the ZOTAC RTX 5090 SOLID to Catch Fire?

The exact cause hasn’t been confirmed. Reports point to a possible short circuit or a cracked board trace, potentially linked to the card sagging under the weight of its cooler over a year of use, but an RMA and inspection are still working through the details.

Has This Happened to a ZOTAC RTX 5090 Before?

Yes, in August 2025 a ZOTAC RTX 5090 Amp Extreme Infinity caught fire during the Battlefield 6 open beta with damage near the memory power rail, not the cable. A forensic board comparison afterward found ZOTAC followed NVIDIA’s reference design closely for outer dimensions, a detail the analyst said deserved credit, but that reference design says nothing about component clearance, which is left to each board partner.

Should RTX 5090 Owners Use a GPU Support Bracket?

Yes, and it addresses a different risk than the cable does. RTX 5090 coolers are large and heavy, and mounted vertically in a PCIe slot for months at a time, that weight can flex the board and stress solder joints near the slot edge, exactly where both ZOTAC fires showed damage.

Will NVIDIA or ZOTAC Replace a Card Damaged Like This?

That’s still unresolved for this case. The SOLID owner’s retailer has an RMA in progress, but as of this week neither ZOTAC nor NVIDIA had issued a public statement on the incident. The August 2025 case ended with a distributor and ZOTAC’s local agent inspecting a prebuilt system rather than a clear public resolution.