TecLab soldered power cables directly to the PCB of a specialized Galax HOF OC LAB XOC RTX 5090D, bypassed both factory 16-pin connectors, and got the card to boot and run benchmarks. The striking result is real; the idea that it is a safe consumer repair is not supported by the evidence.
The original TecLab demonstration shows the soldered board, the RTX 5090D identification in GPU-Z, the boot sequence, benchmark operation, a clamp meter, and an unusual cooling setup. It was published on August 26, 2026. This is direct evidence that one modified sample operated—not a certification, repair guide, or endurance test.
The RTX 5090D experiment
The card was not a standard retail RTX 5090. It was the Galax HOF OC LAB XOC RTX 5090D, an extreme-overclocking model reported to have two factory 16-pin connectors. TecLab used adapters to turn standard 16-pin PSU cables into bare wires, then attached two sets of power wires to PCB areas connected to the original connector traces.
That distinction matters. A standard RTX 5090 generally uses one 16-pin connector, while the HOF test card started with two. Comparing the two without naming the models would be like comparing a dragster with a family sedan and then blaming the sedan for not having a parachute.
| Configuration | Connectors | Power path | Cooling context |
| Standard RTX 5090 | Generally one factory 16-pin connector | Factory connector and cable system | Standard retail cooling was not evaluated in this test |
| Galax HOF OC LAB XOC RTX 5090D | Two factory 16-pin connectors | Factory arrangement before modification | Specialized overclocking hardware |
| TecLab experiment | Factory connectors bypassed | Wires soldered to PCB traces above the connector areas | Stock cooler removed; industrial liquid chiller and two 120 mm fans reported |
The close-up makes the central change unmistakable: the detachable GPU-side connector is no longer the direct connection point. The wires go straight to the board. TecLab presenter Ronaldo framed the modification as criticism of NVIDIA’s connector design:
“Nvidia, learn from this and put in a real connector so there are no more complaints.” — Ronaldo, TecLab. English translation of Portuguese speech in the August 26, 2026 demonstration.
A reported 600 W test under extreme cooling
Tom’s Hardware reported approximately 600 W of consumption, based on an approximately 50 A reading on the 12 V rail. The figure is a reported measurement from specialist coverage, not an independently replicated result. The supplied visual evidence shows a meter and the test environment, but it does not independently verify a readable 600 W value.
The same report says the standard RTX 5090 has a reported 575 W TDP. It also reports that the HOF model supports overclocking BIOS configurations of up to 2,000 W, although that capability was not independently verified in the supplied material. Those numbers describe different things: the demonstration’s approximate draw, a reported standard-card TDP, and a reported extreme BIOS capability. They should not be treated as interchangeable.
The conditions were anything but normal. Tom’s Hardware reported that TecLab removed the stock heatsink and fans, using an industrial liquid chiller and two loosely attached 120 mm fans. The exact coolant description differs between reports, so it should not be presented more precisely than the available evidence allows.
Why bypassing the connector is tempting
The experiment responds to continuing concern about overheating or melting high-current 16-pin GPU connectors, including the 12V-2x6 standard discussed in coverage of RTX 5090-class cards. A connector adds detachable contacts and mechanical interfaces to a high-current power path. In principle, removing the GPU-side plug changes that path. But the experiment does not show that the replacement is safer under sustained consumer use.
Another response is monitoring rather than rewiring. The WireView Pro II Wired video from der8auer EN discusses per-pin current monitoring for the RTX 5090 Founders Edition. It is adjacent context, not evidence that monitoring prevents every failure and not a replacement for the TecLab demonstration.
A working boot is not a safety certification
Yes, the modified RTX 5090D booted and completed benchmark testing. No, that does not prove the modification is safe. The available material contains no controlled failure-rate study, long-term endurance test, validated temperature measurement for the solder joints, or independent engineering analysis comparing the soldered arrangement with the factory connector.
| Claim | Status | What the evidence supports |
| TecLab soldered power cables directly to a Galax HOF RTX 5090D PCB | Confirmed demonstration | The original TecLab video visibly documents the modification |
| The modified card booted and ran benchmarks | Confirmed on video | Booting and benchmark operation are visible in the demonstration |
| The test reached approximately 600 W and 50 A | Reported measurement | Tom’s Hardware attributes the figures to a meter reading; no independent replication is supplied |
| Direct soldering is safer over the long term | Unproven | No controlled thermal, electrical, or endurance comparison is available |
| The modification is a practical repair for owners | Not supported | Specialist coverage identifies warranty, short-circuit, mechanical, thermal-cycle, and serviceability risks |
What normal RTX 5090 owners should do
Do not reproduce this modification at home. Directly soldering high-current wires to a graphics-card PCB can leave exposed joints and conductors, add mechanical strain, create short-circuit hazards, complicate servicing, and likely affect warranty coverage. A short benchmark run cannot answer how those joints behave through sustained load and temperature cycling.
Use the power equipment and connection method appropriate to your specific graphics card and power supply. If you are concerned about a high-current connector, monitoring hardware may offer visibility into the power path, but the supplied evidence does not establish that any monitoring device protects against every failure.
The bottom line is simple: TecLab showed that a specially configured RTX 5090D can operate after its connectors are bypassed. It did not show that direct PCB soldering is safe, durable, approved, or suitable for ordinary owners. The next useful evidence would be independent electrical testing, long-duration results, and a documented response from NVIDIA or GALAX—none of which is included in the supplied material.
