Fresh measurements published today by Chinese tech outlet 快科技 (MyDrivers) have put hard numbers on a long-suspected weakness in NVIDIA’s GeForce RTX 5090 Founders Edition reference power cable: under sustained gaming load, the card’s bundled four-8-pin-to-16-pin adapter distributes current so unevenly across its six 12V lines that individual wires carry up to 11.73 amperes — and the worst-case gap between the two wires inside a single plug can reach 4.52A. The 12V-2×6 connector standard rates each pin at a maximum of 9.5A. At 11.73A, one plug’s leading wire is already running 23 percent above that limit. What makes today’s finding particularly significant for the tens of thousands of RTX 5090 FE owners who have already applied a power cap as a precaution: the test was conducted at 85 percent of the card’s rated 575-watt maximum — a widely recommended mitigation level — and the spec exceedance appeared anyway.

Test Setup and What the Numbers Say

The measurements came from a GeForce RTX 5090 FE owner whose system paired an AMD Ryzen 9 7900X processor with 32 GB of DDR5 memory, a Corsair RM1200e power supply, and a Samsung Odyssey G9 display — a broadly representative high-end gaming build. The workload was Cyberpunk 2077 on Ultra settings with ray tracing and DLSS enabled, a sustained maximum-effort scenario that keeps the GPU’s power delivery under continuous stress. Power draw was capped at 85 percent of rated maximum through the NVIDIA App, placing peak draw at roughly 488 watts. The tester used a clamp ammeter on each of the 16-gauge conductors inside the adapter’s four 8-pin plugs and recorded the following:

The inter-connector balance across the four plugs was broadly even — the problem is not that one plug starves the others, but that within each plug the two 12V wires diverge sharply. Plug C’s higher-loaded wire carries 64 percent more current than its companion. Plug A’s lead wire, at 11.73A, exceeds the connector’s rated current limit by approximately 23 percent. This happened with the power cap applied.

Why Power Limiting Does Not Protect Adapter Users

This is the finding the existing mitigation literature has not addressed directly. The recommendation to run the RTX 5090 at 80–85 percent power to reduce cable stress has been widely repeated since the first connector melt reports emerged in February 2025. It is not wrong — reducing aggregate current does reduce the absolute amperage on each wire. But it cannot close the ratio between the wires, because the ratio is not set by total power. It is set by differential contact resistance.

Each 8-pin connector junction in the reference adapter introduces a discrete set of contact points. A four-8-pin-to-16-pin adapter has four additional sets of contacts between the PSU and the GPU’s 12V-2×6 socket. Any variation in contact quality, spring tension, surface oxidation, or manufacturing tolerance across those junctions creates different resistance values along the six 12V paths. By Ohm’s law, current distributes inversely to resistance: the lower-resistance path carries more current, and the higher-resistance path carries less, regardless of how much total current is flowing. Reduce total power by 15 percent and every wire’s absolute amperage drops by approximately 15 percent — but the ratio between them does not change.

If Plug C shows a 4.52A gap at 85 percent load (488W), the same contact geometry at full 575W load would produce approximately 5.3A of intra-plug imbalance. At 70 percent load it would be approximately 3.9A. The imbalance never reaches zero without fixing the contact resistance — which means either using a different cable or replacing the adapter entirely. Power limiting buys some headroom on absolute wire temperature but cannot protect an unevenly loaded wire from exceeding its rated current.

This physical constraint is why the RTX 5090 Founders Edition’s monitoring architecture compounds the problem rather than compensating for it. The FE reference board uses a single shunt resistor to measure aggregate incoming power. A shunt resistor works by measuring the voltage drop across a known resistance: the GPU reads total current from the combined 12V rail. That measurement cannot distinguish between six wires each carrying 7A and a situation where two wires carry 11.73A apiece while four others share the remainder. From the card’s firmware, both scenarios look identical.

What the RTX 50 Generation Changed — and What ASUS Did Differently

This design represents a generational regression in per-wire safety coverage. NVIDIA’s RTX 30-series cards treated the 12-pin connector as three independent 12V sources, with separate current monitoring on each group. The RTX 40-series reduced this to two shunt resistors, both feeding the same measurement block — no independent per-path reading. The RTX 5090 Founders Edition drops to a single shunt resistor, the most limited monitoring of any high-end NVIDIA card in the connector standard’s history.

ASUS took a different approach on its ROG Astral RTX 5090. That card carries an ITE IT8915FN monitoring chip with per-pin shunt resistors for monitoring, allowing ASUS’s GPU Tweak III Power Detector+ software to display per-pin current in real time and raise a warning at 9.2A. An open-source community utility called 12VHPWR Guard reads per-pin current and shuts down the system if any pin sustains more than 9.5A for longer than 15 seconds.

As hardware power analyst Aris Mpitziopoulos of Hardware Busters has noted in Hardware Busters connector analysis: on every other RTX 50 card, per-wire imbalance is either present or it isn’t, and no system-level software is watching for it. That assessment applies with particular force to any user running the reference adapter cable, where imbalance of the magnitude documented today exists alongside no detection mechanism at all.

How This Connects to a Pattern Four Years Running

Today’s measurement data is the latest chapter in an ongoing connector reliability story that began with the RTX 4090’s original 12VHPWR connector in October 2022. Overclocker and Thermal Grizzly CEO Roman “der8auer” Hartung documented the first detailed physics of the failure in February 2025, finding that a single 12V wire in a thermally stressed RTX 5090 FE carried over 22 amperes — roughly four times its nominal share — at temperatures exceeding 150°C (302°F). Analysis by Igor’s Lab found that even freshly seated, quality native cables could show imbalances at short measurement intervals.

PCI-SIG revised the connector standard in mid-2023, introducing the 12V-2×6 format with longer sense pins mechanically preventing the card from operating unless the connector is fully seated. The RTX 5090 ships with the 12V-2×6 socket. But as the accumulated record of documented melt incidents throughout 2025 and 2026 demonstrates — and as today’s measurements make quantitatively plain — the revised standard addresses improper seating without eliminating the underlying current asymmetry problem in adapter cables.

September 2026 added two additional data points to this record. An ASUS ROG Astral RTX 5090 on Hardware Unboxed’s test bench reached 175°C at the connector while paired with a Be Quiet! Dark Power 13 power supply — the only indication something was wrong was the heat Steve felt when touching the cable. Within the same week, PNY reportedly denied an RTX 5090 warranty claim connected to a prebuilt system, citing native PSU cables as not covered under warranty terms. As of this week, a community-built distributor module by Reddit user DallasGrave — a PCB that splits the GPU’s single 16-pin 12V-2×6 input into two separate 8-pin connectors — held connector temperature to roughly 40°C at 550W across a 48-hour stress test, versus the 91.7°C sense-pin temperature that independent lab testing recorded for a stock connector running Cyberpunk 2077 at 4K with DLSS 5 enabled. That gap — 40°C versus 92°C — is the thermal cost of concentrating 575 watts through a single connector rated at 600 watts with no physical load-distribution across multiple paths.

Does a Power Cap Make the Bundled Adapter Safe?

What today’s clamp test answers, specifically, is the question many RTX 5090 FE adapter users have been asking since the first melt reports: does reducing the power limit to 85 percent put the card out of danger on the reference cable? The answer from the data is: no, not on the adapter cable, and not because 85 percent is the wrong target but because the adapter’s contact resistance distribution remains unchanged at any power level.

Independent testing with Thermal Grizzly’s WireView Pro tool has found that high-quality native 12V-2×6 cables from ATX 3.1-compliant power supplies keep pin-to-pin current imbalance within approximately 0.2 amperes under similar loads. Today’s 4.52A worst-case gap is roughly 22 times larger. That difference in imbalance magnitude is not a difference in gaming behavior or power draw profile — it is a difference in cable path: the adapter’s four additional connector junctions versus a single direct cable with one pair of contacts.

The 12V-2×6 connector standard’s theoretical maximum is 600 watts delivered across six 12V terminals, each rated at 9.5A. NVIDIA’s RTX 5090 carries a 575-watt power ceiling, placing it within 25 watts of that ceiling before overhead factors are applied. When the load distributes unevenly, the margin collapses on the wires that carry the larger share. At 11.73A on a nominally 9.5A-rated pin, the wire itself is within its 16 AWG current-carrying capacity, but the connector contact geometry the standard specifies is not — and the temperature at that contact point rises with the square of current (I² × R heating), while no onboard circuit registers the difference.

How to Think About What This Means If You Own an RTX 5090 FE

The community understanding of RTX 5090 adapter cable risk, synthesized from the measurements accumulated since launch, points to a clear hierarchy:

Native 12V-2×6 cable from an ATX 3.1 PSU eliminates the adapter’s additional contact junctions and the resistance variation they introduce. Testing with Thermal Grizzly WireView and similar tools has found native cable current imbalance as low as 0.2A. This remains the lowest-risk cabling approach.

Reference adapter cable (four 8-pin to one 16-pin) is the configuration tested today. The additional junctions introduce resistance variation the card cannot measure and no power limit can eliminate. Owners using this cable should transition to a native PSU cable where their warranty terms permit.

Using the reference adapter, if transitioning is not immediately possible: ensure all four 8-pin plugs connect to independent, separate cable runs from the PSU — never a daisy-chained or shared run, which concentrates even more resistance into fewer paths. A power cap at 85 percent reduces absolute wire temperatures but does not address distribution ratio.

Aftermarket per-pin monitoring accessories: Thermal Grizzly WireView Pro ($76), Seasonic 12V-2×6 Ti Smart Power Cable, Cooler Master GPU Shield (available in China at 359 RMB — approximately $52 USD at current exchange rates), and Corsair ThermalProtect each detect pin-level imbalance before thermal runaway sets in. None of these changes the contact resistance of the adapter cable; they detect a dangerous condition and alert or shut down before damage occurs.

ASUS ROG Astral RTX 5090: the only retail RTX 5090 with built-in per-pin sensing. The ITE IT8915FN monitoring chip and GPU Tweak III Power Detector+ provide automated protection that no other shipping RTX 5090 currently offers. Combined with the 12VHPWR Guard utility, it provides a shutdown mechanism if any pin sustains more than 9.5A.

One additional complication the test user flagged: many RTX 5090 FE owners continue using the reference cable specifically to preserve their manufacturer warranty. Switching to a third-party native cable can complicate warranty claims if connector damage later occurs — a concern validated by the PNY warranty denial in September 2026. That trade-off — safety versus coverage — has no clean resolution under current industry practice.

Can Reducing Power Further Close the Gap?

One natural question the data raises: if 85 percent power still leaves individual wires over spec, how far must the power limit drop before the problem disappears? The answer, derived from the resistance-ratio framework, is that no feasible power limit on the reference adapter cable will bring the worst-case wire to within spec as long as the contact resistance ratio remains constant. At 4.52A intra-plug gap on Plug C at 488W, the hot wire carries 11.55A. The same ratio applied to a 60-percent power limit (~345W total, approximately 5.75A per wire if evenly distributed) would put the hot wire at roughly 8.16A — still elevated, and still running at a different temperature than its companion. Reaching a point where the hot wire reliably stays below 9.5A under any load would require a dramatic power reduction that would make the card uncompetitive for the workloads that justify its cost.

This is distinct from what a power cap can accomplish on a card with native cable and even distribution. For a native-cable user whose per-pin imbalance is ~0.2A, an 85-percent power cap is a genuine margin improvement. For an adapter-cable user with 4.52A imbalance, the cap is a noise reduction, not a safety solution. Today’s clamp measurements are what that distinction looks like in concrete numbers.

NVIDIA Has Not Responded to Per-Wire Distribution Findings

At the time of publication, NVIDIA has not issued a public statement specifically addressing per-wire current distribution in the RTX 5090 Founders Edition reference adapter cable. The company has previously stated that the move to 12V-2×6 resolved the insertion-related failures documented on the RTX 4090 generation. The systemic imbalance documented in today’s measurements — and in the broader record of independent testing since February 2025 — is distinct from improper seating and is not mitigated by the updated connector geometry.

At its rated 575-watt TDP, the RTX 5090 draws more power than any previous consumer GeForce GPU. The 12V-2×6 standard’s 600-watt theoretical maximum leaves functional headroom of just 25 watts — 4.3 percent — before the connector’s rated limit is reached at full load. When that load distributes unevenly across adapter cable junctions the card cannot see, the margin on the highest-loaded wire collapses entirely.

Frequently Asked Questions

Does applying an 85-percent power limit make the RTX 5090 Founders Edition safe to run on the bundled adapter cable?

Not fully. Today’s clamp measurements show that at 85 percent power limit (approximately 488 watts), individual wires in the reference adapter cable’s four plugs still exceed the 9.5A per-pin spec, with one wire reaching 11.73A — 23 percent above rating. The reason the power limit cannot fix this is physical: the current imbalance is driven by differential contact resistance across the adapter’s connector junctions, not by total power draw. Reducing total power by 15 percent reduces every wire’s absolute current by approximately 15 percent but does not change the ratio between wires, because the resistance distribution that causes the imbalance stays the same. Transitioning to a native 12V-2×6 cable from an ATX 3.1 power supply — which eliminates the adapter’s additional contact junctions — is the only cable-level mitigation that addresses the root cause.

Why can’t the RTX 5090 detect that one of its power wires is overloaded?

The RTX 5090 Founders Edition reference board uses a single shunt resistor to monitor incoming power. That design measures aggregate current from the combined 12V rail — it cannot distinguish between a scenario where each wire carries its nominal share and one where two wires are significantly overloaded while others carry proportionally less. NVIDIA’s RTX 30-series cards treated the connector as three independent monitoring groups; the RTX 40-series dropped to two; the RTX 5090 FE uses one. The ASUS ROG Astral RTX 5090 is the only retail card that adds individual shunt resistors per 12V pin, through an ITE IT8915FN monitoring chip, enabling real-time per-pin current display and a protective shutdown through the GPU Tweak III Power Detector+ feature and the 12VHPWR Guard community utility.

If I switch to a native 12V-2×6 cable, does that fully resolve the connector risk?

It substantially reduces the risk from adapter-specific current imbalance. Independent testing with monitoring tools like the Thermal Grizzly WireView Pro has found that high-quality native ATX 3.1 cables show pin-to-pin current imbalance of approximately 0.2A — roughly 22 times lower than the worst-case 4.52A gap documented today on the reference adapter cable. However, a native cable does not address every failure mode: DLSS 5 can push sustained board power above the 12V-2×6 connector’s 600-watt rated ceiling on single-connector RTX 5090 cards, a separate risk documented in September 2026 testing. Using a native cable, monitoring connector temperature during DLSS 5 sessions, and considering a per-pin monitoring accessory together provide the most layered protection currently available to RTX 5090 FE owners.

What practical steps should RTX 5090 owners take right now?

If you are using the bundled four-8-pin adapter cable: transition to a native 12V-2×6 cable from your power supply as soon as your warranty terms permit. If your PSU does not have a native 12V-2×6 cable, check whether the manufacturer offers one as a replacement. If you must remain on the adapter cable in the interim: ensure all four 8-pin plugs use separate, independent cable runs from the PSU (not daisy-chained runs); confirm the 16-pin connector is fully seated with an audible click; apply an 85-percent power limit as a partial mitigation for absolute wire temperatures; and consider adding a per-pin monitoring accessory such as the Thermal Grizzly WireView Pro (approximately $76) or the Seasonic 12V-2×6 Ti Smart Power Cable to detect imbalance conditions before thermal runaway occurs. Inspect the connector and cable visually every few weeks for discoloration, deformation, or unusual warmth.