Nvidia RTX 5070 cable melts in latest Blackwell card incident: another case of user error?

[midian182]

WTF?! There's been another melting incident involving a Blackwell graphics card. While the previous cases mostly resulted in damaged RTX 5090 cable connectors and sockets, this latest one saw part of the cable itself melt, and the card in question was an RTX 5070.

X user @ere9w posted videos and images of his brand-new system, which includes a Zotac RTX 5070 graphics card and a Seasonic Focus GX-750 power supply.

Most of the melting reports we've seen in the past have involved the connectors, but in this instance, it was part of the cable that melted.

接続端子部分と、ケーブル部分の拡大写真です。 pic.twitter.com/Mwuu7gegex

– ケンケン🐶自作PC依頼受付(クレカ決済◎)@埼玉 (@ere9w) April 10, 2025

VideoCardz writes that the cable was the newer ATX 3.1 design with 12V-2x6 connectors. It appears that a single wire has melted and the connectors on the PSU and GPU sides did not show the same heat damage.

It's likely that the problem stemmed from the RTX 5070. A close-up photo of the card's power socket shows that one of the pins is not visible – it has either bent or been pushed back.

追加の動画です、誰かわかる方いたら是非m(_ _)m

ちなみに、グラフィックカードはZOTACのRTX5070そして電源はシーソニックのGX-750です。

もちろん、両方ともに新品の物になります。 pic.twitter.com/fCO2O7o1cD

– ケンケン🐶自作PC依頼受付(クレカ決済◎)@埼玉 (@ere9w) April 10, 2025

Whether this was a manufacturing error such as a loose pin or user error is unclear. Nvidia suggested that the melting problems seen with the RTX 4090s was down to users not fully inserting the power cable connectors into the card socket, leading to excessive heat buildup and the melting and burning problems. But the fact that over 200 of these cards were going to a repair shop every month in 2024, an increase compared to the 100 per month the previous year, suggests otherwise.

双方の末端部分 pic.twitter.com/PoZk4h9MH4

– ケンケン🐶自作PC依頼受付(クレカ決済◎)@埼玉 (@ere9w) April 12, 2025

There have been several cases of RTX 5090 card connectors melting. There was one in February that some blamed on the use of an unofficial third-party cable from Moddiy. It was an ATX 3.0 PCIe 5.0 16-pin to 16-pin model supporting up to 600W with a 12V-2X6 design.

More recently, there was a case where both ends of a 12VHPWR cable, as well as the RTX 5090 and the PSU, all showed burn damage.

There has also been at least one melting incident involving an RTX 5080.

Overclocker Der8auer replicated the setup of one of these RTX 5090 melting incidents using a Corsair 12VHPWR cable. The cable's connectors reached 150°C on the PSU side and close to 90°C on the GPU side. The problem was an uneven distribution of power: two wires designed to carry 5 to 6 amps of current were carrying more than 20 amps each, while some cables carried as little as 2 amps.

Permalink to story:

Nvidia RTX 5070 cable melts in latest Blackwell card incident: another case of user error?

 

[gamerk2]

This whole mess to me highlights that we sort of need the modern equivalent of the AGP slot of old, something that can provide GPUs the power they require from the slot, without needing external connectors.

Also, the fact the connector doesn't 'snap' into place when fully inserted is almost criminal; I hate when HW guys don't provide connectors that self-lock, since you open up to these types of problems.

 

[Watzupken]

I guess the fatal design flaw stems from the fact that the cable is not deigned to balance the load on each cable evenly. So while it was designed to carry high power across with a single connector, it clearly failed the purpose.

 

[Theinsanegamer]

This whole mess to me highlights that we sort of need the modern equivalent of the AGP slot of old, something that can provide GPUs the power they require from the slot, without needing external connectors.

Also, the fact the connector doesn't 'snap' into place when fully inserted is almost criminal; I hate when HW guys don't provide connectors that self-lock, since you open up to these types of problems.

The irony of this statement is not lost on me, AGPs power limits are what prompted the use of external connectors on the first place, since AGP was insufficient.

I guess the fatal design flaw stems from the fact that the cable is not deigned to balance the load on each cable evenly. So while it was designed to carry high power across with a single connector, it clearly failed the purpose.

The lack of power balancing being required is a major design flaw. USB C has always had negotiation but in and you don't hear about this, despite pushing over 100w on a much smaller connector with fewer pins.

 

[Thatsdisgusting]

Only thing I wonder

- when nvidia decide to scrap this connector to burn in hell (literally), or fundamentally overhaul so it will be brand new and 0 compatible with current one -

What those happy ATX 3.over5000 PSUs gonna do?

 

[AnilD]

If the pin was pushed back in the socket from the start, that would explain the uneven power load and heat buildup. It really emphasizes how even a small physical defect can cascade into a serious hardware failure when you’re dealing with this much power draw.

The scary part is that this wasn’t a flagship 5090 but a 5070. If cables are still melting at mid-tier levels, it suggests there might be deeper design flaws or inconsistencies in manufacturing—either with the cards or the connectors.

 

[user556]

It's not one of the power wires that melted. It's one of the four capacity select wires instead. Dunno how that could happen at all.

Certainly not the usual fault.