New 12v specification could mean the end of melting RTX 4090 power connectors

Jimmy2x

Posts: 239   +29
Staff
Highly anticipated: The curious case of the melting connector has been a hotly debated topic among RTX 4090 owners, manufacturers, and independent third parties since the first case was reported in October of last year. Manufacturers have pointed out that users may not have fully seated the connector, while users and reviewers have blamed the manufacturers for a faulty implementation. Both parties have also highlighted potential shortcomings in the overall design of the connector. Now, a proposed change to the connection's specification could put an end to the debate once and for all.

Igor Wallossek of Igor's Lab has been at the forefront of the 12VHPWR connector issue since problems began surfacing last October. In his latest report, Wallossek provides some much-needed good news for RTX 4090 owners and 12VHPWR connector users: a real solution could be on the way soon.

A revision to the current connector's standard was recently released in a draft engineering change notice filed by the Peripheral Component Interconnect Special Interest Group (PCI-SIG). The new specification aims to address the design issues contributing to melting and connection failures associated with the RTX 4090.

According to the new specification and Wallossek's report, the new 12V-2×6 connector would be compatible with older graphics cards that utilize the original 12VHPWR header, power supply connectors that use the legacy 12VHPWR connection, as well as any existing adapters.

The new 12V-2x6 power connector would maintain the same number of contacts as the existing 12VHPWR power connector. However, the new specification would require 12 larger contacts to carry the power, with four smaller sense pins used for sideband signals. While the previous power connector had a maximum rating of 600W, the limit on the new 12V-2x6 power connector is 675W (600W for the connector and an additional 75W from the PCIe slot).

The legacy and proposed connectors may appear quite similar at first glance. However, users will be able to distinguish between the old and new styles based on their unique identifiers. The legacy 12VHPWR power connectors will have the H+ marking on the top of the connector. On the other hand, the new 12V-2x6 power connector will instead use the H++ identifier.

Like its predecessor, the new specification includes specific requirements for wires and cabling. As mentioned in Wallossek's report, manufacturers can prevent unnecessary stress on the crimp contacts by ensuring that cable bundles are not bent, stretched, or restricted immediately after exiting the connector. Wiring manufacturing and practices are provided in the IPC/WHMA-A-620 standard, which provides guidelines and requirements for the production of cables, wires, and other associated assemblies. The standard describes the materials, test procedures and acceptance criteria for creating crimped, mechanically secured, and soldered connections.

The report provides a detailed account of several differences between the current 12V implementation and draft 12V-2x6 specification, including variances in connectors, temperature thresholds, and power handling and delivery. A comprehensive summary of these changes, along with additional technical information and diagrams from the draft engineering change document, can be foudn in the original report from Igor's Lab.

Image credit: Igor's Lab

Permalink to story.

 
It wasnt, but go ahead and continue believing that.

A connector that is this hard to get to seat correctly, with such thin safety margins, should never have been approved in the first place. The 12v connector was just dumb.
My point when saying that it is the fault of the connector is that the entire job of a connector is to eliminate user error. It failed in the job of eliminating human error.

but here's another point, how about we don't make cards with 700watt TDPs? get 2 birds stoned with one bush
 
My point when saying that it is the fault of the connector is that the entire job of a connector is to eliminate user error. It failed in the job of eliminating human error.

but here's another point, how about we don't make cards with 700watt TDPs? get 2 birds stoned with one bush
Funny, AMD managed a 625 watt slurping monster, more then once, and never had this burning connector issue. The 8 pin is certified for 150 watt but is actually capable of 305 watts while remaining in safety levels for power and heat, or 104% overprovisioning. the 12 pin is capable of 684 watts.....or 14%.

Almost like whomever designed that cable cheaped out to save money. Of course nvidia would never do that.

they should have just made a 12 pin version of the old 8 pin plug that didnt have these heat issues or seating problems.
 
Funny, AMD managed a 625 watt slurping monster, more then once, and never had this burning connector issue. The 8 pin is certified for 150 watt but is actually capable of 305 watts while remaining in safety levels for power and heat, or 104% overprovisioning. the 12 pin is capable of 684 watts.....or 14%.

Almost like whomever designed that cable cheaped out to save money. Of course nvidia would never do that.

they should have just made a 12 pin version of the old 8 pin plug that didnt have these heat issues or seating problems.
I still insist that the purpose of an electrical connector is to eliminate user error(instead of connecting and soldering leads) and that the 12pin connector completely failed at eliminating user error
 
I'm preparing my PSU for next gen GPU's
iu
 
I still insist that the purpose of an electrical connector is to eliminate user error(instead of connecting and soldering leads) and that the 12pin connector completely failed at eliminating user error
Which is something we agree upon. The 12 pin connector is a terrible design that lacks the safety overhead of the 8 pin.

Even if the 8 pin is not correctly inserted, because it has a far higher safety threshold, it wont get hot enough to melt. The 12 pin, OTOH........
 
I still insist that the purpose of an electrical connector is to eliminate user error(instead of connecting and soldering leads) and that the 12pin connector completely failed at eliminating user error
Well the other purpose of not connecting and soldering leads is reusability. A connector should make connecting a PS Idi0t proof.
 
Well the other purpose of not connecting and soldering leads is reusability. A connector should make connecting a PS Idi0t proof.
Never thought of that. Maybe nVidia wanted self-soldering power connectors so that users had to buy new GPUs? I guess limiting VRAM as planned obsolesence wasn't enough for the leatherman
 
I'm preparing my PSU for next gen GPU's
iu

Never cheap out !- get the copper edition - this is just cheap metal -remember Copper for power , silver for signals ( actually looked it up for 5 seconds - copper is pretty good for data like audio ) as for gold plated who knows
 
Why would we need 600 WATT from the beginning...problems or not the fact that need such amount of power means there is something wrong
Because modern process nodes are capable of building GPUs far more complex then they were a decade ago.

There is no arbitrary watt limit for GPUs. In 2010 people were pearl clutching over the 480 pulling 350 watt, before that the same people pearl clutched over GPUs needing two slot coolers and pulling 150 watt. The horror, a whole PC used that not 5 years ago! There must be something wrong!
 
Because modern process nodes are capable of building GPUs far more complex then they were a decade ago.

There is no arbitrary watt limit for GPUs. In 2010 people were pearl clutching over the 480 pulling 350 watt, before that the same people pearl clutched over GPUs needing two slot coolers and pulling 150 watt. The horror, a whole PC used that not 5 years ago! There must be something wrong!
Just visited my YT archives. Who remembers quad fire and quad sli drawing almost 1000 watts of power.

This was one of my favorite channels.
 
Never cheap out !- get the copper edition - this is just cheap metal -remember Copper for power , silver for signals ( actually looked it up for 5 seconds - copper is pretty good for data like audio ) as for gold plated who knows
To follow-on, Silver plated copper wire is substantially better for high-temperature applications. We will likely never see that in consumer computers - due to cost. This link says as much, https://blog.thepipingmart.com/metals/silver-plated-copper-wire-vs-copper-wire-whats-the-difference/ however, I bought some silver-plated copper wire that has a teflon jacket, recently, to replace the wire in a space heater that burned out. The wire will probably outlive the heater. Its rated at 200-degrees C.
 
Just visited my YT archives. Who remembers quad fire and quad sli drawing almost 1000 watts of power.

This was one of my favorite channels.
I remember those days. OCed watercooled 590 SLI with a watercooled OCed 990x. Had to open a window in the middle of January.

This was the era that spawned 1.2, 1.5, and 2 kW power supplies. Those builds in cooler master HAF X cases were totally nuts.
 
Back