Friday, June 16th 2023
be quiet! Dark Power PSU Owner Reports Melting of 12VHPWR Connector
A member of the amusingly named and low populated r/4090Burning subreddit has reported a strange incident where the PSU side of his 12VHPWR connector had melted. Shiftyeyes67k shared his equipment's plight two days ago and included two photos with his story: "Started noticing a smell coming from my PSU (be quiet! Dark Power 13 1000 W) recently that smelled like burnt plastic. Decided to swap it out and noticed that the 12VHPWR cable was burned...From everything I've read this generally happens to the connector on the GPU side so I'm wondering if anyone has seen this yet?" His feedback shows that the relatively new connection standard has ongoing reliability issues, even though most 16-pin connector problem cases have affected beastly GPUs such as NVIDIA's GeForce RTX 4090.
The be quiet! Dark Power 13 1000 W PSU has native support for the 16-pin cable standard, and no type of adapter was used in Shiftyeyes67k's example. Tom's Hardware reported on this sole incident recently, and a be quiet! representative responded to the article's content (updated today): "This is a unique case and we already have reached out to the customer to learn more." The company statement continues: "As our brand is known for highest quality standards, we treat this seriously and have initiated an investigation." be quiet! recommends that any customers experiencing similar issues should contact their support team directly. Tom's Hardware has attempted to get a comment from NVIDIA about the latest problem, but "an Nvidia spokesperson said we may not hear back for a few days due to a company closure."The article's author, Aaron Klotz, signs off with this reckoning: "This issue couldn't have come at a worse time with more RTX 4090 16-pin connector melting reports still coming in. NVIDIA claims that all of the 16-pin issues are related to user error, with the connector not being seated properly. But it's hard to believe that all of the errors were due to user error since some of these latest reports come from people who claim to be veteran system builders. Hopefully, this power supply issue with the 16-pin power connector does not extend to more users. But if it does, this could become an even more serious problem for the graphics card and power supply industries."
Sources:
4090Burning Subreddit, Tom's Hardware
The be quiet! Dark Power 13 1000 W PSU has native support for the 16-pin cable standard, and no type of adapter was used in Shiftyeyes67k's example. Tom's Hardware reported on this sole incident recently, and a be quiet! representative responded to the article's content (updated today): "This is a unique case and we already have reached out to the customer to learn more." The company statement continues: "As our brand is known for highest quality standards, we treat this seriously and have initiated an investigation." be quiet! recommends that any customers experiencing similar issues should contact their support team directly. Tom's Hardware has attempted to get a comment from NVIDIA about the latest problem, but "an Nvidia spokesperson said we may not hear back for a few days due to a company closure."The article's author, Aaron Klotz, signs off with this reckoning: "This issue couldn't have come at a worse time with more RTX 4090 16-pin connector melting reports still coming in. NVIDIA claims that all of the 16-pin issues are related to user error, with the connector not being seated properly. But it's hard to believe that all of the errors were due to user error since some of these latest reports come from people who claim to be veteran system builders. Hopefully, this power supply issue with the 16-pin power connector does not extend to more users. But if it does, this could become an even more serious problem for the graphics card and power supply industries."
98 Comments on be quiet! Dark Power PSU Owner Reports Melting of 12VHPWR Connector
If the GPU needs more than 600 watts, they should be using two of these connectors on the PCB (600w x 2 = 1200w), trying to pull over 600w via one connector is asking for trouble.
BTW ..... There is an RTX 4090 with two of these connectors for just that, it's made by GALAX - RTX 4090 Hall of Fame Limited Edition card that uses a custom BIOS.
We used to have working connectors
those 30~20 years old connectors are still working
I wanna know what went wrong!
Luckily I not planning to have Geforce in near future, at least not for the next 3 years
In seriousness:
Intel was wise to let others 'ride out the teething problems', on their own standard. (Co)Incidentally, the problems have been isolated to Intel's "competitor(s)".
Can't say I feel bad for nVidia, especially since they want to blame the user; ala Apple's: "You're holding it wrong"
On that note:
It's clear that this new connector is far less tolerant to (inevitable) User-Error than PCI-e 8-pin or EPS 8-pin. Blame the user all they want, the rate of failure is higher than any previous standard; *that* 'says something'.
Mostly an aside:
I can't help but feel like there is some 'systemic' problem in the industry at-large. The exact same kinds of 'oversights' are happening in EVs too. Ex. The Mustang Mach-E was recalled for a main battery connector, failing in much the same manner as these +12VHPWR connectors:
"It's rated for +XX% over spec, but they keep melting or trying to catch on fire"
As for their current consumer cards, no, no need to, maybe next gen depending on how much power they need.
Any other cards using this power connector in datacenters?
(I'm curious if the "Pro" sector will start having issues with the connector too?)
The RTX 3090 FE and RTX 3090 Ti pull as much power as the RTX 4090 does, in fact the RTX 4090 FE and Nvidia partner RTX 4090 cards in most cases pulls less power.
If the RTX 30x0 FE and RTX 3090 Ti did not suffer with this issue ..... what else could it be ?
So why the melting connectors on the RTX 4090 from those like MSI, ASUS, Gigabyte ?
So why the melting connectors on the 3rd party cables connectors (CableMod for example) ?
So why the melting connectors on this 12VHPWR cable that comes with this PSU ?
Whats the common denominator here ?
User error ? Maybe
Wrong female pins used in the connector ? ASTRON vs NTK female pin ??!!??!!
Food for thought ??!!
www.techpowerup.com/forums/threads/be-quiet-dark-power-psu-owner-reports-melting-of-12vhpwr-connector.310179/page-2#post-5041951
www.techpowerup.com/forums/threads/be-quiet-dark-power-psu-owner-reports-melting-of-12vhpwr-connector.310179/page-3#post-5042645
8-pin PCIe handles 150W and uses three pairs of wires. That's 50W/wire at 12V, AKA ~4A per wire.
12-pin HPWR handles 600W and uses six pairs of wires. That's 100W/wire at 12V, AKA ~8A per wire AND the wires and connectors are smaller.
Only a total idiot would approve such a downgrade.
Only a complicit industry of idiots would go along with such a stupid decision.
Long before Nvidia existed as a company, there were standards, wikis, AWGs and other info about how much current a wire could handle in the real world; 12V HPWR violates all of them.
OH LOOK EVERYTHING'S MELTING. WOW, I DIDN'T SEE THAT HAPPENING!
/s
Both of you are trying to blame Nvidia here for something they did wrong which is not the case here.
3rd parties are not following the specs / guidelines laid out by Intel - they designed this new connector and it's specs, no one else.
Nvidia followed these specs laid out by Intel from the get go with the RTX 30x0 FE cards that use this 12VHPWR connector - NONE, I repeat NONE of these cards suffered with this melting issue and none of their RTX 4090 FE cards either (except for under test conditions as seen / conducted by Gamers Nexus in their video).
3rd parties are not following these specs, they are using a sub-par component, this being the female pin made by ASTRON.
If your going to build a Lamborghini with sub-par components in it your the one at fault here for not following the specs / guidelines laid out by the developer / designer of this connector.
Intel's 12V HPWR connector leaves ZERO margin for error, despite the manufacturing conformity and adherence to standards being awful where 99% of the hardware is actually made.
8-pin PCIe is overkill in the spec, because in the real world, manufacturers of connectors and cables are penny-pinching crooks who cut corners to make more profit. The 6-pin and 8-pin PCIe standard account for that with big fat safety margins.
The mistake isn't with manufacturers not meeting spec; The mistake is in expecting an industry with 30 years of habitual corner-cutting to suddenly not cut corners. The spec should have headroom in it to account for this and it simply doesn't.
Aris, who used to do the PSU reviews here, weighs in.
Show me a real case (end user case) where they have a cable that uses the NTK female pins with the connector melting.
It's much more likely this person didn't fully plug the connector in, as was exhaustively proven to be the cause of the previous "issues".
NVIDIA sells ~85-90% of the consumer GPUs on the market, many of which have this standard connector for both last and this generation of GPUs, yet this is proof of a design flaw? One person on reddit?
Yet, here we are with RTX 4090 cards from Nvidia partners (MSI,ASUS,Gigabyte etc) using the ASTRON female pins failing.
Same with the 3rd party cables using the ASTRON female pins.
And now a PSU using the ASTRON female pins in the 12VHPWR cable (the connector) that comes with the PSU melting the connector.
Zotac's RTX 3090 Ti AMP HOLOBLACK - the cable that came with it also uses the NTK female pins in the 12VHPWR connector (1st gen connector without the sense pins), I have one of these cards here.
There has not been any reports of these melting, No RTX 3090 Ti FE and not even the other Nvidia partner RTX 3090 Ti cards, as I assume that they are all using the same NTK female pins in the 12VHPWR connector on the cable that came with the GPU.
My bigger concern would be the fact that he got the cable to burn on the GPU side at 650W nearly instantly. This indicates that you could lower the wattage and increase the time under stress to reproduce the same burning. Most 4090s consume around 490 - 500w under full load and the higher end OC models consume more than that. Pretty worrying for anyone OCing or running one of the faster cards out of the box. Also, considering that Arris tests in a controlled environment, I suspect that higher ambient temperatures would also play a small role. It's extremely problematic that the safety margin is so small for this connector due to all the potential variables. I believe GN said something about being unable to get the connector to burn up aside from half seating it but I really wish they had published what didn't work as this video seems to contradict the idea that it takes an unreasonable wattage to burn the connector. If anything it only make me more interested in seeing a video detail the exact real world tolerances of the connector under varying test conditions. The 3000 series and 4000 series have different connectors. The 3000 series connector doesn't include sense pins and if you look at the pin design there's a single contiguous piece of metal with only one small gap, meaning it's significantly more resistant against spreading. No one was reporting this issue with the 3000 series, only the 4000 series and specifically with the 4090. The only card in Nvidia's 4000 series line-up that butts close to the connector's maximum supported wattage.
In addition, the 4000 series has yet to saturate the market and the sales of this generation are very poor. You are implying that said connector has significant market saturation but that isn't anywhere near the case. Mind you it should really only be an issue for 4090, data from 4080s, 4070 tis, 4070s, ect is kind of irrelevant as those cards are well under the safety limit for this connector. At least it should be, if those lower end cards are having issues with the connector there'd be bigger issues at hand.