Support NeoGAF

[OverHeat]

Vertical mounting is a must if you don’t want a bend cable IMO

 

[LiquidMetal14]

I'm sure glad I took the precaution with the 3 4090's that I have. Those babies are going strong.

 

[SportsFan581]

It is indeed a really dumb design. For example, I have a Lian Li Evo Dynamic Mid Tower case which actually does provide a decent amount of width for a graphics card and the adapters. Currently have EVGA 3080 Ti in there which has PCIE Plugs on the top too, but the 3 cables do just fine in the space. If I was to get say a 4090/4080 or upcoming 5000 series (if they keep the same design), I am not certain it would fit without bending the cable too close for comfort.

And this is a case which has plenty of space for other GPUs, length or height. The whole design is just WTF from Nvidia.

I hear you. I recently switched out my ITX case with an Antec P10 Flux which is quite large (175mm from the MB to the side panel), but I'd probably still need a right angle adapter on there.

 

[demigod]

It is indeed a really dumb design. For example, I have a Lian Li Evo Dynamic Mid Tower case which actually does provide a decent amount of width for a graphics card and the adapters. Currently have EVGA 3080 Ti in there which has PCIE Plugs on the top too, but the 3 cables do just fine in the space. If I was to get say a 4090/4080 or upcoming 5000 series (if they keep the same design), I am not certain it would fit without bending the cable too close for comfort.

And this is a case which has plenty of space for other GPUs, length or height. The whole design is just WTF from Nvidia.

My son has this tower. We leave our computers on pretty much 24/7. I built his computer and his 2 cousins computer in June with a different case, no issues. All 3 has the corsair 1000 RM or whatever its called the new one.

 

[winjer]

Adapter Disaster! CableMod Recalls 25,300 12VHPWR Adapters As Damages Reach $75K

CableMod has initiated a "product recall" for its angled 12VHPWR power adapters after widespread incidents about the flaws in the product.

wccftech.com

Consumers should immediately stop using the recalled angled adapters and contact CableMod for instructions on how to safely remove their adapter from the GPU and for a full refund, including cost of shipping, or a $60 store credit for non-customized products, with free standard shipping.

Consumers will be asked to destroy the adapter and upload a photo of the destroyed product to https://cablemod.com/adapterrecall/. The instructions on how to safely remove the adapter are also located on that site. Once destroyed, consumers should discard the adapter in accordance with local laws.

The firm has received 272 reports of the adapters becoming loose, overheating and melting into the GPU, with at least $74,500 in property damage claims in the United States. No injuries have been reported.

 

[YOU PC BRO?!]

When it all comes down to it. Having to plug external power to a pcie card that already receives a portion from the motherboard is silly. The power should all come from the motherboard. Its time that a new standard be created that would allow the motherboard to directly supply enough juice. Someone is going to killed in a house fire due to these connectors.

 

[smbu2000]

When it all comes down to it. Having to plug external power to a pcie card that already receives a portion from the motherboard is silly. The power should all come from the motherboard. Its time that a new standard be created that would allow the motherboard to directly supply enough juice. Someone is going to killed in a house fire due to these connectors.

Motherboards would probably have to be radically redesigned to accomodate that much power, increasing costs. Not everybody is going to need that much power.

Besides video cards have been using external power connectors for a long time now. My 3dfx voodoo 5 5500 has a 4-pin connector for extra power and that card is from 2000.
People have been using 6/8 pin pcie connectors for years now without issue. It’s just this connector that has been having issues.

 

[winjer]

Ongoing Melting Issues with NVIDIA RTX 4090 Graphics Cards: Repair Shops Manage 200 Units Monthly

The NVIDIA RTX 4090 graphics card, launched in September 2022, continues to face serious issues with its power supply interface melting. NorthridgeFix, a GPU repair specialist, recently highlighted the persisting problem in a video that showcased a collection of damaged 16-pin power connectors...

www.guru3d.com

The NVIDIA RTX 4090 graphics card, launched in September 2022, continues to face serious issues with its power supply interface melting. NorthridgeFix, a GPU repair specialist, recently highlighted the persisting problem in a video that showcased a collection of damaged 16-pin power connectors from these cards. On average, the shop addresses malfunctions in 200 RTX 4090 units each month due to this defect. Originally, the RTX 4090 utilized a 12VHPWR connector, which has been prone to poor contact and subsequent melting. In response to the widespread issue, NVIDIA announced a transition to an improved 12V-2x6 power supply interface. Despite this upgrade, many units sold in the past two years still feature the older, problematic 12VHPWR connector.

The melting of the power connectors varies with the intensity of card usage. Some users report damage within months, while others may not notice problems for over a year. Graphics cards that are less frequently used generally show delayed onset of these melting symptoms.

That looks expensive...

 

[StereoVsn]

Ongoing Melting Issues with NVIDIA RTX 4090 Graphics Cards: Repair Shops Manage 200 Units Monthly

The NVIDIA RTX 4090 graphics card, launched in September 2022, continues to face serious issues with its power supply interface melting. NorthridgeFix, a GPU repair specialist, recently highlighted the persisting problem in a video that showcased a collection of damaged 16-pin power connectors...

www.guru3d.com

That looks expensive...

Yeah, saw that and man this connector sucks.

 

[rm082e]

Shitty situation. I looked at a couple of articles and didn't see it mentioned - How much does the repair cost when this happens?

 

[hinch7]

Connector is way too small for that much power running through it. There isn't an issue with the other 4000 series cards, under the 4090.

 

[Kataploom]

I've been thinking whether getting a 4080S or 7900XTX some time this year, then I saw a video saying it's still an issue and nope, not putting my family in danger because of videogames, Nvidia has to fix their shit before I even think on going back to them.

 

[MikeM]

I've been thinking whether getting a 4080S or 7900XTX some time this year, then I saw a video saying it's still an issue and nope, not putting my family in danger because of videogames, Nvidia has to fix their shit before I even think on going back to them.

Nothing wrong with that.

In terms of power delivery- whats the benefit of the Nvidia connector? Why not run 3x 8 pins like AMD?

 

[Celcius]

I honestly wish they had just stuck with the triple 8 pin power connectors.
It doesn't bother me at all with my 3090. The connectors feel snug, secure, and beefy.

 

[Kataploom]

Nothing wrong with that.

In terms of power delivery- whats the benefit of the Nvidia connector? Why not run 3x 8 pins like AMD?

I think the issues lies on AI servers, it's waaaaaay more comfortable to arrange and maintain thousands of cards connected by a single cable each instead of 3 per card, that has been my assumption for months since that's Nvidia main point, for them gaming is secondary market already

 

[Magic Carpet]

My Corsair GPU power cable is sitting in the box waiting for the day I can buy a RTX 5090.
What are the odds that I will never get to use it? Either because they change the connector or I can't find the 5090.

 

[Randall365]

I've been thinking whether getting a 4080S or 7900XTX some time this year, then I saw a video saying it's still an issue and nope, not putting my family in danger because of videogames, Nvidia has to fix their shit before I even think on going back to them.

I think 4080s runs on 320w - same as 3080 etc - it is the jump to 450w for the 4090 that causes the issue maybe?

 

[winjer]

FFS, now this connector is also burning PSUs.

16-Pin Power Connector Failure with RTX 4090 Running 75% power

In a recent incident reported on the Korean Quasar Zone forums, an owner of an MSI GeForce RTX 4090 Suprim X 24G graphics card experienced a simultaneous meltdown of the 16-pin power connector on both the graphics card and power supply sides.

www.guru3d.com

In a recent incident reported on the Korean Quasar Zone forums, an owner of an MSI GeForce RTX 4090 Suprim X 24G graphics card experienced a simultaneous meltdown of the 16-pin power connector on both the graphics card and power supply sides. This event is noteworthy because it occurred even though the power draw was intentionally limited to 75%, suggesting that power restrictions might not reliably prevent such failures. The 16-pin power connector, also known as the 12VHPWR connector, has been a focal point for issues since the RTX 4090's release in October 2022. Numerous users worldwide have reported meltdowns, prompting investigations into the causes, which range from connector design to material quality and the physical configuration of the power cable. Despite a revised connector design introduced by the PCI-SIG, confidence in the modifications remains low among many users.

The graphics card in question was powered by an FSP Hydro power supply, the capacity of which was not specified. This scenario underscores a broader trend where even under reduced power settings, the integrity of the 16-pin connector can be compromised. This raises questions about the overall reliability of the power delivery system in high-performance graphics setups. Further complicating the issue, incidents of connector failures are not limited to the cables supplied by the manufacturer. Many users have sought alternatives from third-party manufacturers. For instance, Cablemod, a well-known provider of aftermarket power solutions, was compelled to recall its line of 16-pin adapters after identifying fire hazards.

 

[ap_puff]

 

[CuNi]

12VHPRW is just a through and through flawed design that either needs to be majorly restructured, with way better margins for error, or completely replaced.

 

[Rivdoric]

12VHPRW is just a through and through flawed design that either needs to be majorly restructured, with way better margins for error, or completely replaced.

Ofc.
The funny stuff is us, consumers, that bought new $250+ PSU in 2023 for this shit flawed connector only to see that we'll probably need to buy another PSU 2 years later for the "revised" version.
OFC no one will take the blame and we can go f*** ourselves.
A true shitshow.

 

[CompleteGlobalSaturation]

I'm scared to disconnect my RTX 4080 for this very reason. To be safe I undervolted my card so hopefully it is not being stressed like an RTX 4090 is with a 450W TDP.

 

[BlackTron]

Fuck me this only make me more worried for the people who are gonna build my pc...

Fucking jock.

 

[winjer]

https://videocardz.com/newz/hong-kong-media-pcm-discovers-melted-cables-and-psu-failure-after-testing-rtx-5090d-5080

Prior to the launch of the GeForce RTX 50 series, we mentioned that since both the RTX 5090 and RTX 5080 GPUs have higher power limits than their predecessors, it was more than likely we would see cases of damaged GPUs and cables soon. We just didn’t know exactly what would happen, but issues related to increased power were a good guess.

Hong Kong media outlet PCM reported that they ended up with a broken 1200W power supply and melted cables after testing the RTX 5090D and RTX 5080 cards last week. They mentioned they “won the lottery” (perhaps because they didn’t receive review units and had to buy them through a lottery draw) but ultimately ended up with a broken PSU. No word on broken GPUs though.

Following our inquiry, PCM confirmed encountering system instability and melted cables after testing the RTX 5090D, 5080, and RTX 4090. They did not clarify the testing sequence or specific methods used, only mentioning that full load testing was conducted. It also remains unclear whether the cables were properly seated or if they used native PSU cables or adapters.

PCM might have simply been unlucky, as testing multiple cards with various connections can lead to mistakes. It’s possible the PSU was already failing, or the cables were loose prior to testing.

However, if you come across these pictures, keep in mind that this is not a widespread issue, nor have reviewers reported any signs of such problems becoming more common. Board partners and NVIDIA have upgraded their cards to the newer 16-pin power design, and some models even include thermal pads for the power connectors. Hopefully, these measures address any potential concerns.

 

[JohnnyFootball]

My Corsair GPU power cable is sitting in the box waiting for the day I can buy a RTX 5090.
What are the odds that I will never get to use it? Either because they change the connector or I can't find the 5090.

It’s stilll waiting I’m sure…

 

[V1LÆM]

https://videocardz.com/newz/hong-kong-media-pcm-discovers-melted-cables-and-psu-failure-after-testing-rtx-5090d-5080

How are they melting if the cable can do 600W?

Doesn’t the 5090 “only” take 500W max through the cable and the other 75W through the pcie slot?

Probably another case of people not plugging things in properly.

 

[winjer]

How are they melting if the cable can do 600W?

Doesn’t the 5090 “only” take 500W max through the cable and the other 75W through the pcie slot?

Probably another case of people not plugging things in properly.

At this point we still don't know what is going on. It it's a repeat of the issues with the 12VHPWR connections.
Or if it was just user error and this is a one time thing.

 

[HoofHearted]

From the site:

Update: PCM updated their post with new information. They now confirm they saw burn marks on RTX 4090, so presumably the issue was spotted after testing 4090. It looks like user error after all.

I was fortunate/lucky enough to be able to buy a 5090 FE on launch day. Yes - I'm sweating on the potential that these burn issues might arise again... but this is the first report I've seen so far...

Yes - the amount of power running through a single cable on this new card is eye-watering (mine pulls ~500-535 average under a stress test/benchmark). I have a 1000W Asus PSU - used a cablemod 90 degree cable with my 4080 FE - but can't use it with the 5090 because of the recessed connector. I went through various CableMod cables/connectors BTW (including the recalled connector) never had an issue for what it's worth.

I'm now using the original cable provided in the box by Asus (never was used) - but probably will switch out and order another CableMod instead...

 

[Maestr0]

If power is now a problem, should we expect new type of cable or another to power up the GPU ?

To avoid unnecessary damage

 

[Mithos]

 

[Celcius]

Based on the above video, is 12v-1x6 just as bad?
I have a Corsair HX1000i psu and I bought the official Corsair 12VHWR cable for it for whenever I can upgrade from my 3090. I didn't want to have to buy a whole new PSU when mine is only like 4-5 years old but I'm starting to wonder if you really need to run a newer native atx 3.1 pcie 5.1 12v-2x6 psu and cable, or is the whole standard just flawed since 12vhpwr and 12v-2x6 both use the same number of thin cables?

12VHPWR Cable | 600W PCIe 5.0 Power Adapter Cable

The official CORSAIR 600W PCIe 5.0 / Gen 5 12VHPWR PSU Power Cable delivers power from CORSAIR Type-4 PSUs to the most advanced PCIe 5.0 graphics cards.

www.corsair.com

Corsair says my cable is rated up to 600w and my PSU is high enough wattage for the 5090 as well...

edit: I think I should be fine based on this: https://www.corsair.com/us/en/explorer/diy-builder/power-supply-units/best-psu-for-rtx-5090/

 

[winjer]

Based on the above video, is 12v-1x6 just as bad?

Seems like it's even worse...

 

[Magic Carpet]

It’s stilll waiting I’m sure…

Yep, still waiting,
I wasted all that time watching Youtube channels with free 5090's review a product that is just smoke and mirrors.

 

[Celcius]

Monitoring software running on an Asus Astral 5090

 

[Celcius]

Another one:



Not reporting on the same case, HIS cable burned too with the 5090 FE

 

[Rivdoric]

Some comments here aged like fine wine.

 

[Lusal]

Another one:



Not reporting on the same case, HIS cable burned too with the 5090 FE

que lástima!​

 

[PaintTinJr]

Based on the above video, is 12v-1x6 just as bad?
I have a Corsair HX1000i psu and I bought the official Corsair 12VHWR cable for it for whenever I can upgrade from my 3090. I didn't want to have to buy a whole new PSU when mine is only like 4-5 years old but I'm starting to wonder if you really need to run a newer native atx 3.1 pcie 5.1 12v-2x6 psu and cable, or is the whole standard just flawed since 12vhpwr and 12v-2x6 both use the same number of thin cables?

12VHPWR Cable | 600W PCIe 5.0 Power Adapter Cable

The official CORSAIR 600W PCIe 5.0 / Gen 5 12VHPWR PSU Power Cable delivers power from CORSAIR Type-4 PSUs to the most advanced PCIe 5.0 graphics cards.

www.corsair.com

Corsair says my cable is rated up to 600w and my PSU is high enough wattage for the 5090 as well...

edit: I think I should be fine based on this: https://www.corsair.com/us/en/explorer/diy-builder/power-supply-units/best-psu-for-rtx-5090/

As I said in the other thread, based on the temperature in the first wire pair and backed by the amperage in the pair - in that de8auer video above your comment - it seem like the cable is having a major issue carrying the maximum power intended power for the first pairing without getting into a cyclical trap of heating the wires, making them more resistive causing them to draw more power.

Historically ISA/EISA/PCI just powered cards from the southbridge bus, and it wasn't until AGP - and then PCIe - that graphics cards optionally became powered by the southbridge and then took power from an external Molex 4pin(on AGP) IIRC, but so long as the card power draw never went above the southbridge bus power being supplied, the external power was never used.

Given the backwards compatibility of external power options for GPUs worked around with cable adapters it seems that GPUs continue to draw everything from the 75watt southbridge bus, and when exceeding draw from the first 6pins (of an 8pin), then draw from the remaining two pins, and then draw from the next 6pins, and eventually the last two pins on a conventional 2x 8pin setup.


So in the video the 23amps on the first 2pins of the 12pin cable is the fault as it is carrying 276watts.
There shouldn't be more than (660/6) 110watts per pairing at the cable limits, but it should be

500watts(575TDP - bus 75watts) / 6 pairs

~= 85watts per pairing

and as the first pair carries that 85watts the other pairings should then fill up to 85watts, until the full TDP is reached (500watt +75watt) based on adapter and external power backwards compatibility.

Obviously the PSU is being dumb and when it exceeds 110watts on a pairing it should supply the next pairing, and so on, but instead gets caught in a resistive feedback loop to a temporary working point of equilibrium until the first cable pairing melts and ultimately failing and surge damaging fries the cable and the PSU and card.

 

[Rentahamster]