• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

It's happening again, melting 12v high pwr connectors

Status
Not open for further replies.
I see your point.

3-5% over spec, 9.7A on a wire... sure, nothing burns. Probably. No big deal.

But honestly, it doesn’t matter if it’s a tiny bump over spec or a massive one - spec limits exist for a reason or else we don't need them.

Brand-new hardware, stock settings, $2,000, $3,000, $4,000 price tag - this shouldn’t be happening. It's not acceptable!

If a $550 mid-tier GPU from AMD or Intel had these issues, we’d riot! So why does Nvidia get a pass? They won’t even deny there’s a problem. No "This is overblown" - just total radio silence. Gaslighting at it's finest.

Yeah, they dominate the high-end market. But does that mean we just accept this? I’d love to hear anything from them - like, “Relax, it’s just a few bad cables,” or “0.2A over spec? You're totally fine.” Instead, they’re just ignoring it and hoping we do too.

At this point, does anyone really believe they care about their gamer community anymore? The very community that kept their business alive all those years from 1999 until AI came along?

This is all you need to understand about where Nvidia’s priorities really are right now:

View attachment 386657

Unfortunately NVIDIA know that they are the leaders and have 90% of Market share, therefore they won't care much as long as people will keep buying their GPUs.
Imo AMD could really start a new momentum for RADEON this year, just like they did with RYZEN in 2017.
 
I do learn something, avoid this shitshow till they really fix the problem.
Except that this is now how many generations with this flawed connector? Do we have any reasonable way to expect they will "fix" anything when the design itself is flawed.

I see your point.

3-5% over spec, 9.7A on a wire... sure, nothing burns. Probably. No big deal.

But honestly, it doesn’t matter if it’s a tiny bump over spec or a massive one - spec limits exist for a reason or else we don't need them.

Brand-new hardware, stock settings, $2,000, $3,000, $4,000 price tag - this shouldn’t be happening. It's not acceptable!

If a $550 mid-tier GPU from AMD or Intel had these issues, we’d riot! So why does Nvidia get a pass? They won’t even deny there’s a problem. No "This is overblown" - just total radio silence. Gaslighting at it's finest.

Yeah, they dominate the high-end market. But does that mean we just accept this? I’d love to hear anything from them - like, “Relax, it’s just a few bad cables,” or “0.2A over spec? You're totally fine.” Instead, they’re just ignoring it and hoping we do too.

At this point, does anyone really believe they care about their gamer community anymore? The very community that kept their business alive all those years from 1999 until AI came along?

This is all you need to understand about where Nvidia’s priorities really are right now:

View attachment 386657
Whoever made that graph need to lay off the wacky stuff.
 
Except that this is now how many generations with this flawed connector? Do we have any reasonable way to expect they will "fix" anything when the design itself is flawed.
The worst part is that NVIDIA seemed to have it right with the power design on the 3090 Ti (with 3 capacitors instead of 1 or 2 as of now) but they decided to not to go with it just to save a few $...
 
The worst part is that NVIDIA seemed to have it right with the power design on the 3090 Ti (with 3 capacitors instead of 1 or 2 as of now) but they decided to not to go with it just to save a few $...
But that's not related to the flawed design of the connector, which is proving once again to be a dangerous fire hazard.
 
Maybe 2 of them wouldn't be a bad idea on cards over 400w, I think its ok for little cards like my 4070Ti.
 
Whoever made that graph need to lay off the wacky stuff.

Oh, you think that’s too wacky to believe? And you need a more trustworthy source for openly reported revenue KPIs - Y'know, the ones big publicly traded corporations are required to disclose?

Alright, how about this one? Better? :)


nvda_statista.png
 
Man.. that Data Center usage.. I see why they don't care about home users too much anymore.
 
Oh, you think that’s too wacky to believe? And you need a more trustworthy source for openly reported revenue KPIs - Y'know, the ones big publicly traded corporations are required to disclose?

Alright, how about this one? Better? :)


View attachment 386706
You're missing the point. Those future projections are hogwash. That bubble is going to pop and it's started already. However, this crap is off topic. So I digress..

Maybe 2 of them wouldn't be a bad idea on cards over 400w, I think its ok for little cards like my 4070Ti.
That would work. But I would say anything that draws over 220W. Doubling up on these flawed connectors could solve the problem.
 
However, this crap is off topic.
I am at work during the day, so all of this stuff is just blending in for me, what a whirlwind of info.

I sometimes feel spent just trying to catch up on what transpired through out the day.

But I would say anything that draws over 220W. Doubling up on these flawed connectors could solve the problem.
My card does like 300w on the core, maybe 350 board power now that I have moved over to a new platform.

On AM4 I have seen my card do 400w board power while folding quite a few times, 380w for MSFS 2020.

But not now... AM4 was bottleneckin this bad gurl.

Anyways.. so far so good that I have not had problems..
 
You're missing the point. Those future projections are hogwash. That bubble is going to pop and it's started already. However, this crap is off topic. So I digress..
No, I think you’re missing the point. First it's "wacky stuff”, now you think it's a bubble? You may, of course, but those figures aren’t projections. Read the axis. Q3 FY25 is not in the future, it's in the past.

Second, this so-called crap isn’t off-topic. It actually explains why Nvidia is uninterested - if not financially indifferent -in resolving the connector issues we’re discussing in this thread.
 
Last edited:
I am at work during the day, so all of this stuff is just blending in for me, what a whirlwind of info.

I sometimes feel spent just trying to catch up on what transpired through out the day.
Right there with you.
My card does like 300w on the core, maybe 350 board power now that I have moved over to a new platform.

On AM4 I have seen my card do 400w board power while folding quite a few times, 380w for MSFS 2020.

But not now... AM4 was bottleneckin this bad gurl.

Anyways.. so far so good that I have not had problems..
Given the energy density, I would really like a second connector for that card. But that's just me leaning on being safe.
EDIT: For the higher end cards, absolutely hell yes do they need a second connector!
 
Last edited:
The way I see it, the 6-pin 75 watt and 8-pin 150 watt PCI Express Graphics (PEG) power connectors were designed with too much safety margin because the designers probably were worried about junk quality power supply units that used 20-gauge wire back when those connectors were adopted. The EPS12V connector has 8 pins, yet delivers 300 watts per connector to CPUs safely without complaint unless the connector is shown to be of poor quality. The fact that EPS12V connectors worked without any demonstrated problems if the connectors are good demonstrated that the PEG power connectors had too much safety margin, wasting video card board space. This probably annoyed Nvidia's graphics board engineers so much that they designed a power connector standard that cut the safety margin to nothing and somehow got that rushed through the standardization process before the flaws could have been adequately tested and found in time for the proposed standard to be rejected.

I think that 12V-2x6 needs to be banned, and should be replaced with the EPS12V connector or a standard with as much safety margin as an EPS12V connector. Maybe if you want to add some smarts like 12VHPWR or 12V-2x6 did to add some signalling, maybe we could have an EPS12V connector with added signal pins to implement a PMBus or similar protocol connection to allow devices to communicate their status to each other and signal the OS if a power fault is detected.
 
The worst part is that NVIDIA seemed to have it right with the power design on the 3090 Ti (with 3 capacitors instead of 1 or 2 as of now) but they decided to not to go with it just to save a few $...
Why is everything a capacitor?
 
The whole reason 12VHPWR exists is to be a single connector. I can't see Nvidia ever allowing dual 12VHPWR. I can't see them ever fixing it though either :laugh: Melting issues are unlikely to affect anyone except 90 series owners and there's so few of them, they probably don't care.

Personally I don't understand why it's not a violation of some electrical standard.
 
Why is everything a capacitor?
Capacitor (when pronounced "cuh-pack-itter") is more fun to say.

Personally I don't understand why it's not a violation of some electrical standard.
This comment and another one or two about UL (which is a company that famously wrote several standards in the US and now their name is still on them even though they're committee-written by members of various companies and you don't have to go to UL the company to do "UL" testing...similar to CSA being the Canadian Standards Association being a company you can have do your testing and their name is on the standards, but you can go to UL, Intertek/ETL, TUV-SUD, etc. to do the actual tests) make me want to put some information out there.

UL Safety standards in the US usually only care about what they consider "hazardous" voltages. They define that as anything at or above 30Vrms/42.4Vpk for AC voltages and 60VDC (for dry location use products...it's 16Vrms/35VDC for wet locations). So for 12VDC, they don't even consider it hazardous. There are other requirements for temperature of touch-points, but that gets much deeper and you'd have to find an applicable standard for this application. The way the standards generally operate, most individual items are tested independently, like the power supplies are tested to IEC 62368-1, which is a global harmonized standard that has replaced US standard UL 60950-1. They would test with general loads, not in a system. The cables that come with the supply are likely not even tested with it, but the supply would be hooked up to some general test equipment that simulates loading. There would be requirements on flammability ratings for the connectors and wires, but it's not a quality control standard either, so it isn't going to force any kind of manufacturing standards on the connectors and cables in question.

Also, the United States doesn't enforce any UL standards testing. The closest thing I've seen to that is an OSHA rule that says if you're a business and you're going to make your employees use a piece of equipment, it has to be certified to the applicable safety standard by an NRTL (Nationally Recognized Test Lab). Most companies don't follow this rule, but some big companies (like META oddly enough) get audited so when they buy equipment to run their Oculus hardware or other stuff, they make sure it's UL-standard tested.

Other countries, like those in EU have laws about testing all products on the market. They have the CE system in Europe and there are directives with standards in them that are intended to test all things. They would also use the IEC 62368-1 standard as IEC is an international electrical equipment certification body that most of the world uses for standards. Most current UL/CSA standards are now harmonized with the IEC standards for example. CE gets a little more interesting because you'd have a standard for every component, and a standard for systems if they're sold that way. So the power supply would be tested, and if Dell (for example) wanted to sell a computer in Europe, that computer should also be tested to an applicable standard (I'm not sure what that standard would be). I personally test a lot of systems to IEC 61010-1, which is "Safety requirements for electrical equipment for measurement, control, and laboratory use". Depending on the customer, I could test a computer to that standard. The difference with the NRTL OSHA requirement above and CE is that CE is a self-declaration (for most markets, not medical, fire alarms, etc.) which means a power supply company can do their own testing (which is still supposed to be to the correct standard), publish their own report, and make their own declaration of conformity. That's the standard practice for safety standards. I've seen bulk DC power supplies from big industrial companies (that take universal AC-input and output 12VDC anywhere from 30-1000W) have their own safety report that they generated that was lacking any kind of clarity on what they actually tested or how but more of a "oh yeah, we definitely did this test, trust us, *wink wink*". I've seen a simple piece of equipment sold to optical labs that had no safety standards applicable to it because it just ran on USB power, but it was supposed to be tested to EMC directive standards for noise immunity...when I asked them for their report, they sent me a declaration of conformity...the standard they listed was for "electric griddles and griddle grills". I was like "you googled 'Declaration of Conformity' and just stole a document from the internet at random, didn't you?" They apparently didn't even look at what the standard was lol.

There's also a lot of products that end up in circulation with "UL" or "CE" marks on them that are fraudulent, usually from one particular country that has had a tendancy to say "oh, that's not the European "CE" mark, that's "China Export". That's another whole topic and I've written a book here that I'm betting very few will read lol.
 
Last edited:
But that's not related to the flawed design of the connector, which is proving once again to be a dangerous fire hazard.
So far there have been no case of melting Connectors & Cables on 3090 Ti so there is definitely something to learn from there... As Buildzoid mentioned the 3090 Ti had 3 separate shunt resistors whereas the 4090 & 5090 only have 1 or 2 depending on models. Maybe that explains why...

1740686789642.png


Why is everything a capacitor?
3 separate shunt resistors, my bad! :)

The whole reason 12VHPWR exists is to be a single connector. I can't see Nvidia ever allowing dual 12VHPWR. I can't see them ever fixing it though either :laugh: Melting issues are unlikely to affect anyone except 90 series owners and there's so few of them, they probably don't care.

Personally I don't understand why it's not a violation of some electrical standard.

I'm pretty sure Nvidia forced AIBs not to use 2x 16-pin connectors to make their FE seem like the "best" variant:
- much cheaper
- all GPUs are limited to 600W (Nvidia probably enforced that too)
- if the Connectors & Cables were only melting on the FE, then everybody would want AIB variants and it would make NVIDIA look even worse...
 
Last edited:
So far there have been no case of melting Connectors & Cables on 3090 Ti
Yeah there was, are you kidding with that?

As Buildzoid mentioned the 3090 Ti had 3 separate shunt resistors whereas the 4090 & 5090 only have 1 or 2 depending on models. Maybe that explains why...
Maybe. If we had a PROPERLY designed connector no one would need such. As I've said elsewhere, they need to go back to the previous PCIe 8pin connector and improve from there or double up on the new one. The engineers who came up with that connector failed to understand electronic physics when they came up with this new connector.
 
Yeah there was, are you kidding with that?

Well I never heard about them... and even Gamers Nexus/Hardware Unboxed/JayzTwoCents/der8auer and other channels all said in their recent videos that they never heard about such cases either. So I'm wondering where you heard that. Do you have any sources?
 
Well I never heard about them... and even Gamers Nexus/Hardware Unboxed/JayzTwoCents/der8auer and other channels all said in their recent videos that they never heard about such cases either. So I'm wondering where you heard that. Do you have any sources?
Are you kidding with that crap?
 
Oh?
The melting jack problems STARTED with the 3090 and included the 3090ti and even the 3080ti. EVGA had a huge problem with them.

Anyone not finding that information is not looking properly.
 
Last edited:
Oh?
The melting jack problems STARTED with the 3090 and included the 3090ti and even the 3080ti. EVGA had a huge problem with them.

Anyone not finding that information is not looking properly.

The 4090 was released on October 12, 2022 and the articles are from October 28th 2022 and June 16th 2023... Also both talk about RTX 4090 and not 3090 Ti
 
WireView is load balancing all 6 wires for my 4090 , got the ideal from JTC .
 
The 4090 was released on October 12, 2022 and the articles are from October 28th 2022 and June 16th 2023... Also both talk about RTX 4090 and not 3090 Ti
+1

The 3090ti and 3090 regulate their power across pins - they have direct telemetry and power draw regulation. The 4090 and 5090 have a simplified connector that relies on
PSU-side regulation.

All of the mass melting started with the 4000 series and is caused by too much power being sent across too few pins, specifically the 4090 at full power and 5090 -- if you undervolt 4090 there is less melting or use 4080/5080 there is no melting that i've seen.
 
The 4090 was released on October 12, 2022 and the articles are from October 28th 2022 and June 16th 2023... Also both talk about RTX 4090 and not 3090 Ti
My bad. Didn't actually read them. I typed in "3090 power melt" and those came up.

You guys can dogpile all you want, I know damn well there was a problem with them as I personally RMA'd two different cards, one EVGA and one Gigabyte. I was not alone.. The problem wasn't as widespread as the 4000 series, but it started with the 3000s.
 
Last edited:
Status
Not open for further replies.
Back
Top