kazuvikingIts not a real solution as it will burn regardless.

Agree, alot of people are shooting blind with DIY safety precautions because all ODMs can give us is plug it in properly while influencers/community tell us to lower our wattage all the while, manufacturers, the most likely to resolve this problem sell us monitoring gadgets that also have this hazardous connector.

In all 3 scenarios cards will still burn.

1. Lower wattage will only delay
2. Plugging it in properly, until the pins move through microvibrations
3. Expensive Monitoring gadgets still very prone and potentially even more so to degradation/movement due to extra resistance.

Work around, get a cable that can tolerate the faulty state because the connector will never not be faulty. 😂

Solution need a connector that grips the pins and will never move.

kazuvikingIts not a real solution as it will burn regardless.

Let's do some maths...
The PCIe bus provides 75 W max.
An 8-pin provides 150 W max.

So three 8-pins allows for a card of up to 525 W. The 5090 has a TDP of 575 W. That's 50 W over the theoretical maximum. And that didn't take the fact into consideration that 50-series GPUs pull very little power through the PCIe bus, and take most of what they eat through the power plugs.

So whoever thought 3 8-pins would work on a 5090 is a moron.

AusWolfSo whoever thought 3 8-pins would work on a 5090 is a moron.

Considering 2 8pins worked fine with the 500W R9 295X2 12 years ago, 3 8-pins will probably work fine with an RTX 5090. It's just that the PSU will need to be able to provide the extra wattage through those 8-pins with safety. The thing is that Nvidia knew where things where going. We can expect the RTX 6090 coming with 2 12V-2×6 connectors, so what might be looking like doable today - use 3 8-pin connectors to power the top model - tomorrow will not be enough. Nvidia simply gone too far with the miniaturization of the connector.

AusWolfLet's do some maths...
The PCIe bus provides 75 W max.
An 8-pin provides 150 W max.

So three 8-pins allows for a card of up to 525 W. The 5090 has a TDP of 575 W. That's 50 W over the theoretical maximum. And that didn't take the fact into consideration that 50-series GPUs pull very little power through the PCIe bus, and take most of what they eat through the power plugs.

So whoever thought 3 8-pins would work on a 5090 is a moron.

Though I understand your point of view, 8pin is capable of withstanding much more than it's rated for. GPUs with 2x8pin connectors can safely do 330W long-term. But yes, it's not good this way, as it violates the specified rated connector load. Fortunately, 8pin is more robust and thus has much higher safety margin than 12V2x6.

With such power hungry cards like RTX 5090, it'd be wise to include both load balancing and more than 10% power safety margin that 12V2x6 has. (In case load balancing does not work properly.)

And if load balancing is too costy, then low cost division of single power rail into 6 separate smaller rails with shunts would do the trick, provided each of the rail is connected straight to single pin of the connector. This was the approach with RTX 3090 (Ti), but there were just 2 rails. Separate rail.per each pin would further extend the safety factor.

As for this new CM "invention", yet another thing that does not solve the fundamental thing about the melting problem. The amount of these so called "splutions" only proves the problem does indeed exist, despite some users not acknowledging the situation and rather pointing to user erros (user abusing the connector).

LittleBroThough I understand your point of view, 8pin is capable of withstanding much more than it's rated for. GPUs with 2x8pin connectors can safely do 330W long-term. But yes, it's not good this way, as it violates the specified rated connector load. Fortunately, 8pin is more robust and thus has much higher safety margin than 12V2x6.

With such power hungry cards like RTX 5090, it'd be wise to include both load balancing and more than 10% power safety margin that 12V2x6 has. (In case load balancing does not work properly.)

And if load balancing is too costy, then low cost division of single power rail into 6 separate smaller rails with shunts would do the trick, provided each of the rail is connected straight to single pin of the connector. This was the approach with RTX 3090 (Ti), but there were just 2 rails. Separate rail.per each pin would further extend the safety factor.

As for this new CM "invention", yet another thing that does not solve the fundamental thing about the melting problem. The amount of these so called "splutions" only proves the problem does indeed exist, despite some users not acknowledging the situation and rather pointing to user erros (user abusing the connector).

Yes, there are solutions that Nvidia (and Sapphire) could apply to make the cards safe, especially considering their price. And 8-pin did very well to include a higher safety margin. Skipping the safeties, and reducing the safety margin is a very short-sighted approach, imo.

AusWolfYes, there are solutions that Nvidia (and Sapphire) could apply to make the cards safe, especially considering their price. And 8-pin did very well to include a higher safety margin. Skipping the safeties, and reducing the safety margin is a very short-sighted approach, imo.

The root cause is money, saving costs while keeping the final price same or higher. Also known as shrinkflation or greedflation.

LittleBroThe root cause is money, saving costs while keeping the final price same or higher. Also known as shrinkflation or greedflation.

I understand the concept of cost saving, but considering how little those safeties would cost, and how expensive those GPUs are, I cannot accept it.

AusWolfI understand the concept of cost saving, but considering how little those safeties would cost, and how expensive those GPUs are, I cannot accept it.

Cost savings on anything related to safety is simply disgusting. Safety first.

I work in heavy industry. After what I've seen so far (fatalities, severed or crushed limbs), there is NO excuse for omitting safety-related equipment.

Similarly, there is no excuse for your GPU setting your room on fire because it lacks equipment it previously used to have.

Para compradores nos EUA, a Cooler Master oferece apenas um ano de garantia para a versão americana do produto.

Source: VideoCardz.com
videocardz.com/newz/cooler-master-launches-50-gpu-shield-in-us-warranty-drops-to-one-year

atila pereiraPara compradores nos EUA, a Cooler Master oferece apenas um ano de garantia para a versão americana do produto.

Even they know its shit.

Offer a warranty (full replacement) for the protected video card if it burns up with the use of this add-on. Then, maybe, I would think about it. If I was gonna buy a card with this connector, which I am not, but that is another story I shan't tell here.

Compare to thermalgrizzly, CM's protector seems have no temprature detection function and can only detect over current. I would rather pay more for more comprehensive function. But CM's price is very attractive, maybe users that only need current protection function will like this.

RejZoRThat's not how it works. Pins themselves are not that much different between 8pin and 12pin HP, 8pin is officially rated to 150W per 8pins. 12pin HP is rated at 600W per 12pins (I'm just ignoring half of them is ground). To apply your logic you'd need a 32pin connector with 150W per every 8 pins. Essentially we'd get a connector shaped like that long thing we still use for motherboards...

Yes, using the current Molex Mini-Fit Jr.

However, there are other alternatives with higher ampacity ratings.

But, the Molex has been an industry standard for so many years... Change is difficult sometimes.

john_An excellent example of "if it works, don't fix it".

Unfortunately, nVIDIA has apparently adopted Microsofts motto; 'If it ain't broke, fix it 'til it is'.

Dr. DroI am so glad you are not an electrical engineer.

He said "brainwashed green ultra fans". Im sure he knows what he is talking about.

FrizzAgree, alot of people are shooting blind with DIY safety precautions because all ODMs can give us is plug it in properly while influencers/community tell us to lower our wattage all the while, manufacturers, the most likely to resolve this problem sell us monitoring gadgets that also have this hazardous connector.

In all 3 scenarios cards will still burn.

1. Lower wattage will only delay
2. Plugging it in properly, until the pins move through microvibrations
3. Expensive Monitoring gadgets still very prone and potentially even more so to degradation/movement due to extra resistance.

Work around, get a cable that can tolerate the faulty state because the connector will never not be faulty. 😂

Solution need a connector that grips the pins and will never move.

That hasn't been my experience at all. Have the cable for 4 years, just bought a wireview pro 2 just to actually check the condition of the cable, it's perfectly fine, evenly spread amperage across all pins and temperature at or below 55. What could I possibly do - after 4 years - to make the cable melt? It looks almost impossible to me unless I do it on purpose.

kazuvikingJust buy the rog equalizer and not this.

What would be the arguments for that one? I am curious to know what makes the equalizer the better option :)

From what i read, this one will beep and will shut off if overloaded. Where as the equalizer also has the shut of feature, but requires what seems to be a windows only software installed to get the warnings.
I obviously want whichever one will protect me the best, but it looks like they are about the same hardware wise? and warnings-wise the beep is more likely to alert me, than a software warning, since the hwardware shutoff will most likely black out my screen?

EvilingWhat would be the arguments for that one? I am curious to know what makes the equalizer the better option

Its completely dumb without any electronics. Its the best 12hpwr cable build quality wise on the market currently.

JustBenchingIt looks almost impossible to me unless I do it on purpose.

Sure that must be it, everybody must be melting it on purpose, it totally must be user error from those who aren't electrical engineers lacking the skill to plug in a connector. /s

AusWolfLet's do some maths...
The PCIe bus provides 75 W max.
An 8-pin provides 150 W max.

So three 8-pins allows for a card of up to 525 W. The 5090 has a TDP of 575 W. That's 50 W over the theoretical maximum. And that didn't take the fact into consideration that 50-series GPUs pull very little power through the PCIe bus, and take most of what they eat through the power plugs.

So whoever thought 3 8-pins would work on a 5090 is a moron.

Really? :laugh:
www.jongerow.com/PCIe/index.html
The Molex Mini-Fit Jr. specs say that using 16AWG it can do 8A per pin with brass contacts. The connector has 8 slots, but it's really a 6 pin since it only has 3x 12V lines.
8A means 288 W.

Using the mechanically compatible Mini-Fit Plus HCS it can do 10A per pin. That means 360 W per connector.

The 75 W for the 6 pin and 150 W for the 8 pin is a thing of the past, or better said it's made to account for very low quality equipment (especially compared to today) from 10-15 years ago. These specs are for the lowest common denominator, but again not from today (unless maybe you actively seek those) instead from the past (when they were somewhat more common).

By the way you can easily spot the bullshit.
The HCS specs say that for 12 terminal slots (like the 12Vhighfailure) with 16AWG it's 9A.
The 12Vhighfailure cable uses 16AWG but it doesn't use big boned HCS pitch and terminals.
And the Amphenol Minitek specs say 9.5A.
But somehow only 50 mating cycles instead of 100 for HCS.
Lying through their teeth!
Because the 12Vhighfailure terminals are closer together and are smaller than HCS ones (which have greater conductivity and lower resistivity) means they can't do 9.5A which is actually a bit more than a 12 slot HCS. Lol, lmao even.
They most likely can't even do the 6A of a 12 slot Molex Mini-Fit Jr. That would be 432 W.

It's way too obvious.
The 3090 Ti implementation was relatively fine because of the greater margins and load balancing. Not ideal but serviceable. The 5090 implementation is built to fail. Officially it's a different story:

Sol_BadguyJensen:
The voltage regulator design is state of the art. Incredible design. The engineering team did a great job.

tpucdn.com/review/nvidia-geforce-rtx-5090-founders-edition/images/front.jpg
tpucdn.com/review/nvidia-geforce-rtx-5090-founders-edition/images/back.jpg
Seems legit!

Hecate91Sure that must be it, everybody must be melting it on purpose, it totally must be user error from those who aren't electrical engineers lacking the skill to plug in a connector. /s

I'm talking about my own connector and experience. Read better and stop baiting.

I'm not baiting, the anecdotal "my connector is just fine so it must be user error" claims are getting tiring.
Those with plenty of experience have said the connector and 12v2x6 standard are flawed, this connector needs recalled since Nvidia decided to cheap out on the 12vhpwr standard by eliminating load balancing from the specifications.

Hecate91I'm not baiting, the anecdotal "my connector is just fine so it must be user error" claims are getting tiring.
Those with plenty of experience have said the connector and 12v2x6 standard are flawed, this connector needs recalled since Nvidia decided to cheap out on the 12vhpwr standard by eliminating load balancing from the specifications.

Of course you are baiting cause I never said it's user error. I said if it is the design of the connector that can't handle that much power then ALL connectors should have issues. Clearly it's not the case. Therefore the other explanation that seems likely to me is that there are a lot of predamaged connectors / low quality from the factory.

Hecate91Sure that must be it, everybody must be melting it on purpose, it totally must be user error from those who aren't electrical engineers lacking the skill to plug in a connector. /s

like Steve from HUB. What an amateur. I am guessing he doesn't know what a graphics card is.

JustBenchingOf course you are baiting cause I never said it's user error. I said if it is the design of the connector that can't handle that much power then ALL connectors should have issues. Clearly it's not the case. Therefore the other explanation that seems likely to me is that there are a lot of predamaged connectors / low quality from the factory.

So, Nvidia accepts bad quality power connectors on products that cost more than $5000 today.
Good to know.

john_So, Nvidia accepts bad quality power connectors on products that cost more than $5000 today.
Good to know.

Nvidia doesn't make power connectors. PSUs and 3rd parties do. What are you even talking about.

And using that logic, AMD accepts bad quality power connectors. Good to know.

Clearly you people are not reacting the way you are cause there is nvidia in the title. Totally not. You are really worried about those GPUs burning someones house :D

Sol_BadguyReally? :laugh:
www.jongerow.com/PCIe/index.html
The Molex Mini-Fit Jr. specs say that using 16AWG it can do 8A per pin with brass contacts. The connector has 8 slots, but it's really a 6 pin since it only has 3x 12V lines.
8A means 288 W.

Just because you can, it doesn't mean you should. Safety margins exist for a reason.

Sol_BadguyIt's way too obvious.
The 3090 Ti implementation was relatively fine because of the greater margins and load balancing. Not ideal but serviceable. The 5090 implementation is built to fail. Officially it's a different story:

tpucdn.com/review/nvidia-geforce-rtx-5090-founders-edition/images/front.jpg
tpucdn.com/review/nvidia-geforce-rtx-5090-founders-edition/images/back.jpg
Seems legit!

This 100%.

AusWolfJust because you can, it doesn't mean you should. Safety margins exist for a reason.

I didn't say it should be run at that value.
Those are the max currents. Just like the 12Vhighfailure claiming 9.5A max (for Amphenol, while for Molex it's 9.2A).
Of course you always want comfortable margins.

So if I take the max for the Molex Mini-Fit Jr. config above we have 8A (288 W) but the PCI-SIG rating is 4.16A (150 W). That is 92% room for maneuver. Playing it a bit too safe, but as I said it was in order to account for the lowest common denominator and the wire/terminal variation that could be encountered in those days.
If I increase the rating to 5.55A (200 W) then the safety margins are at 44%.

Okay let's move to the HCS connector.
That was 10A for both 8 pin PCIe (360 W) and EPS (480 W).
What if I put a PCI-SIG like rating of 7A in order to have some margins? Then 10A is 42.8% higher, almost the same margins as with the Molex Mini-Fit Jr. above.
7A means 252 W for 8 pin PCIe and 336 W for 8 pin EPS again using HCS connectors and 16AWG.

Now let's see how the 12Vhighfailure does.
The 575 W 5090 FE TDP translates to 8A per pin.
9.5A is 18.8% higher. So the safety margins are less than half of those presented above.
And we know other AIB 5090 implementations use 600 W default.
But anyway the halved safety margins are assuming the terminals can actually do 9.5A max.
Like I said previously using the HCS table a 12 slot/pin connector would be rated at 9A max.
How exactly is the 12 pin 12V-2x6 which uses smaller terminals which are closer together capable of similar performance?
www.techpowerup.com/forums/attachments/1688409852822-png.303371/
It doesn't pass the visual test.
Imagining a 12 pin HCS instead of that 8 pin next to the 12Vhighfailure makes it immediately clear that both cannot have the same capability. Well actually it's 9.5A vs 9A so even more head scratching.
This is probably the most "cool story bro" of the last 10 years or something.

Sol_BadguyThey most likely can't even do the 6A of a 12 slot Molex Mini-Fit Jr. That would be 432 W.

44% safety margins for 432 W max would translate to 300 W (4.16A, which is equivalent to two 8 pin PCIe of the "PCI-SIG play-it-safe variety").
And as I said even 6A might be optimistic, if I drop it to 5A max, then 3.5A (252 W) has 42.8% safety margins.
rog.asus.com/power-supply-units/rog-equalizer/rog-equalizer/
Somehow 17A for such small terminals. Mmmkay... :confused::confused::confused:
That is 84.8% higher. Such an increase in durability is astounding, realistically you could expect 20 maybe 30% increase but simply pulling almost 85% more out of the hat is ridiculous.
Nice that they're at least quoting the lower (but still overconfident) 9.2A rating that Molex uses.