Info 12VHPWR PSA on cable bending

[BFG10K]

The same GN video also shows how a person can have it look like it is plugged in, and then wiggle it out while cable routing.

Looked pretty easy to me to screw up.

Exactly right. This is faulty design made even worse by the fact that it's being combined with the biggest slab heater GPU ever made.

LOL. NV isn't in deep stuff. This will go nowhere.

They already paid out for bumpgate and 3.5GB, and none of those were fire hazards. The connector is a defective hazard.

 

[Leeea]

LOL. NV isn't in deep stuff. This will go nowhere.

It is hard to say.

Many high amperage connectors have circuitry that makes sure the plug is correctly seated before drawing high power.


Your PCI slot has such a mechanism, and it only pulls a mere 6 amps.

This high power connector is missing said feature, and it pulls 50 amps.

of course the counter argument is 8 pins never had this for 12.5 amps

the counter argument to that is a 8 pin is a much bigger and more substantial connector pulling a lot less power


This is roulette for class action lawyers. They bet the classes money and keep the winnings.

 

[The red spirit]

LOL. NV isn't in deep stuff. This will go nowhere.

AMD got sued about their FX "cores" and lost, I can see how Lucas could win.

 

[Heartbreaker]

Now the warranty statement will be updated like "Warranty void if the cable was not connected by qualified service personal"

There might be something to keeping the average moron away from 600 Watt power connectors.

 

[sze5003]

Retailers now can offer "professional installation" with a purchase of a 4090.

I was laughing when I saw on the microcenter website when you buy RAM they offer professional installation and I'm thinking what's so hard? But at this point anything is possible.

 

[GodisanAtheist]

Sounds like the connector being bumped or under pressure when closing the case (cause the 4090 coolers are f'ing huge) or even being pulled on jostled during cable management will lead to poor seating and possible melting.

No professional installer is going to touch this with a 100 ft pole.

 

[sze5003]

Sounds like the connector being bumped or under pressure when closing the case (cause the 4090 coolers are f'ing huge) or even being pulled on jostled during cable management will lead to poor seating and possible melting.

No professional installer is going to touch this with a 100 ft pole.

Well if it's connected properly as the last thing you do in the install, cable should already be routed and all you need to do is make sure it's inserted properly with force so that the connector is snug in the pins.

But of course this is extremely hard to verify even if the connector latch is in place so I get what you are saying.

It's bad design and all AIB's and possibly PSU manufacturers too should make use of a redesigned connector that connects much easier, and send this out to everyone with proof of purchase of a 4090.

 

[Leeea]

There might be something to keeping the average moron away from 600 Watt power connectors.

The amps are a bigger problem here then the watts.

Every time a person plugs a hairdryer in they are plugging in a 1400 watt 13 amp item. Although if you examine a modern hairdryer plug you will note it has a mechanism to verify it is plugged in correctly. Typically the wall outlet itself will also have a mechanism to do the same, either GFCI or AFCI. ( I am assuming you live in the US ).


The only thing I know of that I plug in that is 50 amps is my car charger, and that is a complete non-issue. Although the car does take several seconds to verify it is plugged in correctly before audible click, followed by it beginning to charge.


It is completely possible for people to use high amp plugs every day. There are safe ways to do it.

 

[Heartbreaker]

The amps are a bigger problem here then the watts.

Every time a person plugs a hairdryer in they are plugging in a 1400 watt 13 amp item. Although if you examine a modern hairdryer plug you will note it has a mechanism to verify it is plugged in correctly. Typically the wall outlet itself will also have a mechanism to do the same, either GFCI or AFCI. ( I am assuming you live in the US ).


The only thing I know of that I plug in that is 50 amps is my car charger, and that is a complete non-issue. Although the car does take several seconds to verify it is plugged in correctly before audible click, followed by it beginning to charge.


It is completely possible for people to use high amp plugs every day. There are safe ways to do it.

You are comparing using an appliance to internal installation and wiring of a PC.

They aren't comparable, and there are people that should probably have nothing to do with latter.

I was concerned before, but now that I know what the issue is, I have zero concerns about ever getting one of these connectors on a GPU. I have no issue properly seating and latching the power connector.

 

[Leeea]

You are comparing using an appliance to internal installation and wiring of a PC.

As previously mentioned, the 6 amp PCI slot in a PC also has a mechanism to verify correct connection.

They aren't comparable, and there are people that should probably have nothing to do with latter.

Perhaps, but the deciding factor of who is going to have something to do with the latter is going to be $, not technical skill.

I was concerned before, but now that I know what the issue is, I have zero concerns about ever getting one of these connectors on a GPU. I have no issue properly seating and latching the power connector.

If you remember the GN video prior, there were people who received connectors that were so tight they were unable to determine if they had latched.

Your experience may not reflect everyone's experience.


However, the whole unplugging and plugging it back in ritual should probably go away, replaced by a just inspect it closely if your worried about it. Perhaps a prayer to the machine god would also be beneficial.

 

[Heartbreaker]

As previously mentioned, the 6 amp PCI slot in a PC also has a mechanism to verify correct connection.

It's primarily a data slot, it's going to fail unless the data pins are seated, and it virtually impossible to seat the data pins without the power pins.

If you remember the GN video prior, there were people who received connectors that were so tight they were unable to determine if they had latched.

That only makes sense if it's so tight that you can't remove it, where it likely won't be an issue.

This is a problem that really only happens with a very loose connection.

 

[Muadib]

I ordered a replacement cable from Cablemod. It was $30 shipped, but unless someone has seen a burnt one, I'm keeping it.

 

[lopri]

TLDW:

Two main failure modes identified (they affect both manufacturers):

1: Improper insertion (not fully seating the socket, then some sideways force to pull it out of alignment).
2: Debris in socket, Possible metallic debris from insertion cycles. Creates an alternate, higher resistance path.

Hard to get over the size of that thing

 

[DooKey]

Hard to get over the size of that thing

That's what she said...

Yeah, they're big.

 

[IEC]

The irony is that GN's test shows that pouring all the 12V into 2 pins is still enough not to run into the issue as long as the pins are properly seated. So they could have stuck with 3x 8pin older style PCIe connectors with ZERO issue.

 

[gorobei]

The irony is that GN's test shows that pouring all the 12V into 2 pins is still enough not to run into the issue as long as the pins are properly seated. So they could have stuck with 3x 8pin older style PCIe connectors with ZERO issue.

nv couldnt stick with 3x 8pin because they couldnt fit them on the pcb with how much space the through fan cutout chopped away from the pcb. pushing the power limits has consequences in cooling and size, and now people are starting to pay for those consequences.


realistically the whole atx format needs a full rework or something brand new.

stick the video card slot on the top edge of the motherboard on a 90deg turn so the gpu pcb runs parallel(rather than perpendicular) to the mb pcb and stick a tower cooler on the gpu. move the m.2 ssd slots to the back of the motherboard. re-align/organize/standardize everything so power cables dont need to be 2ft long "just in case" you have to route something around the long way.

with 3.3 and 5v moving out of psu to the motherboard, manufacturers might as well start trying something new/better.

 

[AnandThenMan]

Never thought I'd see the day where a power cable had to be subjected to a scanning electron microscope.

 

[Stuka87]

Never thought I'd see the day where a power cable had to be subjected to a scanning electron microscope.

But it was so cool!

 

[coercitiv]

Never thought I'd see the day where a power cable had to be subjected to a scanning electron microscope.

Here, cure your wounds.

 

[dr1337]

nv couldnt stick with 3x 8pin because they couldnt fit them on the pcb with how much space the through fan cutout chopped away from the pcb.

I think they had plenty of space for 3x8 pin. But the main thing with the adaptors is they allow for either 3x or 4x 8 pin. So nvidia can make one single chonker card that accepts different amounts of power all using the same single 12+4 pin connector. Saves them money and looks a lot better than 4x 8 pins on a card with one being optional or something else weird.

 

[JustViewing]

Why it has to be 12 pins? Can't they design a connector with just 2 thick cables?

 

[Stuka87]

Why it has to be 12 pins? Can't they design a connector with just 2 thick cables?

There are downsides to using larger diameter wire, as well as downsides to having fewer conductors.

Larger wires are harder to bend, and can put more stress on connectors as they try to straight themselves. More conductors allows for smaller diameter and offers resiliency in case of a defect in one of them.

But this connector could have easily been at least the same size of a 8 pin connector, which would have made it more durable. Preferably the same size pins as the 8 pin, just with four more added to the end.

 

[adamge]

Why it has to be 12 pins? Can't they design a connector with just 2 thick cables?

Good point. Here's another solution: Why does it have to be a hard-mount connector directly on the PCB? Why can't the PCB have 15cm long power wires permanently attached to it and connectors on those wires, so the user connection to the power supply doesn't touch the graphics card at all?

 

[MrTeal]

So a pigtail of wires with a connector at the end of that that connects to your PSU power cables?

That's a pretty ugly and unconfigurable setup when using a connector has always been fine and still remains so. There's obviously some issues with this particular implementation in terms of margin and installation difficulty, but that's not a good enough reason to move away from the general concept of the system.

It would be awesome for cable management if they updated the x16 slot to have a long 12V and GND land like in a server PSU capable of 50A though, and then you plugged your 12VHPW cable into a right angle connector on the MB.

 

[Aikouka]

AMD got sued about their FX "cores" and lost, I can see how Lucas could win.

I don't think bringing up AMD losing a lawsuit over attempting to use the ALU count as the core count is really equivocal here. I mean... Nvidia isn't being sued for false advertising. Frankly, the GamersNexus report gives Nvidia the best ammunition that it needs in a lawsuit: "Failures have been attributed to user error." (I'm sure Nvidia has been looking into this themselves as well.)

Although, it does bring up an interesting question... if we think the 12VHPWR connector is just too haphazard to use, who should be at fault for it? I do like the idea of having the sense pins connect after the power/ground pins have already made proper contact. However, if the main issue is with latching, it seems like that's probably the biggest area that needs to be addressed. Even if the sense pins are moved to connect last, it may not help if the connector is still not latched in properly. Although, that makes me wonder what happens if the sense wires are impacted during use. Does the card only check them on boot-up or does it continuously monitor them? It seems like it's a boot-up only process.