News Right-Angle 16-pin Connector May Save A Lot of RTX 4090 GPUs

[bigdragon]

The only problem with adapters is the extra potential points of failure.

True, but I think the risk of unseated connections would be lower than the risk of overly-curved wires when the tolerances are so restrictive. People need time to get familiar with the quirks, limitations, and dangers of the new HPWR connector.

Only one adapter is necessary to plug into the connector on the card. The 4 plugs on the other side of the adapter don't need any additional adapters.

Yeah, the majority of people should be able to make 1 right angle adapter work. I said 4 because some setups may benefit from up, down, and left power wire connections. I didn't want to assume everyone is using the traditional ATX case with standard motherboard orientation -- I know I'm not.

 

[watzupken]

In my opinion, all these discussions only shows that the connector design is a fail. The reason is because its a cable, and having restrictions as to how you plug it, you cannot bend it, how many time you can plug and unplug it, etc, just shows that it's very limiting. And here we can see the results, melted connectors which did not happen even with shunt modded RTX 3090/3080 that is drawing very high level of power. The point is that they have designed a cable with less and thin wires to deliver massive amount of power looks nice, but it's a dangerous balancing act to make sure it does not melt due to the heat. And on top of that, one need to be mindful of the restrictions even more so as not to melt their expensive hardware.

 

[InvalidError]

In my opinion, all these discussions only shows that the connector design is a fail. The reason is because its a cable, and having restrictions as to how you plug it, you cannot bend it, how many time you can plug and unplug it, etc, just shows that it's very limiting.

All cables and connectors have limits of those natures even if we may not be aware of them. Bend an AC power cord the same way you need to bend the octopus on an RTX4090 to make it fit what few cases can accommodate it without modifications and you'll have failures there too.

The connector itself is fine, just not the right choice for this application at least in its straight incarnation.

It is kind of funny that Nvidia did not foresee the problems with a straight connector when it placed the power connector on the side-panel-facing side of their over-sized GPU. Makes it that much more surprising that they didn't bundle a right-angled cable. The straight version was never going to fit any application without a relatively hard bend.

 

[Alvar "Miles" Udell]

You could cast the first 2" away from the plug in rock-hard epoxy inside a steel reinforcement box and still have problems with copper getting ripped out of the crimps from hard bends after the rigid "strain relief" because only the insulating jacket is constrained. Conductors inside the jacket are still able to move and the extra tension on outer conductors which have to stretch more to make their longer path will still get all the way back to the pins. Using thicker gauge wires won't solve that and HPWR is already calling for #16 which can already handle twice as much current as what HPWR requires.

No cable and connector spec supports having the heck bent out of it for cable management purposes. If you need a sharp or otherwise out-of-spec bend, use an appropriate angled plug or adapter.

There's a difference between "having the heck bent out of it" and curving it so it goes into the back of your case. The way some of these reports sound, especially the CyberPower pre built, it sounds like they have just curved it and not mashed it into a right angle, which makes it sound incredibly fragile, especially compared to PCI6 and PCI8 pins which could take a large amount of abuse.

 

[drivinfast247]

Or, you know, don't put the side panel on the case. Ever. I've been building for nigh on twenty years and the first thing that's cast in the pitch is the stupid useless side panel. Stick a standard table fan in front of the open case interior and boom automatic hundred+ dollars savings on a gazillion stupid (RGB) fans. I prefer the no nonsense industrial grey-black no FREAKING RGB crap design. My towers look less like Enterprise-Es and more like Borg cubes. And they do assimilate many components against their will.

Just because you do something for a long time doesn't mean you've been doing it correctly.

 

[massivevoid]

About the void warranty issue, see CableMod comment here:

View: https://old.reddit.com/r/nvidia/comments/ye9ede/will_be_cablemod_12vhpwr_cable_void_my_4090/


NVIDIA never said that using 3rd party adapters/cables will void your warranty.

 

[demonicus]

use a different adapter voids warranty pfff .........so use a different one if it messes up put in the other one and say hey it messed up how can they tell

 

[InvalidError]

use a different adapter voids warranty pfff .........so use a different one if it messes up put in the other one and say hey it messed up how can they tell

If the GPU's connector is melted and there is no corresponding damage on the cable you return with it, then they know for sure that you used some other cable.

 

[TheOtherOne]

Soon, we will end up with something like this just for GPUs!

 

[Integr8d]

Don’t know about you clowns. But I’ll be watercooling my cables.

 

[alceryes]

Just stop with the anti-Nvidia FUD already. Nvidia is the one that reported the hotspot/melting problem to PCI-SIG as a result of their own testing.
As it turns out, the problem was reported by NVIDIA as a member of PCI-SIG, who were testing various prototype power cables and supplies. They observed that some cables were prone to showing hotspots or even melting due to severe bending or due to excess connection cycles (about 40

Please explain to me what part of my statement is anti-NVIDIA FUD?
Regardless of whether NVIDIA went to PCI-SIG and said 'test this' before PCI-SIG showed their results, NVIDIA ignored the results. Results that were published BEFORE the RTX 4090 was released. No part of my statement is untrue. In fact, NVIDIA knowing that there was a potential issue before the card was released, asking someone else to test it, getting the same results from the tester backing up their idea that there IS AN ISSUE, and then releasing the fire hazard GPU anyway, just shows an even more disastrous management structure at NVIDIA than if it was just PCI-SIG raising the red flag.

I'm definitely not anit-NVIDIA. Just before my RX 6900 XT I had the NVIDIA Titan Xp. A spectacular GPU that is still a great performer today.

I am anti some things though.
I am anti-ignoring the warning signs. I am anti-corporate greed for the sake of safety. I AM anti-let's get a third party to test this because there may be a problem but let's ignore their results and release the GPU anyway.

 

[ssj3rd]

You still can’t purchase this cable…

 

[kmw350]

Why is the right angle design angling the cable downwards though? Surely it will run right in front of the GPU badge, (in many cases covering RGB lighting too).

 

[InvalidError]

Why is the right angle design angling the cable downwards though? Surely it will run right in front of the GPU badge, (in many cases covering RGB lighting too).

What is above the GPU? The CPU's HSF if using direct air cooling.

 

[Truwolf]

Concerned about only being expected to use a connector 30 times. Are these made of knockoff LEGOs? For a $1600-$2000 item I want some Tesla space-age engineering I can twist around, plug/unplug all day without being able to halfarse it.

 

[InvalidError]

Concerned about only being expected to use a connector 30 times. Are these made of knockoff LEGOs? For a $1600-$2000 item I want some Tesla space-age engineering I can twist around, plug/unplug all day without being able to halfarse it.

Just about every internal PC connector is only officially rated for 30 cycles. That includes normal PCIe slots and 6/8-pins AUX connectors. Nothing unique about HPWR cables there.

LGA sockets are only officially rated for 15 cycles.