Check Your 4090's 12vhpwr Connectors To Make Sure They are in Pristine Condition

Twinz · 16 Nov 2022 at 23:43

Grim5 said:
GN melted both Astron and NTK connectors. If you're seeing more reports of Astron connectors melting it's simply because there are more of them out there, they are Nvidia's preferred source partner and NTK is only used when Astron is out of stock and another possibility is that the Astron connector is designed in a way that makes it more prone to user error

That's not the "type" of connector I'm talking about.

The adapters have 12VHPWR and PCIe connectors on them.

Users are responsible for properly connecting both. They don't seem to have issues connecting the PCIe power connectors.

pc-guy · 16 Nov 2022 at 23:44

uscool said:
Well they can shove it where the sun doesn't shine I ain't buying anything anyway unless they one day have something at value around £700 mark

It’s ok, Rtx 4090 will be £700 by next gen

All the previous halo products end up like that.

pc-guy · 16 Nov 2022 at 23:45

Twinz said:
That's not the "type" of connector I'm talking about.

The adapters have 12VHPWR and PCIe connectors on them.

Users are responsible for properly connecting both. They don't seem to have issues connecting the PCIe power connectors.

You aren’t right about PCIE8pins

People do melt them properly or improperly connected.

You just need to Google.

Sc00p007 · 16 Nov 2022 at 23:46

Angilion said:
How do you explain how these users who are incompetent to insert a connector correctly manage to do so with at least half a dozen other connectors and probably have done so with dozens if not hundreds of connectors?

To be fair in my limited number of experiences, many times I have found friends or family not fully insert cables. It was often the cause of an issue, or they managed to get away with it. Now we have a connector that is unforgiving, I could quite believe that many of these cases are user error.
I was taught to always check to fully insert connectors and pc building wasn't quite as streamlined so needed to be more careful.

pc-guy · 16 Nov 2022 at 23:53

Would be funny if OCUK’s prebuilds systems have melting connectors.

Good test to do in an interview for technician.

Twinz · 16 Nov 2022 at 23:54

pc-guy said:
You aren’t right about PCIE8pins

People do melt them properly or improperly connected.

You just need to Google.

"Goggle" a connector that has been produced in much larger quantities and used in the field over far more time? Serriously?

We have multiple PCIe connectors for every 12VHPWR connector on these adapters...for this particular application.

With the 4090, users are not erring with the PCIe connectors at the same rate as the 12VHPWR connectors.

Adrian1976 · 16 Nov 2022 at 23:58

deuse said:
It wasn't only user error.
It was a number of things.

True but that's GN's nicest way of not totally laying the blame on the user, In all honesty there is these defects in any plug or metal that is manufactured but not to cause this sort of damage, What he is clearly trying to say is it is it is user error, After watching that video it's quite clear people are not inserting them to lock and then they're twisting off to an angle as they mess about with their cable management.

The percentage is so low, If this really was a metal or plastic defect thousands of users would have had this issue in the last month of the first batch and I do not know anyone who has or in here neither.

Don't get me wrong the tolerance for not a good fit the connector is **** but you still can't blame Nvidia if people don't plug them in fully it's just asking for trouble.

Twinz · 17 Nov 2022 at 00:02

Sc00p007 said:
To be fair in my limited number of experiences, many times I have found friends or family not fully insert cables. It was often the cause of an issue, or they managed to get away with it. Now we have a connector that is unforgiving, I could quite believe that many of these cases are user error.
I was taught to always check to fully insert connectors and pc building wasn't quite as streamlined so needed to be more careful.

The mode of failure, once user error or manufacturing defect occurs, is what matters.

"These cables don't really click." (as Steve mentions at 18:30 in the video) Combine that with most cards' inability to fit in most cases without pushing the cable against the side panel and you get a situation where vibration + pressure on the cable can slowly extract the cable with catastrophic results.

pc-guy · 17 Nov 2022 at 00:03

Twinz said:
"Goggle" a connector that has been produced in much larger quantities and used in the field over far more time? Serriously?

We have multiple PCIe connectors for every 12VHPWR connector on these adapters...for this particular application.

With the 4090, users are not erring with the PCIe connectors at the same rate as the 12VHPWR connectors.

Well, I can use your logic to you then. More pcie8pin are burnt and defective so they are clearly worse.

You see the fundamental flaw in your own arguments there?

0.1% failure rate isn’t indicative of a manufacturing issue or defect. If it was manufacturing issues or defect it would be 10x or 100x higher. Batches would be affected you are talking about thousands.

Apparently Nvidia made 100,000 4090 cores - let’s assume that’s all sold now. 0.1%-0.15% is 10-15 cards…which is about the number in that mega thread.

Twinz · 17 Nov 2022 at 00:08

pc-guy said:
Well, I can use your logic to you then. More pcie8pin are burnt and defective so they are clearly worse.

You see the fundamental flaw in your own arguments there?

0.1% failure rate isn’t indicative of a manufacturing issue or defect. If it was manufacturing issues or defect it would be 10x or 100x higher. Batches would be affected you are talking about thousands.

That's a straw man, not my logic.

When each adapter for the 4090 has more opportunities for user error with PCIe than 12VHPWR, user error should occur more often with the PCIe than the 12VHPWR side. (If it's *just* user error)

Even though users are required to plug in 3-4 times as many PCIe power adapters when installing a 4090, it's the 12VHPWR connector that gets a melting megathread.

Rogueywon · 17 Nov 2022 at 00:12

Sc00p007 said:
To be fair in my limited number of experiences, many times I have found friends or family not fully insert cables. It was often the cause of an issue, or they managed to get away with it. Now we have a connector that is unforgiving, I could quite believe that many of these cases are user error.
I was taught to always check to fully insert connectors and pc building wasn't quite as streamlined so needed to be more careful.

This is fine when the fail state is "no signal from the GPU" or "BSOD". It's not fine when the fail state is "stuff burns".

pc-guy · 17 Nov 2022 at 00:14

Twinz said:
When each adapter for the 4090 has more opportunities for user error with PCIe than 12VHPWR, user error should occur more often with the PCIe than the 12VHPWR side. (If it's *just* user error)

I don’t entirely disagree. There is clearly something to be bettered in the design.

But it is also heavily user error related. You cannot say it is 100% design issue. Clearly many of the early cases, user was at fault as they got found out that the connector wasn’t seated properly.

Conflating pcie8pin is irrelevant to this argument. Cos pcie8 also get burnt through primarily user error but also to faulty products. I would hazardously guess similar rate or lower of the even happening. But in mining scene, you get WAY more burnt out connectors. 1) loads of idiots and newbies 2) cheating out on cables

Sc00p007 · 17 Nov 2022 at 00:28

Twinz said:
The mode of failure, once user error or manufacturing defect occurs, is what matters.

"These cables don't really click." (as Steve mentions at 18:30 in the video) Combine that with most cards' inability to fit in most cases without pushing the cable against the side panel and you get a situation where vibration + pressure on the cable can slowly extract the cable with catastrophic results.

My comment was in reply to someone claiming that someone incompetent at plugging the 12 pin manage to correctly plug in all other cables. I was just pointing out that I've seen many instances of cables not being inserted correctly.
For mode of failure, we are still speculating - you are saying that people are inserting them correctly but vibration and area restrictions are causing the adaptor to be 2/3 plugged in. I am saying I have experienced many people to only plug in cables 2/3. Probably a bit of both

Sc00p007 · 17 Nov 2022 at 00:42

Rogueywon said:
This is fine when the fail state is "no signal from the GPU" or "BSOD". It's not fine when the fail state is "stuff burns".

Depend's on what is causing it. If it is someone who isn't fully inserting the cable, then it's on them, be more careful and learn the lesson. If it turns out that the majority is an adaptor defect and as claimed by some are vibrating out the sockets, then yes that is bad. Still all pure speculation, but we have a lot more good information than all the pseudo science and opinion passed as fact the last couple of weeks.

Twinz · 17 Nov 2022 at 00:45

Sc00p007 said:
My comment was in reply to someone claiming that someone incompetent at plugging the 12 pin manage to correctly plug in all other cables. I was just pointing out that I've seen many instances of cables not being inserted correctly.
For mode of failure, we are still speculating - you are saying that people are inserting them correctly but vibration and area restrictions are causing the adaptor to be 2/3 plugged in. I am saying I have experienced many people to only plug in cables 2/3. Probably a bit of both

I'm not saying that people are plugging them in "correctly". I'm suggesting that they may be getting plugged in without actually latching. If the cable isn't latched, (Because "They don't really click.") pressure on the cable (because the cards don't fit in most cases without the side panel making contact with the cable), could further unseat the connector.

Grim5 · 17 Nov 2022 at 00:52

If the cable isn't latched then lots of things can make it come even more loose over time, for example just the vibration inside your case from the spinning fan can and will make things can loose that are not held down properly and even the bending on cables can make the connector came loose from tight cable management or other things pulling against the cable.

That's how someone can plug the cable in and it runs fine (or so they think) for a few weeks before finally the cable becomes loose enough to overheat

Sc00p007 · 17 Nov 2022 at 00:55

Twinz said:
I'm not saying that people are plugging them in "correctly". I'm suggesting that they may be getting plugged in without actually latching. If the cable isn't latched, (Because "They don't really click.") pressure on the cable (because the cards don't fit in most cases without the side panel making contact with the cable), could further unseat the connector.

Don't really click doesn't mean it hasn't latched though. Just means you can't rely on a sound, again I've had cables like this in the past so I double check them to make sure.
So it then comes down to personal opinion. If I plugged in a cable and didn't hear a click and then didn't look carefully and do then "pull" test

. I'd blame myself if the cable started to uninsert due to me putting pressure on it or via case vibrations. I guess others, and I see their pov, would blame the connector for being too difficult.

Purgatory · 17 Nov 2022 at 02:17

well they got it to melt....

Grim5 · 17 Nov 2022 at 02:23

Purgatory said:
well they got it to melt....

About 12 hours too late mate

Angilion · 17 Nov 2022 at 02:24

Sc00p007 said:
Don't really click doesn't mean it hasn't latched though. Just means you can't rely on a sound, again I've had cables like this in the past so I double check them to make sure.
So it then comes down to personal opinion. If I plugged in a cable and didn't hear a click and then didn't look carefully and do then "pull" test . I'd blame myself if the cable started to uninsert due to me putting pressure on it or via case vibrations. I guess others, and I see their pov, would blame the connector for being too difficult.

I'd blame the connector for being badly designed or manufactured if;

1) A non-trivial number of people with extensive experience of inserting a connector into a socket (which should be a trivial task achievable by a toddler anyway) can insert the cable incorrectly and not notice because there's no indication when the cable is correctly inserted and little to no indication when it isn't.
2) There's no failsafe in place, to the extent that a failure can not only ruin the product but also cause a fire.
3) The product is an ultra-premium product overpriced to an extreme level.

Even if the straw that broke the camel's back is user error (which is unproven) it's still clearly a seriously flawed design and/or manufacturing process.