Jul 18, 2018
5,905
PXL-20221025-170708728.jpg
PXL-20221025-170703549.jpg

Hopefully better for the cables. Not much room tbh
That dirt/dust looks like sand in your case...lol
 

Zoyos

Banned
Oct 30, 2017
322
All they had to do was beef up the connector a bit in thickness to make sure it doesn't pull the pins loose. Cablemod will come through with a superior design and will do quite well for themselves, I'm willing to bet.

They are going to have to rewrite the spec it is just too much current for the practical contact and bend stability of the connector. Furthermore the metal pins on the GPU insert into a folded socket that reduces contact by splitting the socket when any leverage is applied just like the 8 pin, just smaller with more current.

74775_01_this-is-an-even-better-look-at-nvidias-new-12-pin-power-connector_full.jpg



All the top pins are burning because those are all the 12V connection. Ensure those are the ones being secured the most.

12VHPWR-pinout.jpg


While th 8 pin connector had 3 12V connections.

images


Literally doubling the current through much smaller contact area.
 

gabdeg

Member
Oct 26, 2017
6,033
🐝
Just found out the Asus TUF has its power plug flipped upside down so stuff like this Seasonic adapter wouldn't even work with it because the cable would go out the wrong side. This and the be quiet 12VHPWR cable having issues just makes me want to return all my stuff while I still can and come back at a later date when proper ATX 3.0 PSUs are out. Maybe by then I can figure out a card model that has less coil whine as well.
 

Theswweet

RPG Site
Verified
Oct 25, 2017
6,475
California
I was already set on waiting for AMD's move before getting my new GPU, but this whole 12-pin nonsense is likely gonna have me stick with team red just for peace of mind. What a nightmare!
 

Zoyos

Banned
Oct 30, 2017
322
Just found out the Asus TUF has its power plug flipped upside down so stuff like this Seasonic adapter wouldn't even work with it because the cable would go out the wrong side. This and the be quiet 12VHPWR cable having issues just makes me want to return all my stuff while I still can and come back at a later date when proper ATX 3.0 PSUs are out. Maybe by then I can figure out a card model that has less coil whine as well.

ATX 3.0 PSUs are going are just going to increase the likelihood of this being an issue because it puts this same connector on the PSU end.
 

GameAddict411

Member
Oct 26, 2017
8,600
They are going to have to rewrite the spec it is just too much current for the practical contact and bend stability of the connector. Furthermore the metal pins on the GPU insert into a folded socket that reduces contact by splitting the socket when any leverage is applied just like the 8 pin, just smaller with more current.

74775_01_this-is-an-even-better-look-at-nvidias-new-12-pin-power-connector_full.jpg



All the top pins are burning because those are all the 12V connection. Ensure those are the ones being secured the most.

12VHPWR-pinout.jpg


While th 8 pin connector had 3 12V connections.

images


Literally doubling the current through much smaller contact area.
This all speculation on your part. In a circuit the current has to loop so if there is a connection issue, it will affect all the pins including the ground pins. Second of all, you are basically assuming that all of this effort by Nvidia and their AIBs is done by a bunch of idiots and you are somehow know more than everyone from their team. The reality is that this issue is extremely blown out of proportion. And from the issues we have seen online, we don't know how much power their cards were really pulling. They could be using custom BIOS for all we care. And the pics indicate that the connector is pulled to the side. We have YET to see a connector with the recommended bend have issues melting.

Also Gamer Nexus just did a world record run with the 4090 with that very same connector pulling in probably twice the amount of power the regular bios allows. Their connector didn't melt.
 

rsfour

Member
Oct 26, 2017
16,982
I'm not gonna bother picking up a 4090, but going forward just in case this is still a thing, even for lower end cards for whatever reason, aren't pinouts different for some psus?
 

Akronis

Prophet of Regret - Lizard Daddy
Member
Oct 25, 2017
5,471
For the... none of you out there with a Hyte Y60, here's the GPU clearance. I took off the glass panel to show it better. It's maybe 1.5cm. GPU at stock runs at around 66-68c with the default fan curve on full load.

20221025_083453.png


ICue can control the FE LED. That's cool. I have it synced up with the rest of my lights :)

20221025_083342.png


Cable situation still isn't perfect. But it's better. I know the issue is some of the pins not making full contact when it's turned. So I'm not gaming until I can try and get in there and wiggle some cables a bit more.

OT, but I bought a Y60 to replace my O11D XL but haven't gotten around to actually swapping it out. How are you liking the case and how do thermals seem?
 

Dave.

Member
Oct 27, 2017
6,190
ATX 3.0 PSUs are going are just going to increase the likelihood of this being an issue because it puts this same connector on the PSU end.
I dread to think of the bends some cables have been forced in to as I cram the huge surplus of cables in to a tiny shroud at the bottom of my cases over the years 😬
 

crespo

Member
Oct 27, 2017
4,678
ATX 3.0 PSUs are going are just going to increase the likelihood of this being an issue because it puts this same connector on the PSU end.
This is what I was going to ask. Will this only be an issue with the adapters, or is it a general issue with the way the plug assembly itself was designed?

I'd like to believe a single 12-pin going from the PSU straight into the vidja cahd won't have anywhere near the issues 4 8-pins melting into one 12-pin via shoddily-put together adapter currently has. Guess we'll have to wait for more ATX 3.0 PSUs to be out in the wild to get a solid answer.
 
Last edited:

Zoyos

Banned
Oct 30, 2017
322
This all speculation on your part. In a circuit the current has to loop so if there is a connection issue, it will affect all the pins including the ground pins. Second of all, you are basically assuming that all of this effort by Nvidia and their AIBs is done by a bunch of idiots and you are somehow know more than everyone from their team. The reality is that this issue is extremely blown out of proportion. And from the issues we have seen online, we don't know how much power their cards were really pulling. They could be using custom BIOS for all we care. And the pics indicate that the connector is pulled to the side. We have YET to see a connector with the recommended bend have issues melting.

Also Gamer Nexus just did a world record run with the 4090 with that very same connector pulling in probably twice the amount of power the regular bios allows. Their connector didn't melt.

Your statement comes off all kinds of wrong.

I'm an electrical engineer who deals with aircraft electronics. You're calling a random person on the internet a comparative idiot.

There are a ton of really smart people at Nvidia. I've meet a few of them. They are still people, and people make mistakes. Especially when being pushed to maximize profit.

This will probably be a 4090 exclusive issue based on it's power draw / core cut relative to the other 40 series cards and the time until another card this powerful comes out. Maybe 1% of all PC gamers will buy this card. And this won't affect all of them either.

BUT, You can't blow a real fire hazard out of proportion.

Maybe it's a cable manufacturer using the lowest grade of plastics that can pass or lower. Maybe the sockets aren't exactly on spec.

Nvidia didn't put out a recommendation on cable bends, some individuals there actively put out warnings to PCI SIG while others brushed it off when reviewers brought it up.

Using one of the most meticulous reviewers setting a world record isn't exactly typical or representative of normal use case.

Regardless, these connectors are supposed to have overhead built into the spec so things like imperfections in cable management don't cause a fire.

This was unacceptable.

This is what I was going to ask. Will this only be an issue with the adapters, or is it a general issue with the way the plug assembly itself was designed?

I'd like to believe a single 12-pin going from the PSU straight into the vidja cahd won't have anywhere near the issues 4 8-pins melting into one 12-pin via shoddily-put together adapter currently has. Guess we'll have to wait for more ATX 3.0 PSUs to be out in the wild to get a solid answer.

It's the connector melting at the contact point. The number of wires is irrelevant as they are not the point of failure. In fact reducing the number of wires creates a different hazard, but that is a different part of the spec for consideration. Less likely to be an issue.
 
Last edited:

Patate

Member
Oct 15, 2022
551
The 12900K will do the same thing if you let it have unlimited power. It will hit 100c in just a couple seconds in CB23 with an AIO 360mm.

My 12900k is overclocked, with unlimited power and doesn't go above 90C even after 10 minutes of cinebench R23 with a Lian Li 360 Galahad AIO.
 

DieH@rd

Member
Oct 26, 2017
10,725
I'm eagerly awaiting picture of a side panel with two holes cut into it to make space for power cables. :D
 

Patate

Member
Oct 15, 2022
551
This all speculation on your part. In a circuit the current has to loop so if there is a connection issue, it will affect all the pins including the ground pins. Second of all, you are basically assuming that all of this effort by Nvidia and their AIBs is done by a bunch of idiots and you are somehow know more than everyone from their team. The reality is that this issue is extremely blown out of proportion. And from the issues we have seen online, we don't know how much power their cards were really pulling. They could be using custom BIOS for all we care. And the pics indicate that the connector is pulled to the side. We have YET to see a connector with the recommended bend have issues melting.

Also Gamer Nexus just did a world record run with the 4090 with that very same connector pulling in probably twice the amount of power the regular bios allows. Their connector didn't melt.

If you are talking about the run they did this Sunday then no, it was with a stock bios at 600w max output like any other card. No shun mod, no nothing. They talked about it during the stream. They tried to have an unlocked bios, talked to a rep from AIB, were told to speak to Nvidia, were told to speak to AIB then they forwarded the discussion so that both parties could talk to each other and then never got a reply.
 

F34R

Member
Oct 27, 2017
12,035
My 12900k is overclocked, with unlimited power and doesn't go above 90C even after 10 minutes of cinebench R23 with a Lian Li 360 Galahad AIO.
Undervolt? Which motherboard do you use?
Mine is 5.2/3.9 and it hits 99c really quick with stock settings with the Galahad 360. Gigabyte z670 aorus master.
 

NaDannMaGoGo

Member
Oct 25, 2017
6,010
This all speculation on your part. In a circuit the current has to loop so if there is a connection issue, it will affect all the pins including the ground pins. Second of all, you are basically assuming that all of this effort by Nvidia and their AIBs is done by a bunch of idiots and you are somehow know more than everyone from their team. The reality is that this issue is extremely blown out of proportion. And from the issues we have seen online, we don't know how much power their cards were really pulling. They could be using custom BIOS for all we care. And the pics indicate that the connector is pulled to the side. We have YET to see a connector with the recommended bend have issues melting.

Also Gamer Nexus just did a world record run with the 4090 with that very same connector pulling in probably twice the amount of power the regular bios allows. Their connector didn't melt.

They put in some hard facts in that comment just for you to call it all speculation. After which you speculate that the issue is plainly extremely blown out of proportion. Like, not that it might be. You just state that it is.

Baffling.
 

Moebius

Member
Oct 28, 2017
5,427
Whelp, guess cablemod is getting my business.

Would cablemod cables have the same issue eventually due to the very small pins? Someone mentioned something about the pins becoming corrosive eventually and due to the small size, it may be dangerous.

Also, if you had the 90 degree adapter + the cablemod cables would there be less of a risk of a fire here since you wouldn't have any crazy bends? I'm wondering if adding the 90 degree adapter would cause any issues because now you're daisy chaining cables.
 

crespo

Member
Oct 27, 2017
4,678
It's the connector melting at the contact point. The number of wires is irrelevant as they are not the point of failure. In fact reducing the number of wires creates a different hazard, but that is a different part of the spec for consideration. Less likely to be an issue.
Ah fuck, right. Buildzoid's video did a great job explaining this, not sure why my brain decided to forget about that when I posted. Thanks.
 

Much

The Gif That Keeps on Giffing
Member
Feb 24, 2018
6,068
So... these 90 degree adapters? Looking like a good idea to buy for the 4080 as well? I'm assuming instance of a fire is less probable on a 4080 than the 4090s that we're seeing, but good to take precautions, right?
 

Captainslow

Member
Dec 3, 2020
364
Just found out the Asus TUF has its power plug flipped upside down so stuff like this Seasonic adapter wouldn't even work with it because the cable would go out the wrong side. This and the be quiet 12VHPWR cable having issues just makes me want to return all my stuff while I still can and come back at a later date when proper ATX 3.0 PSUs are out. Maybe by then I can figure out a card model that has less coil whine as well.

Can you help me understand this a bit more? I've backordered a TUF and I also have a Seasonic PSU. What do you mean "...the cable would go out the wrong side..."?
 
Oct 25, 2017
41,368
Miami, FL

gabdeg

Member
Oct 26, 2017
6,033
🐝
Can you help me understand this a bit more? I've backordered a TUF and I also have a Seasonic PSU. What do you mean "...the cable would go out the wrong side..."?
The 90 degree adapter would point upward instead of downward because of the flipped plug on the Asus cards. It's not a huge deal, CableMods for example will offer adapters that go both ways. But I assume most official 90 degree adapters from PSU manufacturers etc. will be made for the standard plug orientation that everyone but Asus seems to employ.
 

Patate

Member
Oct 15, 2022
551
Undervolt? Which motherboard do you use?
Mine is 5.2/3.9 and it hits 99c really quick with stock settings with the Galahad 360. Gigabyte z670 aorus master.

No undervolt, 1.32V, 5.1/4.0, 4.4ghz ring bus, MSI Pro Z690-A. I was hitting close to 99C on my old case. With my new Corsair 5000D and a top mounted radiator, I get the temp I've mentionned and I can post screenshots if needed.
 
Last edited:

F34R

Member
Oct 27, 2017
12,035
No undervolt, 1.32V, 5.1/4.1, 4.4ghz ring bus, MSI Pro Z690-A. I was hitting close to 99C on my old case. With my new Corsair 5000D and a top mounted radiator, I get the temp I've mentionned and I can post screenshots if needed.
Screens not necessary, I believe you. I just wanna figure out why mine are worse lol. Lian Li o11 Dynamic Evo, 6xAL120 intake fans (bottom, side). 3xAL120 fans on the Galahad 360mm rad, 1xAL120 exhaust fan on the back of the case.

CBR23 one pass with the PL1/PL2 at 241w.
 

No Depth

Member
Oct 27, 2017
18,449
Starting to think it may be better to just set my own card on fire than wait for the inevitable.

Anyone recommend propane and propane accessories that would get the job done?
 

Patate

Member
Oct 15, 2022
551
Screens not necessary, I believe you. I just wanna figure out why mine are worse lol. Lian Li o11 Dynamic Evo, 6xAL120 intake fans (bottom, side). 3xAL120 fans on the Galahad 360mm rad, 1xAL120 exhaust fan on the back of the case.

CBR23 one pass with the PL1/PL2 at 241w.

cinengcmv.png


This is after a 10 minutes loop. Did you leave every voltage set to auto in your bios?
 

Deleted member 16908

Oct 27, 2017
9,377
So I just found out (in the 4090 owners thread at overclock.net) that there's a discord where you can get notifications when your local best buy has a canceled order at the store. This is the server: https://discord.com/invite/gpu

You can get roles for USA/Canada notifications, which I think is amazon, newegg, bestbuy, etc. And also notifications for local best buys for your state.

2022-10-24_19_00_57-usa-drop-roles___Fixitfixitfixit_Drops__Tech_-_Discord.png


Not a few hours after joining, I got a notification for my local best buy. So I drove down right away and talked to the guy standing by the door as soon as you walk in (I also called them on the phone on my way there but don't bother doing this, as the person on the phone was clueless as to what I was talking about). The guy in the store looked it up on their system and said that they had 1 canceled order sitting over at customer service. He went over and grabbed it and I paid for it. Also chatted up another guy who was picking up his online order from the stock drop last Wednesday. I told him about the discord since his friends are also trying to find them.

Anyways, wooo.

20221024_170911.png


I checked the channel for my state again and it says out of stock for that location already. I don't know how it works, but it seems pretty accurate and timely with the updates.

You're a legend for posting this.
 

GameAddict411

Member
Oct 26, 2017
8,600
Your statement comes off all kinds of wrong.

I'm an electrical engineer who deals with aircraft electronics. You're calling a random person on the internet a comparative idiot.

There are a ton of really smart people at Nvidia. I've meet a few of them. They are still people, and people make mistakes. Especially when being pushed to maximize profit.

This will probably be a 4090 exclusive issue based on it's power draw / core cut relative to the other 40 series cards and the time until another card this powerful comes out. Maybe 1% of all PC gamers will buy this card. And this won't affect all of them either.

BUT, You can't blow a real fire hazard out of proportion.

Maybe it's a cable manufacturer using the lowest grade of plastics that can pass or lower. Maybe the sockets aren't exactly on spec.

Nvidia didn't put out a recommendation on cable bends, some individuals there actively put out warnings to PCI SIG while others brushed it off when reviewers brought it up.

Using one of the most meticulous reviewers setting a world record isn't exactly typical or representative of normal use case.

Regardless, these connectors are supposed to have overhead built into the spec so things like imperfections in cable management don't cause a fire.

This was unacceptable.



It's the connector melting at the contact point. The number of wires is irrelevant as they are not the point of failure. In fact reducing the number of wires creates a different hazard, but that is a different part of the spec for consideration. Less likely to be an issue.
Like I said it's just speculation based on cases you see online. The connector issue is trivial for a basic electrical engineer when it comes to testing out the limits. The only logical thing I can see is that either some connectors are just not made properly or the connection was poor due to a how the cable was bent. I am electrical engineer too by the way. If the connector really could not handle the rated power like you are claiming, you are basically saying they literally didn't do any testing. During my time doing testing, we usually tested above the rated power and temperature tolerance to see how it fails. I have a hard time believing this would be missed. Also there are tons of threads from previous gen GPUs with burned 8 pin connectors:
spvc9fk492a81.jpg

cablemod_melted_connector-jpg.2536142


Does that mean 8 pin connectors have wide spread issues? No. This is why I think people are blowing it out of proportion. They see a small number of people having issues on a new design which automatically means it's a bad design.
They put in some hard facts in that comment just for you to call it all speculation. After which you speculate that the issue is plainly extremely blown out of proportion. Like, not that it might be. You just state that it is.

Baffling.
They just provided the pinouts. Nothing that proves anything. Maybe you should inspect the "hard facts" before commenting. And yes it's speculation because anecdotals can be misleading. We haven't really seen any proper testing of the power connector from reputable sources.
 

unknown_nut

Member
Sep 12, 2022
1,617
Until we have more cases, it's most likely blown out of proportion. So far we have 3 users? Out of thousands. We need more data before panicing.
 

Zoyos

Banned
Oct 30, 2017
322
Like I said it's just speculation based on cases you see online. The connector issue is trivial for a basic electrical engineer when it comes to testing out the limits. The only logical thing I can see is that either some connectors are just not made properly or the connection was poor due to a how the cable was bent. I am electrical engineer too by the way. If the connector really could not handle the rated power like you are claiming, you are basically saying they literally didn't do any testing. During my time doing testing, we usually tested above the rated power and temperature tolerance to see how it fails. I have a hard time believing this would be missed. Also there are tons of threads from previous gen GPUs with burned 8 pin connectors:
spvc9fk492a81.jpg

cablemod_melted_connector-jpg.2536142


Does that mean 8 pin connectors have wide spread issues? No. This is why I think people are blowing it out of proportion. They see a small number of people having issues on a new design which automatically means it's a bad design.

They just provided the pinouts. Nothing that proves anything. Maybe you should inspect the "hard facts" before commenting. And yes it's speculation because anecdotals can be misleading. We haven't really seen any proper testing of the power connector from reputable sources.

In aircraft we have to deal with ensuring the connection of wires under stress while pulling at times multiple Gs of force.

In our case this is handled with backshell strain reliefs.

As an electrical engineer you should know what happens when current is pulled over a smaller cross section such as a poorly mated connector.

They literally did testing and reported it to pci sig as a concern and then released the 4090 anyway.

GN shared a few powerpoint slides from pci sig confirming this was found in testing.

"measuered 36.4A on a single pin, or 436W on a single pin"

The problem isn't the technical values, but the tolerances required to make them practical in a consumer product that doesn't require an aerospace standard backshell.
 
Last edited:

Deleted member 1594

Account closed at user request
Banned
Oct 25, 2017
5,762
You're a legend for posting this.
No problem. I feel like people are somewhat secretive about their methods so I wanted to be as detailed as possible as to how I went about getting one. Especially since the FEs in particular are annoying to get.

Also, I'd be on the lookout for another bestbuy stock drop tomorrow. Last wednesday was the previous one.
 

Tovarisc

Member
Oct 25, 2017
24,520
FIN
is saga with connector peaking now? We have DIY guides for how to make your own 16-pin connector with better pins, https://www.reddit.com/r/nvidia/comments/ydndbi/psa_you_can_reterminate_your_nvidia_4090_adapter/

Relevant post from OP
I discovered when building my own adapter that Nvidia is using double seamed/double split terminals, which bend and seperate much easier than the normal single seam (see https://youtu.be/ip03WmCWthc?t=122 for seperation action).

I found this weird as when building my own cable using OEM terminals, they are single seamed and seem to be much more resistant to splitting (https://www.mouser.com/ProductDetail/Amphenol-FCI/10132447-121PLF?qs=KVgMXE4aH4kD2xtDkHxOJw==)

Replacing these crappy Nvidia terminals with high quality, single seamed ones should result in a much better, damage resistant connection. They appear more copper in color as well so checking quality should be easy, also, look for single seam.

A stopgap would be to check your connector and manually bend back any separating terminals with a small screwdriver. Don't hulk it, these are thin wall copper, a pinch will do ya.

I wrote up a blog on how-to and my discoveries:

https://rossfisher.blog/2022/10/20/...ned-to-build-my-own-atx-3-sfx-power-solution/
Source: https://www.reddit.com/r/nvidia/com...?utm_source=reddit&utm_medium=web2x&context=3