Seems like GN got to the bottom of the 12PIN connector debacle, or at least provided the best insight so far

[ravenshrike]

3 hours ago, pas008 said:

you have source on this?

remember nvidias 12pin for 3k cards seemed fine not sure if there were reports on anything bad on that?

Sorry, It was PCI-SIG itself who stated that this was problem. Of course, they tried to pass it off as being from cheap adapters, when in reality it was a design issue as shown by the fact that nVidia's own adapter can cause melting. 

 

https://www.techspot.com/community/topics/pci-sig-warns-that-some-12vhpwr-adapters-can-get-dangerously-hot.276959/

[NumLock21]

9 hours ago, mr moose said:

 

How do you know they aren't in constant communication with PCI-SIG?  How do you know they haven't done due diligence before releasing a statement?  and what makes you think they just parroted GN?

GN published a video on Nov 16, 2022, saying their findings for the melted 12 pin was due to user error.

2 days later, Nvidia responded to GN on Nov 18, 2022 their findings for the melting 12 pin was also due to user error.

 

If that is not parroting off what GN exactly said, then I don't know what is. If GN have never published their findings, is Nvidia going to say "User Error"?

[starsmine]

55 minutes ago, NumLock21 said:

GN published a video on Nov 16, 2022, saying their findings for the melted 12 pin was due to user error.

2 days later, Nvidia responded to GN on Nov 18, 2022 their findings for the melting 12 pin was also due to user error.

 

If that is not parroting off what GN exactly said, then I don't know what is. If GN have never published their findings, is Nvidia going to say "User Error"?

Exactly you don't know what is.

These reports takes weeks of testing and days to write by their technical writers before being passed on to a community facing group that is allowed to make statements on the companies behalf. They are not likely to release the full report publicly because its internal documents, not a publication, and require a level of confidence higher then a publication from gamers nexus would require. IS or IS NOT something rather then a MAY.

To even view PCI-SIG publications, you need to be a member. These are not generally things for public use.

[NumLock21]

18 minutes ago, starsmine said:

Exactly you don't know what is.

These reports takes weeks of testing and days to write by their technical writers before being passed on to a community facing group that is allowed to make statements on the companies behalf. They are not likely to release the full report publicly because its internal documents, not a publication, and require a level of confidence higher then a publication then gamers nexus would require. IS or IS NOT something rather then a MAY.

Like to even view PCI-SIG publications, you need to be a member. 

Hopefully they'll redesign the connector, even when it's partially or not fully plugged in, it isn't going to melt.

These mini connector isn't something new. There are them mini 24 pin where HP and Dell uses and those for PICO PC builds. Those are 24 pin, perhaps Nvidia can talk with the entire PCI-SIG to see if they can adapt that design for their 12 pin, cause so far those 24 pins haven't melt yet. 

 

 

[starsmine]

9 minutes ago, NumLock21 said:

Hopefully they'll redesign the connector, even when it's partially or not fully plugged in, it isn't going to melt.

These mini connector isn't something new. There are them mini 24 pin where HP and Dell uses and those for PICO PC builds. Those are 24 pin, perhaps Nvidia can talk with the entire PCI-SIG to see if they can adapt that design for their 12 pin, cause so far those 24 pins haven't melt yet. 

 

 

That seems like more engineering then the problem requires.
Its still user error for not plugging it in until it clips in, the two cheap engineering solutions are a better tactile response when it clicks, and shorter sense pins to disengage when sense pins pull out. 

Like a reverse gear lock out in a manual. it is ALWAYS user error on cars without reverse lock out, money shifts are on you, not the car, but a lock out is nice to prevent the user error.

[pas008]

1 hour ago, ravenshrike said:

Sorry, It was PCI-SIG itself who stated that this was problem. Of course, they tried to pass it off as being from cheap adapters, when in reality it was a design issue as shown by the fact that nVidia's own adapter can cause melting. 

 

https://www.techspot.com/community/topics/pci-sig-warns-that-some-12vhpwr-adapters-can-get-dangerously-hot.276959/

were there reports on the original 12 pin from nvidia on 3k series not this hvpwr12 which is technically 16 pin

 

just wondering how nothing happened on those from 3k series technically they are compatible according to nvidia just dont have those 4 extra pins

 

 

[NumLock21]

22 minutes ago, starsmine said:

That seems like more engineering then the problem requires.
Its still user error for not plugging it in until it clips in, the two cheap engineering solutions are a better tactile response when it clicks, and shorter sense pins to disengage when sense pins pull out. 

Like a reverse gear lock out in a manual. it is ALWAYS user error on cars without reverse lock out, money shifts are on you, not the car, but a lock out is nice to prevent the user error.

I'm not saying to just change the 12 pin to a 24 one. It's more about adapting the individual parts of that mini 24 pin to their 12 pin one like the internal spacing of each individual connector down to the pin itself. Those 12 pins almost look like sewing needles for being that thin.

[mr moose]

19 hours ago, ravenshrike said:

Except nVidia had issues with that connector before the product was released during their prototyping phase. So they could have changed the design, but that would have meant no nifty passthrough fan, full PCBs, and at least 3, probably 4 8-pin connectors on the 4090.

As was pointed out it was PCI-SIG that had a warning, it is quite possible that because the number of faults is so low that all of nvidia's in house testing didn't uncover a problem which could plausibly be put down to PCI-SIG encountering cheap connectors.

13 hours ago, NumLock21 said:

GN published a video on Nov 16, 2022, saying their findings for the melted 12 pin was due to user error.

2 days later, Nvidia responded to GN on Nov 18, 2022 their findings for the melting 12 pin was also due to user error.

 

If that is not parroting off what GN exactly said, then I don't know what is. If GN have never published their findings, is Nvidia going to say "User Error"?

 

Again, have you got anything to substantiate your claim, all this proves is that Nvidia came to the same conclusion.  

 

 

[NumLock21]

9 hours ago, mr moose said:

As was pointed out it was PCI-SIG that had a warning, it is quite possible that because the number of faults is so low that all of nvidia's in house testing didn't uncover a problem which could plausibly be put down to PCI-SIG encountering cheap connectors.

 

Again, have you got anything to substantiate your claim, all this proves is that Nvidia came to the same conclusion.  

 

 

No, but the time of it all, is how I see things.

If GN never published their findings, would Nvidia still have released that statement?

[LAwLz]

6 minutes ago, NumLock21 said:

No, but the time of it all, is how I see things.

If GN never published their findings, would Nvidia still have released that statement?

I don't see any reason to assume they wouldn't. 

They have made public statements in the past when users online have been displeased or had issues with something. For example the uneven GTX 970 memory bandwidth. 

 

Be careful with "it makes sense in my mind, so it must be that way" lines of thinking. It can often lead you astray because it's so easy to assume things, and then back those assumptions up with even more assumptions.