Reverse engineering a 12VHPWR cable

[ShrimpBrime]

Yep, because thousands upon thousands of cards are having problems.

Source? Links? Even to just this forum more than 5 cases of melted connectors at TPU?

 

[Braegnok]

It's not the cable, and not the connectors at fault.

It is the lack of current and or load balancing and shunt resisters on the PCB that causes the 12+4 pin connectors to melt on the 4090 and 5090.

Using the power stages with current and load balancing and having shunt resisters on the PCB prevents issues,.. Nvidia decided to skip this with the 40 and 50 series cards.

 

[ShrimpBrime]

Right, hours waiting for a reply. So asking Google AI search instead, the overview simply states.
Question:
How many reports of 12 pin pcie connector melted

While the exact number of reports is difficult to pin down definitively, multiple sources indicate a number of cases where the 12-pin PCIe connector (also known as 12VHPWR) has melted or experienced damage, especially with the RTX 4090 and RTX 5090 graphics cards. Some sources report over 250 cases of failures with CableMod adapters alone. Other sources mention 23 confirmed cases and 6 unconfirmed cases on a Reddit megathread. Additionally, a survey on a different board showed 3.3% of users who had used older PCIe connectors had had a failure

Everything with a grain of salt obviously, but a 3.3% failure rate is not much different than AMD and Intel Processor failure rates.

But let's reverse engineer it... lol

how do you even reverse engineer something we have all the data for easily accessible ...

 

[ty_ger]

Yep, because thousands upon thousands of cards are having problems.

Funny, AMD is a PCI-SIG member, you would think they would say something - even through back channels - that there is a problem with Nvidia’s implementation of the specification.

Or, it could be there’s not an issue and people are falling for internet sensationalism.

It doesn't take thousands of failures. All it takes is a large group who could be affected. This is a large group, and the design problem affects everyone similarly. All that needs to be argued is that it is a safety problem that they failed to prevent due to negligence. Easily provable. But much harder when it is an economic giant.

 

[Visible Noise]

It doesn't take thousands of failures. All it takes is a large group who could be affected. This is a large group, and the design problem affects everyone similarly. All that needs to be argued is that it is a safety problem that they failed to prevent due to negligence. Easily provable. But much harder when it is an economic giant.

What safety issue? What large group?

Do you know nvidia literarily sells millions of video cards a year? If there was a safety issue it would be headline news, and reported by the CPSC:

Recalls & Product Safety Warnings

www.cpsc.gov

An echo chamber of 100 cases out of millions cards sold proves nothing, especially when it’s not happening to pre-builts.

Actually I need to correct myself. It proves there’s not an issue that can’t be chalked up to user error.

 

[Bill_Bright]

there might be a class-action lawsuit against Nvidia over this particular design at some point.

I'm not certain NVIDIA would be the proper defendant here. Does NVIDIA, a California based company, actually manufacture anything? I don't think so.

As far as I know, they do the R&D on the chips but pretty much outsource everything else, including the manufacturing and design of the cards their chips will be mounted on.

In fact, they have a name for those companies, AIBs (add-in board partners). It is these AIBs who take the GPUs from NVIDIA (which were actually made by TSMC in Taiwan), mount them to their boards, add their own video memory, power delivery, cooling solutions, etc. (source).

That explains why boards made by ASUS, Gigabyte, MSI, etc. that use the exact same GPU, have a different component layout and perform differently too.

If NVIDIA is supplying ASUS and Gigabyte the GPUs only (through TSMC's fab labs), and simply giving those AIBs the specs (including the power requirements) and it is the AIBs who are deciding what connectors to use, then I would think it is the card makers who should be sued, "IF" a suit is even warranted.

As ShrimpBrime noted, a 96.7% success rate is pretty good, and on par with other similar products so again, not sure a lawsuit is warranted. Now if the card makers refused to honor warranties, or if resulting heat issues resulted in house fires (or worse) and the companies refuse to do anything about it, deny the problem, or worse, try to cover the problem up, then that is a different story.

@ShrimpBrime - It would be interesting if you refine your AI search to break the results down by brand.

 

[ShrimpBrime]

@ShrimpBrime - It would be interesting if you refine your AI search to break the results down by brand.

The search engine does give links to the data its pulled from.

But as it mentioned a reddit megathread, which I have looked through several times, is 95% bullshit comments and 5% actual data. Forgive those numbers, thats a guesstimate on my part.

All this is similar to Intel 13th gen processors failing, but it originated with mobile 13900H/X processors that some how blew up to desktop chips. Also very low numbers actually reported that would be considered a confirmed case.

None the less, its all for glory, sites like GN are after. Sure they dig for data, but the main goal is to get the public in a ruckus for views which brings a revenue.

Concluding: I don't feel the need to dig deep for actual confirmed cases of 12+4 connectors melting.

 

[Bill_Bright]

Concluding: I don't feel the need to dig deep for actual confirmed cases of 12+4 connectors melting.

I don't blame you. I just wonder if this is another case of a report of a connector melting, and then every blogger, IT media outlet, and you-name-it parrots the report and suddenly we have 100s of reports of the same incident, that then is reported as 100s of incidents, and so on and so on.

I am sure there is more than one - just using that to illustrate.

 

[Shrek]

And if I just solder two 8-pin connectors there it wouldn't work, right?
View attachment 402614

I wonder if the two 8 pin connectors can be used to supplement (not replace) the main connector; if so, this would seem the best solution to me.

 

[lukart]

is there even much benifits to the connector to make some AMD partners use it?

 

[Macro Device]

is there even much benifits to the connector to make some AMD partners use it?

No, this connector shouldn't have existed. Utter rubbish.