NVidia GeForce RTX 40x0 cards - x2 the power consumption, x2 the performance

The problem appears to be that it burns if its slightly out of its pin connector and bending the cable lets it easily slightly slip out (THATS WHAT SHE SAID) without you noticing if you don't have a volt meter attached.

i'm sure they will figure out cable and PCB layouts that don't "make" you bend the cable, stronger connects that lock pin connections, etc but at this point you are basically alpha testing

Mist said:

People with ATX 3.0 PSUs aren't having this problem, right?

Click to expand...

Surely cable quality matters, but they should still have problems.

The size of the cards require tight radius bends on the cable-to-card connection to fit in most cases. This bend means less contact area for the pins on the outer radius of the bend, so these pins push the same 8+ amps but have less area to spread the heat generated. Keep in mind that the 12vhpwr connector uses 6 12v pins for up to 600 watts of power, where as the old 8pin PCIe plus used 3 for 150, so high end cards had 3x8pin or 9 pins for 12v delivery - IOW, the power and heat is more localized on the new cable..

The spec says to give the cable 35mm of length coming off the card before starting any bend. Given the height of the card this is impractical for most cases. If concerned one work around is the leave the side panel off to give the cable that 35mm of space.

Big Phoenix said:

Jay just did a video doing that trying to replicate these burnung cables and couldnt get it to happen.

Click to expand...

Is it a quick and easy replication of the issue? Or just throwing numbers out there, a few days cumulative...melting before burning?

I do not think anyone truely knows yet. Buildzoid suspects that the bend is putting excessive strain on the female receptacle, bending the receptacle at the seam and making the connection loose. That, plus the bend pulling the pins on the outer radius out make for loose/less connection area. I would be interested in seeing close-ups of the connector on the card along with the melted cable connector.

Jovec said:

I do not think anyone truely knows yet. Buildzoid suspects that the bend is putting excessive strain on the female receptacle, bending the receptacle at the seam and making the connection loose. That, plus the bend pulling the pins on the outer radius out make for loose/less connection area. I would be interested in seeing close-ups of the connector on the card along with the melted cable connector.

Click to expand...

Anytime an electrical connector/pin or whatever is loose it generates a ton of heat. We routinely check our electrical cabinets for hotspots using an infrared camera to find all the wires that weren't tightened down correctly. More than likely this is the case.

However, this could be due to pins not being within spec and not fitting properly within each other and the stress on the wire is compounding the issue. With the wattage these cards pull if the card senses it's not getting enough juice it's probably pulling even more power due to the loose connection and making the issue even worse.

mkopec said:

Why cant they just relocate the fucking plug on these cards to another area? ITs just shit design and hasd been for a long time. Or make a 90deg connector or some sort of add on 90 piece that goees in the middle of the plugs from the PS and the card? It cant be more than like a .25c piece.

Click to expand...

Ive been asking why they havent put that fucking connector down at the bottom on the END of the card going towards the front of your PC for fucking ages. Running that cable up to the top of that card looks fucking ugly no matter how well you cable manage. Or even just the top of that end if they cant have the current travel from that direction because then its within relative range. Its not like Nvidia gives a fuck about their board partners designs etc.

Igor has a great article on this

Adapter gate? NVIDIA briefs all board partners this morning and makes damage an absolute boss issue | igor'sLAB (igorslab.de)

So, regarding the PC I ordered. I already changed it from a 12900 to a 13900k, but I asked them if they knew when the ATX 3.0 were available, and they didn't have any or know when they would get any. They said then if I wanted that, I'd just have to cancel my order and purchase it again when it's available.

So, should I do that?

Mist said:

I'm not building anything until some company that doesn't suck puts out a good Titanium ATX 3.0 PSU.

Click to expand...

I canceled it right after I wrote the above. Not taking the chance even if the chance is minimal. Hopefully it won't be too long until they're available.

Kirun said:

That begs the question of why you'd need a 13900k at all then, no?

Click to expand...

Just to tell other people you have a 13900k and 4090. Its 90% bragging rights.

Daidraco said:

Ive been asking why they havent put that fucking connector down at the bottom on the END of the card going towards the front of your PC for fucking ages. Running that cable up to the top of that card looks fucking ugly no matter how well you cable manage. Or even just the top of that end if they cant have the current travel from that direction because then its within relative range. Its not like Nvidia gives a fuck about their board partners designs etc.

Click to expand...

That end of the card is all cooling stuff. They could run an extension down to the end, but $.

External power pack would be a good idea at this point. $ though.

I'm of the opinion that the whole mess inside needs to be tossed out and redesigned, including all the dumb plugs and piles of wire everywhere.