News RTX 4090's 16-Pin Connector Melted After One Year of Usage

[InvalidError]

That's fair but it's still a Mickey-Mouse way of handling it because there will still be cables to deal with. I think that it would be more advantageous just to add this power-handling capability to the motherboard itself. Then there would be no cables at all. ASUS' "power toe" idea is actually a great idea, if something like it could be standardised under the ATX umbrella.

SIs don't care that there are still cables to deal with, they care that fewer cables means lower overall costs. Delegating power distribution to the motherboard means likely more expensive motherboards, which in turn likely increases overall costs, which is a bad idea when mainstream buyers are already complaining that motherboard costs are out of control.

Also, the PCIe slot and anything that might become a standard extension to it would be PCI-SIG territory, not Intel/ATX. The case and PSU specs have never had anything to say about board-to-board connections, the PCI-SIG dictated the 6/8-pin PCI-AUX power connectors that went into the ATX PSU spec, just like it did the HPWR spec in PCIe5 and ATX 3.0.

 

[Avro Arrow]

SIs don't care that there are still cables to deal with, they care that fewer cables means lower overall costs. Delegating power distribution to the motherboard means likely more expensive motherboards, which in turn likely increases overall costs, which is a bad idea when mainstream buyers are already complaining that motherboard costs are out of control.

Also, the PCIe slot and anything that might become a standard extension to it would be PCI-SIG territory, not Intel/ATX. The case and PSU specs have never had anything to say about board-to-board connections, the PCI-SIG dictated the 6/8-pin PCI-AUX power connectors that went into the ATX PSU spec, just like it did the HPWR spec in PCIe5 and ATX 3.0.

I suppose that makes sense, I just don't understand how it managed to be so problematic on the RTX 4090. It's not like the other RTX cards that use it had these issues (at least, not nearly to the same degree).

 

[InvalidError]

I suppose that makes sense, I just don't understand how it managed to be so problematic on the RTX 4090. It's not like the other RTX cards that use it had these issues (at least, not nearly to the same degree).

Manufacturers can screw up their connector designs and fail to meet the spec with the intended safety margins, cable manufacturers can screw up their cable assembly in a way that invalidates the connector manufacturers' assumptions. User error can invalidate everything manufacturers assumed even when manufacturers did everything right.

Hopefully the PCI-SIG and others have learned that splitting high current across so many small pins might not be a good idea and HPWR's successor will be somewhere in the XT60-style direction.

 

[Brubbler]

Just what I want! A blatantly faulty gpu that may burn my house down from a company who shows zero accountability. Slow clap Nvidia, really inspires me to buy your product. How has this not yet resulted in a broad legal action?

 

[InvalidError]

Just what I want! A blatantly faulty gpu that may burn my house down

The connectors may melt but nobody has ever shown one actually catch fire, which is the whole point of requiring that the connectors and cables be made with 94V-0 compliant materials which cannot sustain a flame, as pretty much everything else that goes inside a PC. If you want to start a fire with the connector, you'd have to put cotton balls on to help it along.

If you watch a video of a 94V-0 compliant cable or extension cord getting grossly overloaded, the insulation will melt off or turn to ash with the wire glowing red-hot but still no flame.

 

[purpleduggy]

told you so