Except we do know. Because der8auer already made a video about it....and found, using his own 5090 FE for testing, that one of the 6 pins that actually carries power was pulling 23 amps, while another was pulling 20amps, while the remainder were pulling almost NOTHING.
Had load been properly distributed, each pin would pull 8 amps to hit the rated 575w, and no more than 9.5 amps if it were hitting the maximum of 680w the connector is technically rated for. It took a couple of minutes for the temp of the wire carrying that overload to hit 90c.
How exactly do you think someone is going to "overclock it within an inch of things"? The power slider tops out at 104% on FE cards, which would still "only" be 598w.
JayzTwoCents likewise already found that his AIB 5090, at stock settings has a default power draw of 630w. Which is....ya know, too high. While Igor's lab testing has shown the 5090s to have a burst power draw of 738w.
It's a bad design. It was a bad design on the 4090 when 450w cards were burning up, it's a worse design on a 575w card.
I believe the issue was long term using, especially in a SFF case where heat is trapped inside literally like an oven and things sprial to oblivation soon.
De8auer and the redditer might have been unlucky on the FE having a slightly defective socket, or that the cable is literally unusable after a few cycles of unplugging and re-plugging in and leading to resistance difference.
Some other respected guys like Wendel from levelone tech and Falcon NW have tested their own setups with 5090FE and luckily they didn't see the imbalance draw on wires
But either way, even assume both cases are slightly worn cables without apparent sign, there are few issues IMO:
1) As Buildzoid explained, the GPU design literally have no warning mechanism to just bloody cut the power draw and beep the s__t out your pants if such imperfection occurrs, be it worn out cables or manufacturing defect of the socket/cable
2) Redudancy is way lower than the previous 8pin standard
And even with those "balanced power delivery" cases, you can see the connector temperature is literally going from 68C to 80C in full load, and when one is measuring, it's either open bench or a tower case with open side panel, aka semi-open bench, and these temperatures are a big red flag at least to myself, it's now still winter/spring in the US, I assume ppl test it in heated room temp of ~21C, at summer you can easily put it at 25-27C for normal operation, and if ppl live in tropics without AC, it could be 30C+, which will put 10C on top.
Then there's the heat dumping of the card to the air inside the case, as most gamers likely don't put the fans in jet engine mode, 500W+ of heat circulating inside would easily rise the internal air temp another 10-20C (guesstimated from reports that in a FE card using with a 9800X3D compared to 4090 or similar raise the air cooled 9800X3D by 20C and throttling)
Adding these two together you could've encountering 85-100C on the exterior of the plug, and we know that plastic is a good heat insulator, so the internal temp is of anyone's guess, which doesn't sound good anyway...
And more ridiculously... putting the same shunt resistor circuit inside the GPU as in the 3090Ti config could at least save the card from fire, and from the good old days going from Molex 4pin in the 9700pro subsequently to 6pin and then 8pin and MULTIPLE 8pin have already got a good old solution... when you are too close to the rated capacity of the connector, just get another one, EVGA did that in kingpin 3090Ti, I don't see any reason of having the 5090 to use a single connector and call it a day
Facebook
Twitter
Instagram