You know it: New technology comes onto the market, and someone always manages to either grill it within the first week or to flash it into the hereafter. But with Nvidia's flagships, this happens occasionally.
Because while players should only think about Framerates and not about fire protection, the RTX 5090 ensures that they have rare individual cases with its power connection - in the truest sense of the word.
A look back
Let's remember: The first two years ago diving first two years ago. Actually a problem that is known from sci-fi films:The reactor overheated!
But that actually happened - smeared cables, charred plugs, and even entire graphics cards and power supplies had to let their lives.
The explanation back then?Badly connected adapter or cheap cable from third -party providers. Anyone who used the original adapter from Nvidia and did not twist it too much saw themselves on the safe side.
That also applied to me. Sure, I was a little queasy, but as long as everything was infected properly, nothing should happen - at least that's how the theory.
But now, with the RTX 5090, the drama is going to the next round. But this time with even higher temperatures and therefore more melting potential.
Why have cables melted again?
In fact, the 12VHPWR plug was improved after the first melting problems of the 4090. Sensor pins have been shortened so that the contact only establishes when the cable is really bombable.
But lo and behold: .
Romanthe8ver
Hartung, hardware guru and hobby fire investigator, took a closer look. An affected user had given him a melted RTX 5090 Founders Edition with a burning power supply and cable. But also with his own test system.
Two of the six power lines get too hot - up to 90 degrees. And on the plugs themselves? Sometimes more than 150 degrees! This is no longer justwarm
, that is the moment when you ask yourself whether the graphics card immediately replaces the smoking oven.
The cause
But why does that happen? Although the connector is correctly sitting, individual power lines do not seem to have full contact. This can have to do with the fact that Roman as a tester often insert the cable into and out and therefore worn out.
In the cables of the user, a production error or an unfavorable combination of gold -plated contacts and tin may have had. Either way, in both cases, this obviously means that not all six 12-volt cables are loaded evenly.
And if one or two instead of six cables wear the entire load of up to 575 watts, there is suddenly a current, which is usually sufficient for a small saw. I speak of up to 48 amps instead of eight.
No wonder that the cable will eventually open in flames and tears the power supply and graphics card into the abyss.
A construction error?
In my opinion, the real error is not even on the plug or on the cable, but on the circuit board design of the RTX 4090 and RTX 5090.
There are a maximum of two shunts (small resistors for electricity measurement) for the 12VHPWR connection. Nevertheless, all six 12-volt lines ultimately run together-even if a shunt flies.
With the RTX 5090 Founders Edition, even a shunt is therefore installed because it makes no difference if in the end all lines lead together to the board anyway.
Means:The card has no idea whether a single line is just overloaded. Switch off? None! Instead, it moves on until it melts.
NVIDIA apparently rely on the fact that the revised 12VHPWR connection is sufficient as a security measure. Statistically, this should also be true.
But it used to be different:The RTX 3090 Ti, for example, has three shunt resistors on the 12VHPWR connection, of which the lines lead individually onto the board, which is why there were never any problems with the same power consumption as with the RTX 4090 (both 450 watts).
And this time the problem is potentially even larger than the 4090, because the RTX 5090 pulls up to 125 watts more.
What worries me the most: You just can't be 100 % sure whether everything goes well without checking. Even with an improved plug, there is apparently a residual risk - however small it may be.
Panic is nevertheless unfounded:For now, these are isolated cases that can be due to special circumstances that I have described above.
What to do?
My tip:If you have an RTX 5090, there is a stress test - for example with 3DMark Speed Way (for a fee) or Furmark (free of charge).
Then simply feel with your finger on the coats of cables and plugs. If you can easily touch them after long operation (15 to 30 minutes or more), this is a good sign.
However, this is not a guarantee, and in my opinion there should be in the form of suitable protective measures.
Asus, for example, recognized the problem with his ROG Astral RTX 5090 OC Edition and installed six additional shunts so that each line can be monitored individually. A good decision, because the card can react in an emergency, adapt the current flow or switch off.
I have already contacted Nvidia on the subject. If I get an informative answer, you will find out on Gamestar.de.
