Probably Bound to Happen: Increased Power Draw for DLSS 5 Reportedly Claims Its First Victim After a User Finds an RTX 5090 with a Melted Power Ada...

Brian_B said:
Well, had they done the testing that way, then it blew up later on without the test equipment - that would also be news-worthy. And even if it didn't - still would be very interesting to know, particularly since this issue is very much out there.

But they didn't, did they. Apparently they didn't do any of that valuable meltdown-related data, and they were focused on performance-related metrics. Which is why I asked how measuring meltdown-related data would affect that.

If you are reviewing a card with a 12VHPWR connector and you aren't evaluating volts and amps and balancing external temps and meltdown-related items... you really aren't doing your job. So many of those cards have very publicly went up in smoke that I would consider that data essential to a GPU review... if you have that connector. And interesting, but non-essential, if you don't have that connector.
Click to expand...
they should buy a brand new 5090. test dlss5 in various games. establish when the connector blows up. publish that as a video
 
I think bad case airflow is also a factor. If there is no serious air current pushing the heat out, the GPU will slowly get cooked. I can't recall if there's been any reports of a liquid cooled 5090 failing.
 
igor_kavinski said:
I think bad case airflow is also a factor. If there is no serious air current pushing the heat out, the GPU will slowly get cooked. I can't recall if there's been any reports of a liquid cooled 5090 failing.
Click to expand...
A quick gemini query says yes this does happen. VERY rarely... and lets be real. More people dropping a high end card in that are experienced system builders are doing it with VERY solid connections the first time, NOT doing multiple re-seats 'because it might be having problems.' and generally have solid cooling.
 
Usually, for electrical fires - air flow often makes the problem worse by spreading the spark/heat out faster

Electrical is usually a high impedance/resistance connection with just too many amps going through it (P = R * I^2), and it's a very small, very hot point source - and that spreads to nearby material which can ignite or melt. Air flow ~could~ help but most of the time it's buried inside a connector or a junction box or something, and blowing air across it just makes the flame spread faster when the material ignition eventually occurs.
 
Brian_B said:
Air flow ~could~ help but most of the time it's buried inside a connector or a junction box or something, and blowing air across it just makes the flame spread faster when the material ignition eventually occurs.
Click to expand...
Yes but I was referring to the local heat build up. Glass cases are especially bad in trapping heat. The airflow may help reduce the likelihood of reaching the ignition point.
 
How do you thermally protect something so localized though, and that can happen quite fast i suppose. I know !! Need a combination gfci, afci overload protection breaker... Yeah thats it a giant breaker to go with the tiny connector.
 
Uvilla said:
How do you thermally protect something so localized though, and that can happen quite fast i suppose. I know !! Need a combination gfci, afci overload protection breaker... Yeah thats it a giant breaker to go with the tiny connector.
Click to expand...
From reading, it sounds like they put a thermocouple in the plug itself. But yeah, just looking at typical amps/volts, nothing there really protects you from high point source resistance.
 
Standard Molex was never immune to meltdowns either - it certainly can happen. Mostly it was with cheap extension cables or cheap no-name power supplies though - so it was easy to point the finger at a manufacturing flaw rather than a design defect.

12VHPWR though - doesn't seem as if anything is immune, despite all these protections and theatrics that people are trying to avoid it - which points at a design flaw.
 
Now that they've been out a while, time seems to be a factor in how long they will last before failure as well. /shrug

I think the proof is in the pudding on this connector.
 
Grimlakin said:
This right here makes me want to think about undervolting. Sigh...
Click to expand...
Started this with mine, maybe a month or two out of the gate and that combined with power limiting and sometimes FPS limits. More than anyone should ever have to do but nearly two years in with them and everything still going smoothly. The reward is more than just keeping these alive though as I've seen NV really screwed up the power profiles with them. I and more than a few others out there think this was just to claim high clock records.

There are many posts out there explaining how you can get at least 90% or more performance while dialing things back and I've often seen between 250W-350W doing nearly the same performance as when it was drawing 450W to over 550W. DLSS 5 makes this more challenging but users can still dial in to keep in the 250W-400W range and get good performance. The stock power profiles are a bad joke imho.

Edit: Also thinking very heavily about getting one of these to slightly lessen the workload while still looking aweseome, was pricing over the weekend and saw them for between $750-$800: https://www.msi.com/Monitor/MAG-341CQP-QD-OLED-X28
 
You must log in or register to reply here.
Become a Patron!
Back
Top