jankerson

 
It might run longer than that, but they go out of spec, and that is the real problem.
 
Quality of the PSU doesn't matter here, even the best PSU on the market would need to be replaced in say 10 years.
 
It's actually scary to see how far they can go out of spec over time, and sometimes the protection doesn't work anymore either.
 
So if they go they can take the whole system them.

SladeX
A friend's 1k PSU has 4 PCIE ports on his PSU, but 2 are rated lower in amperage than the #3/4 plugs.

SladeX
Have both a XC3 and FTW3 3090. On some older power supplies, I've noticed that not all VGA/PCI power can output what the 3090 is demanding. A friend's 1k PSU has 4 PCIE ports on his PSU, but 2 are rated lower in amperage than the #3/4 plugs.
 
On my unit, I've seen a peak 475W draw from my FTW3 where PCI3 was pulling 160W.
 
This has been shown on my UPS as much as a 930W draw from the wall.

AWK16

DarkTAO
[Kurztest] Inter-Tech SL500A Netzteil Test auf Teststation SunMoon SM5500 ATE | Die Hardware-Community für PC-Spieler - PCGH Extreme (pcgameshardware.de) 
 
Is that you, or are you re-using images of others?
 

I can assure you its not his

nomad47
Wow it seems my PSU query stirred quite a debate here. Chill fellas. From an engineering standpoint, there is a life of every thing, PSU included and I think its prudent to change the component near End of Life (EOL). MTBF (mean time between failure) is another good metric to judge it on. Lets be realistic here, not everyone has the equipment or technical acumen to go on testing a PSU output, stability, variance, unless they do it for living.

z1nonly
Of all the protection circuitry that goes into these cards, I have a hard time believing the engineers assumed that everyone would use an excellent specimen of a PSU that not only met the wattage requirements stated by Nvidia, but then went on to ensure that their PSU *also* met some unlisted, transient-response capabilities that other cards have hitherto managed to survive without. 
 
These cards should not be so sensitive and fragile. 

jankerson

z1nonly
Of all the protection circuitry that goes into these cards, I have a hard time believing the engineers assumed that everyone would use an excellent specimen of a PSU that not only met the wattage requirements stated by Nvidia, but then went on to ensure that their PSU *also* met some unlisted, transient-response capabilities that other cards have hitherto managed to survive without. 
 
These cards should not be so sensitive and fragile. 

The 3080 and 3090, especially the FTW3s pull a lot more power than the 1000 and 2000 series cards. Which also means the transient power spikes will also be higher. And they are a lot higher.
 
Nvidia has no control over what the PSU manufactures do or don't do when they design PSUs.
 
It comes down to PSU design really in the end so doing some research and checking professional reviews before buying a PSU is important.
 
In the end it's important to get a good high quality PSU, not the cheapest one in the list.
 
A difference in even $20 in price can make the difference in it working and not working.
 
 
 
 

jankerson



 
MOST as in almost all of the shut down issues are related to PSU design and issues.
 
Some models just can't handle the transient power spikes, that's a PSU design problem, not a GPU problem. 
 
 

marcula

jankerson



 
MOST as in almost all of the shut down issues are related to PSU design and issues.
 
Some models just can't handle the transient power spikes, that's a PSU design problem, not a GPU problem. 
 
 

What data are you basing this statement on?
 
I've seen reports from folks with dead cards who are using relatively new high-end PSUs, e.g. the OP from the "3090 FTW3 Ultra Dead After 2 Days" thread was using a 2 year old EVGA 1200w P2. 
 
 
This is all pointing to a straight up product defect with a batch of FTW3s, and not some unicorn PSU that's needed to run it. Causing a PSU to trip or the system to shutdown because a PSU can't handle the power draw is one thing-- but I think *actually smoking* and completely bricking is another. I don't know if other AIB cards are having the same issues, but I haven't heard anything close to the number of failures being reported on this forum yet.