It is partially EVGA at fault here, at least on cards with increased power limits compared to the reference design. The reference design also has its own share of issues.

Unfortunately there is no one available to talk about the issue, the cause and possible workarounds.

lets just agree theres an issue with the 3000 series across the board.

The FTW3 GA102 cards were flawed Day 1 - this is well known and has been proven even though there's no tacit acknowledgement from EVGA directly.  Unpacked BIOS revealed the design of the FTW3 initially was intended to be 2x8-pin and 1x6-pin PCIe power connections - this has totally led to the uneven power draw issue.  Some of the component choices are less than ideal like the use of analog VRM controllers.  All of these combined have contributed to the red LED of death - EVGA also chose to use fuses on the PCIe rails, and the bad power distribution kept blowing fuses.
 
It was far worse for the 3090 (top end card) than the 3080 but both cards IMO suffers from the same drawbacks.  Later revisions have sought to correct these problems but the weakness of the initial design choices has crippled EVGA's ability to really move forward, a total ground up redesign probably would cost them too much.  There's other weird EVGA choices that causes problems for enthusiasts - like the odd height thermal pads.
 
Bottom line - you can fall back on the warranty, but EVGA's cards this generation save for the Kingpin are not best in class.  XC3 cards and non GA102 are totally fine in regards to failure rates though - its the high profile top end cards failing that has raised eyebrows.

Please add me to the list of those who have experienced GPU failure. My RTX 3070 Ti FTW3 just died this morning. I'm glad I went with a company that has decent customer service.

neteng101
The FTW3 GA102 cards were flawed Day 1 - this is well known and has been proven even though there's no tacit acknowledgement from EVGA directly.  Unpacked BIOS revealed the design of the FTW3 initially was intended to be 2x8-pin and 1x6-pin PCIe power connections - this has totally led to the uneven power draw issue.  Some of the component choices are less than ideal like the use of analog VRM controllers.  All of these combined have contributed to the red LED of death - EVGA also chose to use fuses on the PCIe rails, and the bad power distribution kept blowing fuses.
 
It was far worse for the 3090 (top end card) than the 3080 but both cards IMO suffers from the same drawbacks.  Later revisions have sought to correct these problems but the weakness of the initial design choices has crippled EVGA's ability to really move forward, a total ground up redesign probably would cost them too much.  There's other weird EVGA choices that causes problems for enthusiasts - like the odd height thermal pads.
 
Bottom line - you can fall back on the warranty, but EVGA's cards this generation save for the Kingpin are not best in class.  XC3 cards and non GA102 are totally fine in regards to failure rates though - its the high profile top end cards failing that has raised eyebrows.

To some extent I agree, but there are some things that don't make sense though.  A bunch of the 3090 cards use the uP9511R controller, not just EVGA.  In fact it's the most common.
 
https://www.overclock.net/threads/official-nvidia-rtx-3090-owners-club.1753930/
 
The power balancing might be contributing to the failures, obviously doesn't help.  But none of the cards pull enough power on any of the 8pins with even the 500w XOC to explain failures.  This definitely does not blow the 20 amp fuses as none of them get to 240w on a single 8pin (or 120w slot on the original rev of the 3090).  The only way they blow is for something else to fail (or run a 1kw bios possibly).

Fennario
Statistically invalid basis for comparison - user forums have different levels of participation and typically consist of complaints.  EVGA gets even more due to their excellent RMA and customer service (people post knowing there will be a resolution). 
 
A proper metric would be each manufacturer's total number sold/total number RMA.

I think a lot of the forum participation here at the moment is due to EVGA instituting the queue system. That keeps people checking in to see how the queue is moving and having questions about how it works and all that.

While I agree the cards have issues I can tell you from first hand experience trying to get an rma from asrock or asus is not easy…. So you better hope you don’t have an issue if you are going non evga

My 3090 FTW3 Ultra failed after about a year of use. Luckily I had purchased the advanced rma on the card when I bought it, and the rma and replacement process through EVGA was very quick and pain free. It could not have gone any easier. Replacement has been doing great so far.
 
I have been buying EVGA GPU's since the 280 card, and this was first one that ever failed on me.

I’ve been reading bad card and RMA threads for many months, but my chief complaint is more generic. The 3080 and 3080 Ti FTW3 models particularly run very hot. Yes, case cooling plays a factor in keeping the cards cool, but companies really need to start focusing on power efficiency to go along with improved performance. It’s getting to the point where water cooling is becoming a must if one doesn’t want their GPU fans sounding like a turbine jet engine when gaming or under a heavy load in general.

ObscureEmpyre
I’ve been reading bad card and RMA threads for many months, but my chief complaint is more generic. The 3080 and 3080 Ti FTW3 models particularly run very hot. Yes, case cooling plays a factor in keeping the cards cool, but companies really need to start focusing on power efficiency to go along with improved performance. It’s getting to the point where water cooling is becoming a must if one doesn’t want their GPU fans sounding like a turbine jet engine when gaming or under a heavy load in general.

The fact that this thread exists falls on the engineering team, and even more so on EVGA_Jacob (a product manager in name only), for letting that team's failings bloom into this continued brand reputation decay.
With that said, I think the fact that these air-cooled SKUS can pull 450W cant be blamed on EVGA, as they had to release BIOSes that allowed them to be competitive(with equally overwhelmed air flagships from others). Short of making FTW3s hybrid only, I don't see how any three slot air cooler can handle what these BIOSes are allowing any better. I don't think an FTW3 air-cooler without assembly faults is any worse than competing 3-slot air coolers from others, with perhaps 3090FTW3 VRAM cooling's being an exception.