Hmm, well - that was addressed quite some time ago (faulty component (capacitor? I forget now)). A search should have brought that up for Halo.
 
You had an original run with the problem, it was addressed and fixed. You might get an RMA'd card that still has the issue (but I doubt it, as EVGA should have been able to identify the batches and fix them) - if it happens again just ask nicely for EVGA to cover the shipping costs. They've done so for me and others with repeat problems in the past.

I saw that the original issue with a weak solder joint on a mosfet was taken care of, but I’ve also seen issues with 3080ti having this issue with halo mcc, and I’ve seen 3090s drawing 135% power limit, or the cards pulling 75 - 85% GPU power while the limit is set to 50%. It’s just a terrifying thought. I had seen where other brands gpus stayed much closer to the power limit.

I hope the issues have been fixed because I would prefer to stick with EVGA. I have a 2080, 3080 ti and 3090 from EVGA. I just don’t want to deal with continuous RMA issues, and the constant fear/possibility of a failure dealing permanent damage to other components.

Go ahead and make an RMA and start this process. 

I don't think it has ever been fixed. The Rev 1.0 just balances a little better and is less likely to die, in my opinion.

I never for a second believed the "bad soldering" excuse and supposed bad mystery batch which EVGA never would provide any more specifics about. If it were actually a bad batch, why is EVGA not telling us the batch number, why are customers with affected batch numbers not notified, and why is EVGA still sending out Rev 0.1 cards via RMA which fail the same way? If a bad batch of soldering was the cause, EVGA should have been able to mitigate it long ago.

Be careful about the "different account" thing. That is against the terms of service of this website. You should work with EVGA to get your accounts consolidated to only one.

Edit:
BTW, do you have a link to that Buildzoid video about power balancing exacerbating the issue? I started to be convinced that was the case, but haven't heard anyone else say it. I am convinced that NVIDIA's choice to use average power is at the heart of it, NVIDIA's choice to have no instantaneous power limit of any kind is at the heart of it, transient loads kill the cards, cards with higher power limits with reference VRM are more likely to die, cards with poor power balancing are more likely to die since more of a transient spike will be provided by one bus instead of being better distributed among the 4 buses, and Rev 0.1 cards have poorer power balancing due to their inferior analog voltage controller chip.

Also, just for a matter of ranting... EVGA created a special 3090 OC exchange program where people could send in their crappy Rev 0.1 card which balanced badly and exchange it for a Rev 1.0 model which balanced better (and therefore power limited less often/ performed better). Where do you think the Rev 0.1 cards that people sent in are going? To poor people like you with a RMA. So, if you get a Rev 0.1 from RMA, it is almost certainly one that EVGA previously accepted as unacceptable from a previous customer. What gives EVGA?!? EVGA doesn't want to take the loss, so they hope customers won't know better and will gamble with another one. That's why you have to specifically tell EVGA that you don't want a Rev 0.1 reject.

The cards blow up because their transient response requirements are too much for some power supplies. Old rev 0.1 cards use a pwm on the vrm that is too slow to respond to the severe overcurrent/undervoltage condition in time to save the card. Newer rev 1.0 use a faster pwm that usually triggers the protections in time.

Considering the PSU you used, it is closer to "it will probably take more time until it does blow up".

It is actually a matter between Nvidia and the PSU makers being shady on the way they interpret the ATX spec.

Evga just went and increased the power limit of their cards on top, so the issue becomes even more notable.

Nah, buildzoid is just rambling cause he does not have a working ftw3 card in hand to see what is going on and they fry. He specifically states that he makes assumptions about this at some point.

In reality, the affected cards just draw power so massively that some PSUs cannot give it in time. Just that.

ty_ger07
Bogus. Hook one up to the best DC battery source you can, and they will still have transient overshoots.
The transient overshoots are a product of the design and are guaranteed to happen regardless of which power supply you use.

I still fail to understand how you think a power supply that isn't able to keep up with demand can overpower a card. It is illogical. If your theory were true, the card would shut off due to undervoltage protection and would be perfectly fine.

It's overcurrent protection that the card is lacking, and that is what is killing the card when a transient spike is too high and the card has no built-in protection against such a condition.

At some point we have to stop blaming power supplies. It is a waste of everyone's time.

I am very specific about this situation, so do not put words in my mouth.

I put the blame where it is.
The power supplies were supposedly guaranteed to handle the aforementioned transient spikes. Some do not. It is that simple.

I’ve seen cards that have died regardless of psu choice. I seriously doubt PSU choice matters at all. Hence the redesign of the highest end variants from board partners power deliver systems. Suprim X, Strix, Kingpin, Extreme and Hall of Fame don’t have these issues. All the other models do.

Will EVGA do refunds instead of sending another card that will blow up out? I would much prefer money back, and I’ll just buy a 3090 TI since it has much better power management.

NobleHonor7
Will EVGA do refunds instead of sending another card that will blow up out? I would much prefer money back, and I’ll just buy a 3090 TI since it has much better power management.

"EVGA offers a 14-day return for refund policy from the date of purchase on items purchased directly by you from EVGA.com. EVGA does not offer any refunds on shipping once the order has shipped to the customer and the customer is responsible for any transit, customs, or brokerage fees due to a non-deliverable or rejected package." Per https://www.evga.com/legal/store/
 

NobleHonor7
Will EVGA do refunds instead of sending another card that will blow up out? I would much prefer money back, and I’ll just buy a 3090 TI since it has much better power management.

It took 3 RMAs for EVGA to give me a refund on my 3090 FTW3 Ultra

ty_ger07

Michapolys
In reality, the affected cards just draw power so massively that some PSUs cannot give it in time. Just that.

Bogus. Hook one up to the best DC battery source you can, and they will still have transient overshoots.
The transient overshoots are a product of the design and are guaranteed to happen regardless of which power supply you use.

I still fail to understand how you think a power supply that isn't able to keep up with demand can overpower a card. It is illogical. If your theory were true, the card would shut off due to undervoltage protection and would be perfectly fine.

It's overcurrent protection that the card is lacking, and that is what is killing the card when a transient spike is too high and the card has no built-in protection against such a condition.

At some point we have to stop blaming power supplies. It is a waste of everyone's time.

I tend to agree that it is less to do with the PSU and more to do with understated TDP requirements on the 3090's (in conjunction with sub optimal power delivery).  Too many people were limping in with bare minimum / maxed out PSUs to run 3090s which isn't a good idea. 
 
The capacitors on a maxed out PSU aren't big enough to handle those load spikes coming from the GPU.  Not the PSU's fault if the power demand goes way over the wattage limit of the PSU, even for a split millisecond. 

yaymz
The capacitors on a maxed out PSU aren't big enough to handle those load spikes coming from the GPU.  Not the PSU's fault if the power demand goes way over the wattage limit of the PSU, even for a split millisecond. 

That is the problem. The power demand is within the recommended limit as defined by the ATX spec. It is just that most PSUs currently out there rely on a loophole of the spec to reduce their transient load capabilities.