Here we go again, another person who assumes that its about maxing the card out. Its not, its about the transient power spikes.
If your card uses the 12vHP connector (the original version from the 3000 series), and that connects back to say 2x 8-Pin on your PSU and you run a really awesome multi-rail PSU. If the card for just a 100ms or so asks for like 1200w as a spike, it will certainly trip OCP on the PSU. OCP has like a 5ms response time or something like that from memory.
From what I have read its always Gigabyte cards (very few other brand of cards, the more expensive ones). Where I am from in AU, Gigabyte is one of the cheaper brands, and you can see it in the component quality on the cards.
A piece of software cannot physically kill a piece of hardware unless its maliciously designed to override the hardware protections. I somewhat doubt that D4 has malicious code in it that is just trying to override the cards limits.
Most likely the cards in question that are being bricked have inherent faults in thier quality of construction that just somehow they cant handle the loads being asked for by the game engine.
Its not about giving maximum sustained load to the GPU, its about those little spikes that can absolutely bring even the best PSUs to thier knees, if at the PSU end you are connected to 2x 8-Pin connectors and the card requests a momentary spike of 2.5x its max load, chances are its going to cause problems.
Who’s to say that the card has not been overstressing the PSU for a while with spikes, and in this case the PSU has finally said enough is enough and its ended up killing the card.
I wished people would stop blaming games for hardware failures. Its more than likely co-incidence that it just happens to be when you are playing D4 that the card dies.
The game is in Beta, if you dont want issues wait for official launch.
Can you imagine the size of the class action if someone reverse engineered the D4 beta code and found that they had some form of malicous code in there that was killing GPUs by pushing them beyond thier limits?
We have known now that RTX30 and now worse off RTX40 series cards are great for performance, but they are better at generation astronomical transient spikes. Unless you have the professional monitoring gear connected to your card to measure the spikes, you are not going to even notice they are happening until something else in your machine says “not happening” and the card stops working.
My bet is that the card requests the spike, the PSU finally says not happening, the card chucks a hissy fit and its that resulting time period after the PSU refuses to supply the spike (when OCP is tripping) or the power provided is momentarily not as clean as one would expect from a super high grade PSU that causes damage to the card.
18 months ago it was New World killing GPUs, now its D4, next year it will be another game. From memory back when New World was killing eVGA 3090’s some people pointed out that the cards themselves had a lower quality set of caps on the back of them, and that it was not happening with the higher quality set of caps.
I have never seen a Gigabyte card with all high quality SMD caps on the back of the card around the GPU area. They always use the cheaper ones. Mostly ASUS ROG STRIX OC cards use the higher quality caps, or at most 1 of the cheaper ones and the rest are the smaller high quality ones. Gigabyte mostly always uses 5 or 6 of the cheaper larger ones in place of the more expensive better quality components.
When eVGA started sending out cards with less of those cheaper caps, the issue seems to dissapear. So was it really the games fault? Or did the game just find a scenario where the card couldnt cope due to poor assembly?
My vote is on the latter.