Interesting speculation:
EVGA GeForce RTX 3080, RTX 3090 and (not only) New World - when design flaws cause graphics cards to run amok | Tests | igor´sLAB
Disclaimer: The following article is machine translated from the original German, and has not been edited or checked for errors. Thank you for understanding!www.igorslab.de
I find it hard to believe the fan control would cause this issue. But I find it easy to believe the fan speed sensor, which detects pulses from the fan rotation, could be spoofed by graphics card wide voltage noise caused unusual load from a different part of the card.
Relevant video from someone familiar with CPU/GPU power delivery.
Gets good at about the 20:00 mark. 6-phase power delivery for the . . . MSVDD? Sorry having problems understanding what he's saying there. Anyway looks like somehow New World was overloading one of the memory power rails.
I think the main take away from Buildzoid's video is that Nvidia's power protections aren't working or aren't responding fast enough and that New World can't be blamed for a probable design flaw. Also, Nvidia signs off on all partner board designs so any design flaw, ultimately, lies with Nvidia and not EVGA or Zotac, etc.
Well that's part of it. But I think @aigomorla was partly right or at least gave us a hint at what's going on: it's something to do with the way New World is hitting the card's VRAM without really stressing any other part of the card.
While I am right along with everyone being concerned about the rear VRAM temps being too high, for me it is more of a concern about longevity.
If it was just a matter of heat, the memory would throttle. But if throttling wasn't possible or wasn't sufficient because of some unusual electrical circumstances due to a design flaw in power delivery, then the VRAM overheating is a symptom rather than a cause of the failures.
I agree. What aigo was pointing out is that a lot of power can be pushed to the RAM. But what happens if that occurs when the rest of the card isn't really being stressed? Like when sitting at a menu with unlimited fps?
I agree, but that is kind of my point. If Nvidia/EVGA/Zotac/ASUS/etc. designed their cards in such a way that the situation you are describing is a possibility, that is a design flaw/oversight. Nvidia and their board partners know what each component's power limit is. If they only have a whole-of-card power limiting setup and not per-component, that is a design flaw IMO.
My interpretation of the Buildzoid video is the exact opposite of that: the card HAS the ability to enforce current limits on a per-rail basis, but for whatever reason it fails to do so. Buildzoid shows manufacturer quotes for the 3 most commonly used voltage controllers in 3090 cards, and they all have configurable per-phase over-current protection.It is. It appears as though the card has no way of enforcing individual power limits on a per-rail basis, which is weird since each rail has its own VRMs which should be capable of such.
My interpretation of the Buildzoid video is the exact opposite of that: the card HAS the ability to enforce current limits on a per-rail basis, but for whatever reason it fails to do so. Buildzoid shows manufacturer quotes for the 3 most commonly used voltage controllers in 3090 cards, and they all have configurable per-phase over-current protection.
So, for what we know in the current situation, are those failing cards just junk or are the games junk ? Or both ?
At this point it looks like a faulty hardware issue that collides with improbable (but possible) software, such as menus with unlimited framerates.
That seems very likely to me.More faulty solder or part(s) because cards have been in such a high demand I am confident corners are being cut to meet demand.