Geforce 3080 issues

[LN400]

First, I don't know if this, or the Science sub forum is the right place for this so mods, feel free to move it if it belongs over at the other sub forum.

 

Now, there might be a good chance some of you have your eyes on the new 3080 family of graphics cards from Nvidia/multiple manufacturers. Some of you might have heard these cards are reported to have issues leading to the GPU crashing. I am more than a bit confused since the internet is going all out in reporting this, and the responses from Nvidia and the manufacturers,  with some new reports only repeating old news posing as new, some reports contradicting others etc.

So far, I've gathered, there is a problem with 6 capacitors, more specifically what type exactly, of capacitors the various manufacturers have gone for. At first, it was reported that ASUS cards did not see the same crash frequency as other brands but just today I read a new report stating that ASUS cards crashed just as much as any other brand. Some (honest) reports say the capacitors MIGHT be the cause, others say they ARE the cause case closed, and yet others say the capacitors have nothing to do with it.

Have any of you followed the 3080 debacle, and if so, what are your thoughts on this whole thing?

Edited September 28, 2020 by LN400
typo

[James Kerman]

Basically some manufacturers are using 6 all in one power capacitors instead of 4 all in ones and 2 arrays of 6 individual capacitors. This video explains the problem and shows the different configurations:

 

[LN400]

35 minutes ago, James Kerman said:

Basically some manufacturers are using 6 all in one power capacitors instead of 4 all in ones and 2 arrays of 6 individual capacitors. This video explains the problem and shows the different configurations:

That was the video that piqued my interest in the issue  Thing is, there are now coming out reports from several places saying this might not be the issue after all. In this article https://videocardz.com/newz/manufacturers-respond-to-geforce-rtx-3080-3090-crash-to-desktop-issues it is suggested that the capacitors are the cause but nothing is. as far as I'm able to tell, for sure and the cause could be something else.

 

EDIT: A follow-up article from Igor's Lab https://www.igorslab.de/en/nvidia-geforce-rtx-3080-und-rtx-3090-and-the-crash-why-the-capacitors-are-so-important-and-what-are-the-object-behind/

EDIT 2: It's also worth noting that Igor's Lab's report from Sept. 25th https://www.igorslab.de/en/what-real-what-can-be-investigative-within-the-crashes-and-instabilities-of-the-force-rtx-3080-andrtx-3090/ never states that the capacitors definitely are the cause.

EDIT 3: In this twitter post, it is claimed that ASUS cards crash as well https://twitter.com/HardwareUnboxed/status/1309659834468298753 This was linked in this article https://www.extremetech.com/computing/315511-a-stability-problem-is-brewing-with-nvidia-rtx-3080-3090-gpus

 

Edited September 28, 2020 by LN400

[Elthy]

Something about Nvidia seems of. Its almost as if they are extremly afraid of what AMD is going to show in a few weeks:

-While using a well established (almost outdated) process for the chips its almost a paperlaunch, with a few thousand cards at max all over the world
-The biggest chip on a XX80 card and clocked to the breakingpoint, using unseen levels of energy
-Extremly short time for AIBs to develop and test their custom cards, which were still released with the founders edition, not later as usual
-Now crashes that happen so often they make news on all hardware-websites

Navi2 sounds good, but that good?

[steve_v]

22 minutes ago, Elthy said:

clocked to the breakingpoint

Citation needed.
 

22 minutes ago, Elthy said:

using unseen levels of energy

Meh, not that far off a GTX295, GTX480 or Vega64 though. Performance per watt actually looks quite respectable.
 

25 minutes ago, Elthy said:

Now crashes that happen so often they make news on all hardware-websites

Given the shambles of a launch, not all that surprising. This sounds like a power delivery/board design issue, and there are very few cards in the wild so far. By the time normal people can buy one this will likely be sorted.
 

21 minutes ago, Elthy said:

Navi2 sounds good, but that good?

Maybe. Personally I'm not ready to attribute complex motivations to something which can be adequately explained by garden-variety incompetence and bureaucratic bungling just yet...
If and when the fabled Navi2 actually makes an appearance and gets some benches in, that'll be the time to make educated guesses.

Edited September 28, 2020 by steve_v

[Elthy]

6 minutes ago, steve_v said:

If and when the fabled Navi2 actually makes an appearance and gets some benches in, that'll be the time to make educated guesses.

Thats another strange thing. Usualy AMDs stuff leaks a lot before presentations, while Nvidia comes from the dark. This time its the other way around, hardly anything surprising in the Ampere reveal, while there is almost nothing known about AMDs chips other than 80 CU on the biggest one, the other leaks are often inconsistent. There are rumors about quite small memory interfaces on AMDs new chips, which would be the complete opposite of their usual strategy. Also Nvidia went the way of extreme raw processing power which is usualy AMDs job, while it seems for now that suddenly AMD could be ahead in energy efficency. Somehow they completly traded roles.

[LN400]

A summary so far (Sept. 28th) https://www.pcworld.com/article/3583778/nvidia-geforce-rtx-3080-crashing-issues-what-you-need-to-know.html

This article refers to Igor and the capacitors but is careful in stating that it is still theory, not proven.

Edited September 28, 2020 by LN400

[LN400]

In this article (Sept. 28th) https://www.pcworld.com/article/3583894/tested-nvidias-new-drivers-fix-rtx-3080-crashes-by-sacrificing-clock-speed.html it appears that the crashes are related to overclocking the GPU, and with new drivers from Nvidia, capping the clock, the issue is (reportedly) fixed.

Update (Sept. 29th) https://www.pcgamer.com/nvidia-rtx-3080-gpu-crashes-instability/

Edited September 29, 2020 by LN400