Hi All,
 
I think I found the issue why some users here report that the Card stop working or Crashes or similar.
 
I did a furmark test with the fresh recieved card on stock settings. Therefore I saw in the Software absolut acceptable temperatures on all sensors (picture 1). Then I did a small measurement on the card with my flir (picture 2) and then I was shocked. On the Back side of the card the area nearby the memory Chips get 94° hot! On the same time the temperatures shown on the sensors of the software did not change. So there is a massive issue with the cooling of the backside of the card. The Backplate is not nearly enough.
You need to mention the Furmark test only runs about 3-4 minutes with an opened case.
 
So is this known to evga? The bad feeling for me is that you cant actually trust the software. Maybe some users overclock their cards more and more cause the software says everything is in good temperature ranges. But in real the memory cooks above and nobody knows...
 
Until yet I did not reciev a crash, but I think its clearly if people pushes OC, then this could clearly happen.
 
pic1

pic2

LK

Yeah sure, my Flir works great. For those high temperatures the flir only deviates for around 1-2 °.
The only greater deviation is on reflective metal surfaces like the deep blue/black shown metal bracket. Flirs do have a problem with reflective surfaces. But the PCB on the Back and the memory chips are no problem and only have a small deviation like mentioned above^^.
 
Until yet I measured all of the hardware which I built for my friends and customers...

I wonder if some adhesive heat sinks would help. Or the water cooling backplate product as a step up from that.

I currently have a 3500RPM fan pointed at the back of my card until my waterblock arrives.

Bradypan
Also, edit your image in your original post, it has your latitude/longitude coordinates.

It does!!?!  I didn't notice.  Now I will have to type them in and figure out where he lives.
[sarcasm]
 
I noticed the coordinates hours ago.
If you are genuinely worried about it, stating it publicly doesn't help him in any way.  It just increases the realization of the coordinates existence.

Trust me... if you figured this out within a few hours of opening the package... both NVIDIA & EVGA are well aware of it.
 

Zetta
 
Nice. This is very interesting. I'd like to know if this is at all normal. My cards have typically failed when running in non-demanding games with small OC and a power target of 107.
 
I doubt EVGA will comment on this though.
 

It's not just an EVGA issue.  Unfortunately it's an issue with all AIB 3090 cards.  I remember having a similar overheating issue with VRAM on the Pascal Titan X where it would get too hot to touch.  Back then I fixed the problem by using an active cooled backplate from AquaComputer and that brought the temperatures down from 80-90C into the 50s.  A lot of people over on OCN are basically sticking a thermal pad on the backplate, and putting a heatsink on top of that.  It really helps.  Especially if you have airflow in your case.

AWK16
Hardware engineers that have build the card from ground up and you think that's the cause of literally dying cards, you have to look elsewhere, although at first hand sounds pretty convincing

Example:
Hardware engineers had built the first revisions of X299 Mainboards and selled them all around the world - all brands together had the Temperature issues with their vrms regarding crashes...
Der8auer - a simple youtuber - not an hardware engineer found this issue and addressed that to the developers... After that all Mainboard brands made a new revision with better cooling.
 
I think I had no more to say here, exept iam not der8auer.

AWK16
Hardware engineers that have build the card from ground up and you think that's the cause of literally dying cards, you have to look elsewhere, although at first hand sounds pretty convincing

Hardware engineers make mistakes.  Hardware engineers' suggestions are also sometimes ignored by management to lower costs.