Hi guys,
What is wrong with my temperatures? Do you think I need a new waterblock (thinking of PWR5 and MEM3)? The temps are the same in IDLE and running 3DMark (1-2 degrees).
Thanks for any input! Take care!
 

Maybe not making correct contact.

The only reason a VRM temp would be that high is if there's bad/no contact with it, there's no reason for them to be running that hot otherwise, ever, even on an air cooled card.
 
The VRAM temp, do you have any back plate or any cooling on the back of the card?  It's possible the VRAM temp that's high is a backside chip measurement.
 
The 3090 FTW3 Hybrid I had didn't run anywhere near that hot on any of the temps, so an actual aftermarket liquid block should run much cooler than that.  Something definitely isn't right.

The backplate might help with the VRAM temp, but the VRM temp is something I don't see that effecting.  All of the VRM temperature sensors, as far as I know, are on the face-side of the card.  Pretty sure one of the VRAM sensors is on the backside, but it still shouldn't be running THAT hot at idle. 
 
You checked that the card is making good contact with the single RAM chip that's near the PCIe slot right?
 
You can always go full ham on cooling the back part of your card, I have a thread going about doing just that here: https://forums.evga.com/3090-Kingpin-Get-that-backside-VRAM-cool-Also-Thermal-Pad-size-backplate-side-m3219773.aspx It says Kingpin in the title but I'm trying to get visuals for all the different 3090s that I can... finding XC3 card PCB shots has been tough so far.

Tonyboj
Backplate is on it's way. But then MEM3 is still left to be solved?
 
Contact seems to be excellent (se photo). And temperatures got a little bit better after reapplying thermalstrips, but still the same with MEM3 & PWR5.

I quickly checked the thread you linked to. Great work! I'll be sure to do something along those lines. Am I correct that to mount a backplate I basically need to use less screws for securing the waterblock on the faceside?
 

Did you also put a pad or putty on the vrms? They are missing on your photo (check the bottom right) 

Okay. We'll that's disheartening. I was hoping for the backplate to do something. Yes, thermal pads with good contact on all the VRMs (LR22s), photo was taken just before I reapplied the last of the thermals. Installation was done by the book both before and after.

I'll try placing an active fan pointing on the back until the backplate arrives. Hopefully that will do some good.

Idle temps are 84c on MEM3 and PWR5, in game about 100-110c. I've now installed the backplate perfectly with precisely cut thermal pads and the temperatures, at least in idle, are identical.
 
IDLE temperatures
1 With stock cooler 35 C (~35 C for all PWRx/MEMx except PWR5/MEM3 which are at 90 C)
2 With EKWB FTW3 Vector waterblock 26 C (~27 C for all PWRx/MEMx except PWR5/MEM3 which are at 84 C)
3 With EKWB FTW3 Vector waterblock and backplate 29C (~27 C for all PWRx/MEMx except PWR5/MEM3 which are at 85 C)
4 With EVGA Hybrid Kit, will update in about 2 weeks.

LOAD temperatures
1 Stock cooler 80 C (MEM3/PWR5 around 112 C)
2/3 EKWB with or without backplate 60 C (MEM3/PWR5 around 110 C)
 

• Do you guys think these temperatures will affect the stability of the card? Can't see any signs of that during stresstest in 3DMark. Unfortunately overclocking is out of the question.
• Does anyone know the exact location of MEM3/PWR5?
• Maybe it's just the sensor? I don't know if this kind of temperature glitch falls under any warranty. The availability of these cards make me hesitant to even consider asking for a warranty replacement.

I have ordered the EVGA Water-cooling kit. 
 

Tonyboj
Idle temps are 84c on MEM3 and PWR5, in game about 100-110c. I've now installed the backplate perfectly with precisely cut thermal pads and the temperatures, at least in idle, are identical.
 
IDLE temperatures
1 With stock cooler 35 C (~35 C for all PWRx/MEMx except PWR5/MEM3 which are at 90 C)
2 With EKWB FTW3 Vector waterblock 26 C (~27 C for all PWRx/MEMx except PWR5/MEM3 which are at 84 C)
3 With EKWB FTW3 Vector waterblock and backplate 29C (~27 C for all PWRx/MEMx except PWR5/MEM3 which are at 85 C)
4 With EVGA Hybrid Kit, will update in about 2 weeks.

LOAD temperatures
1 Stock cooler 80 C (MEM3/PWR5 around 112 C)
2/3 EKWB with or without backplate 60 C (MEM3/PWR5 around 110 C)
 

• Do you guys think these temperatures will affect the stability of the card? Can't see any signs of that during stresstest in 3DMark. Unfortunately overclocking is out of the question.
• Does anyone know the exact location of MEM3/PWR5?
• Maybe it's just the sensor? I don't know if this kind of temperature glitch falls under any warranty. The availability of these cards make me hesitant to even consider asking for a warranty replacement.

I have ordered the EVGA Water-cooling kit. 
 

The EVGA kit will not affect the temps you are seeing. The temps will be worse than with the EK block.  If it were me I would guess this may be sensor issue since the temps are way too high compared to the other temps. Anything over 100C is not good or acceptable for regular use. I would definitely consider this a warranty issue. I would rma it to be on the safe side. The other option is to keep using it until it or if it fails then rma it providing you're still under warranty. This is a perfect time for an EAR program rma. I wouldn't wait and rma the unit now but that decision is up to you.

You would think there may be some type of throttling, with those temps. Does GPU-Z show anything happening? You may notice similar or better temps with the hybrid cooler.
I'm basing that off of my earlier GPU's results, one was on water, the other hybrid.. But, it won't do anything about the two sensors erratic readings. 

 
So far, no stability issues whatsoever. And it seems I get the performance I should. But temperatures are worrying. After playing Hunt Showdown for a short while MEM3 and PWR5 reaches 120 C (!).
 
I have tried both with the original stock cooler (all thermal pads perfectly mounted as from the factory) and EKWB Vector with backplate, the only difference is that max temperatures of MEM1-2 and PWR 1-4 are about 20 C lower with the water block. MEM3 and PWR5 are basically the same. Hovering around 100 C - 120 C at load.
 
In HWInfo I can tell the "Performance Limit - Power, Reliability Voltage, Utilization, Thermal" all turn to Yes some of the time.

I don't think that's correct, because there wouldn't be much point in having 4 sensors on one VRM.  But it's possible that's fairly accurate.
I can tell you that I'm pretty sure GPU2 is backside of the GPU Die, GPU1 is "edge temp" of the Die, which is opposite to what it looks like in the image... I didn't do enough experimenting with my 3090 FTW3 Hybrid to be able to tell you for sure what sensors are where unfortunately.  I would ASSUME it's similar to the Kingpin at least in terms of memory temperatures, I.E. Mem 1 or Mem 2 is backplate memory, and Mem3 is definitely face/Die side memory - but that could be incorrect.
 
I really think this is a case of sensors reading out wrong.
Your card has a bare-back currently, correct?  If so (or even if you have a backplate, this should still work), do this: put a non-conductive rubber glove on of some sort, run a stress test, ground yourself on something OTHER than your PC first to make sure you don't shock anything, then feel the back of the GPU PCB in a few places.
 
If something is ACTUALLY running at 84C-102C, you will DEFINITELY feel it - keep in mind, 84C is 183F, that's nearly the boiling temperature of water, so if something is that temperature, you should get that instant "yikes" if you actually touch near it.
 
You should be able to tell if at least the memory is actually running that hot anywhere, because if it is, you'll definitely feel it.  
 
Have you checked HWInfo64 to see what it's reporting for Memory Junction Temperature?  If that's not reporting the high temp seen on Mem3, I would definitely say it's a sensor error - Junction Temp is supposed to be the hottest internal temperature of any of the memory modules, from what I understand.
 
It would be really nice if eVGA actually gave us information on where those sensors are located, especially in cases like yours where there's supposed hot-spots - you could easily determine if it's indeed an error or not simply by feeling/thermal camera at the area it is in.

Check your mounting pressure. Maybe the board is flexing and not making good contact with the block. Its really easy to do.
A backplate is a must with a 3090. I tired without one and my temps went to 90c pretty quick.
Ek has a ton of pads included with their quality backplate. The stock one is pretty flimsy               

126C is ridiculous. That's nearly 260F. Components would fail in short order at those sustained temps. It has to be sensor issues. I would rma the card just for piece of mind.

Were you able to solve this issue? I have EK FTW block and backplate and also have high temps on MEM3 and PWR5. contact seems pretty good as well. these sensors are 20c higher (compared to other mem sensors) and 30c higher (compared to other pwr sensors).

Follow up... My average temps last night after 7 hrs of gaming were in the mid 40's C.  Based off of other threads, I think I must have had some faulty reading spikes to the 95 C; I saw those readings yesterday and the day before.