Hello - I purchased a RTX 3090 FTW3 Ultra Hybrid, back in 2020, directly from EVGA. The card itself runs great, and I have never really had any issues with temps playing anything on max graphics. The GPU temp rarely reaches over 62c but over the last year, when temperatures approach the 60c mark, I noticed the GPU hot spot will ramp up to about 89-92c alongside with it. So, roughly about 30c delta and an average of about 24-26c between GPU temp and hot spot after a session of gaming, depending on how demanding the title is. The highest I have ever seen the hot spot reach was 97c but that was a once or twice thing. The ambient room temp is 21c, and I use Afterburner to run a pretty aggressive fan curve and undervolted the card. Memory junction temperatures rarely go beyond 80c at max load and there is usually a 15-20c delta between GPU temp and Memory Junction temp. I have ran multiple benches with 3DMark and get the same results as if I were to game a demanding title, although the clock frequency will sometimes slow down during heavier benches such as Fire Strike Extreme (natively running 1440p).
 
I am not really the most knowledgeable person when it comes to these types of issues. Are these temperatures within acceptable range? Is a re-paste in order? Information on this model of card has been hard to acquire and would appreciate some advice before I even consider dismantling such a huge investment. I am sorry if this is a repost. I tried to post a couple days ago with pictures but for some reason it didn't post. If there is any critical information I have left out please let me know. Thanks in advance for any help!

Just know that "PTM 7950 pad", might not be the best choice for any water cooled application .... the "anti-pump-out feature is great for air cooled cards" 
 
I'm not saying it can't or wont work, just that it has been discussed as an issue for the cooler running AIO & water blocked cards
 
Not over-hyped: Honeywell PTM7950 in a lab test and a real game changer for graphics cards | igor´sLAB (igorslab.de)
 
(excerpt)
 
"Where I see problems are with water coolers and water blocks, where there might not be a complete phase change, at least on the cooler side. But you need at least 55 to 60 degrees throughout the pad for a complete change of state. “Half-defrosted” pads are counterproductive and are even more prone to cracking. At least that’s what one of the OEMs has now confirmed to me. Everyone has to work out for themselves how to solve this properly. I wouldn’t use a pad on a water block, at least not a PCM,"

Use the paste you have
 
The difference of GPU vs CPU ..... use the spread method for the GPU die, make sure to cover the entire die ... there is really about 0 room for movement when assembling the GPU cooler & Keep the Assembly Level. 

Install all screws loosely before you start tightening them to get everything aligned.  You want a screwdriver that is in great shape & fits those tiny screws, (Philips #0 & #1 if I remember correctly) so you do not damage them.  Be careful not to cross-thread those tiny screws, start by turning them "backwards" /Counterclockwise, until you feel the threads align (click/clunk feel)
 
Look over EVGA guide for the AIO kit for your card -- this will give you a better understanding of where everything goes & where the screws are.
 https://www.evga.com/support/manuals/files/400-HY-1988-B1.pdf
 
Easier to remove parts from the card & not damage pads if everything is still warm, or at least card is not cold   So a warm place in your house for a couple hours or wrapped in a heating pad (if you have one) can help somewhat.
 
Card Pads:
EVGA pads seem to be on the softer side - you need the pads to compress, so the GPU die to cooler contact can be optimal
 
Ask EVGA if pads are available - after your card is registered - EVGA will use the email address,  from your registered account for correspondence
 
Contact EVGA if you should have any additional questions or concerns.  Support@EVGA.com
 
 

Copy pasting this from another post (my results) :

I installed PTM7950 on my hybrid 3080ti. Please note that I also have a copper plate mod, so my vrms are getting cooled by the radiator vice the vrm fan.

So far temps are GREAT. On 3dmark time spy, with a fresh MX4 paste and push/pull with noctua redux p12-1700 I used to get 54C (I know there are better tests, but for comparison wise I always use that). Over time it gotten worse and with recent tests I would get 66-68C on 100% rad fans, vrm fan and case fan (yes it's louddddd). In gaming (cs2), I would start in the 62-ish C and finish my night in a warm room at 72C, with 50% case fans, 60% rad fans and 35% vrm fan.

With PTM7950 so far:
1. 3dmark time spy: 54C with push ONLY with noctua A12x25 at 60% rad fans, 35% VRM (I could turn it off completely but I would still like some airflow in there)
2. Gaming: same fan speed as above (nice and quiet!), under 60C no matter what. Even if my room got warmer, I would fluctuate between 52-58/59.
3. Furmark: for some burn in, I left regular fan speed as above, and I would be maxed out at 58/59C same as gaming. I did turn down the rad fans to 35/40% and I was getting 66-68C max.

Overall super happy. I think it's the same temps or better with a fresh thermal paste, BUT I'm using only 2 rad fans instead of 4. I was getting annoyed with the very loud PC that I started looking at quiet options (obviously gotta throw noctua a12x25 in there!). Now I'll never need to re-paste again and temps should stay pretty consistent.

Temps are above 45C, so a phase change should occur. I also have a hotspot that is in the 60s, and VRM that although not touching the ptm7950, it is getting heat to the pump plate. Pump plate should be above 45C, so I see the PTM7950 working as intended. And if it doesn't, I mean whatever, it's efficient enough to give me low, consistent temperatures.

scarecrow666
I appreciate all the advice and providing a link to the guide. Now, I actually can see what I am up against. How thick should you spread the thermal paste on? Pea size and spread? I contacted EVGA in regards to the thermal pads. I will update when the problem is resolved. Thanks everyone. 

You just need enough to coat the entire GPU die, think of it as painting - as long as you cant see the die & it looks even that is enough

scarecrow666
EVGA was nice enough to send me replacement thermal pads. Although, not a whole set. Didn't include backplate pads. Shipping took forever. Will update when finished.

 Good to see you were able to get pads from EVGA
 
Just reuse the backplate pads, unless they shred

scarecrow666
I used the replacement pads from EVGA and repasted with NT-H1. I was able to drop the hotspot temp by about 10c. Unfortunately, I have memory junction temps that are in the 80-90c range when they were previously 72-80c. When disassembling, the pads ripped on the backplate and I salvaged what I could to make it work. Any thoughts?
 
Timespy Stress 20 loop: imgur.com/a/WR6AHhw
 
Sorry, I can't figure out how to post the picture and the spam bot is not letting me post a link.

 
 
PSA adding a photo to your Post - EVGA made some changes - Post Directions below
 
Learn your way around the EVGA Forums, Rules & limits on new accounts Ultimate Self-Starter Thread For New Members
 
Just buy the new thermal pad(s) of the correct thickness to replace the backplate pads that got damaged - (as you mentioned EVGA shipping was slower than you liked)

 
FAQ # 59822  What are the thermal pad sizes for my 20 or 30 series graphics card?
 
(your photo)

scarecrow666
I will order replacement pads for the backplate and update when finished. Thank you for all your help. 

Happy to Help

scarecrow666
I replaced the backplate pads with some 2mm thermals, as you can't order 2.25mm. Or, at least I couldn't find any available. Was only able to lower mem junction temps by -2c. So, I improved hot spot temps by roughly -10c after the repad/repaste. But, now I am unsure how to proceed on fixing this memory junction issue, as it has increased around 6-8c. I looked into the X1 Precision tool but the sensor areas on the card are pretty ambiguous to me.
Port Royal before repad/repaste:
Port Royal after repad/repaste: 

95c or lower is fine for memory, so looks fine for now.

No problem. 👍

 

Cool GTX
 
Did you check the thermal pad imprint after making a test assembly?  (to make sure that all parts were contacted)
 
 

 
If spec is 2.25 - then 2.5 mm would have been the way I would have gone -- as long as you use a soft pad
 
Your results fail to mention Delta T device temp in relation to room temp:

Your testing needs to consider changes in room temp (ambient) & case fan CFM /RPM
 
 
The memory chips on the back of the GPU are the warm ones? -- you can add an additional fan to move air across you GPU backplate
 
What are your OC settings?

Big oof! I do not believe I checked thermal pad imprint in the test assembly. I remember seating it and lifting it back up to inspect the paste but not the pads. 
 
2.5 mm was an option for the top pads, but I feared going too big would make it worse, as I am not well versed with these components. I figured with a set of cheapies first, I could at least gauge if this was the issue. All other pads came directly from EVGA, and as far as I know, they are to specification.
 
Room temperature remains a constant 21-22c (71f). The case fans maintain a constant 800-900RPM. I am not exactly sure which chips are the hot ones now, since I replaced all the pads. I couldn't exactly find the most reliable information on where the components are actually being measured at. I have the card undervolted at 0.925 V with a 1950Mhz clock (mem is untouched). Normal BIOS switch is activated. A fan across the backplate would be quite the squeeze, as I have the massive NH-D14 cooler from Noctua. If there is any information I failed to include please let me know. 
 
Side note: I forgot to mention that when I initially disassembled the card, EVGA had originally used thermal paste on the rows of LR22 chips. I guess they decided that was a horrible mess and switched to pads in the hybrid kit (thank god). Now, they have been replaced with pads instead of the globs of dried thermal compound they used. 

I took the card apart again and checked the thermal pad contact like I had forgot the first time. There was some minor contact issues in a couple of areas. After another clean and re-paste, I am seeing a noticeable difference. Was able to drop the hot spot another -7-8c and lower memory junction temps -2-4c. Memory junction temps aren't what they used to be that's for sure, but I think they are looking more like they used too. 
 
Port royal original temps: 
Port royal after repaste: