Falkentyne,
I got excited too early
After 4 hours of playing CP2077, the core temperature is 68C, but the temperature on the hot spot is 90C. The delta is 20C.
What could it be, where to look, what to do?
Or is this normal?

All this effort and it all goes to waste...?

 

Take the card apart and take a clear picture of the thermal pads on the front side of the card, VRAM and mosfet pads and picture of the thermal paste on the core and heatsink
Do not remove the thermal paste. You can remove the thermal pads from the heatsink to take a picture of the contact surface (thermal pad contact to memory and VRM surface). Then you can post it here.

Is there compression marks on the pads? Imprint of VRM and chip borders and/or writing on the pads?
The thermal paste should be pushed outside to the outer edges around the chip (with only a very little left, touching the core->heatsink evenly) and the pads should have complete pressure "map" (imprint) of the GDDR6X and VRM chips on the pads.

 

These temperatures were obtained after 7 minutes of MSI Kombustor
That doesn't sound like a 10 to 14C delta to me. What do you think of this, Falkentyne?

 

Another two things I noticed.
1) the memory pressure looks almost too good. Almost like the pads were too thick. The writing seems to be too easy to read.

You can compare my 1.5mm gelid extremes


With yours.

http://images.vfl.ru/ii/1637164418/aab328a5/36728709.jpg

Yes I know my picture is more blurry but you can still see what I am talking about.
Your 1.5mm gelid extremes look like the pressure was too high, which means those pads must not really be 1.5mm...

2) It looks like the bottom left section of your VRM pad is cut way too short. Like half the bottom left mosfet is not covered.
Again compare mine and yours. You can use the little white sticker by the VRM to compare.

3) and of course, the thermal paste spread issue and consistency issue (very big problem too).

 

Falkentyne, Thank you for answering me. This is very important to me.
I will now try to respond to your messages one by one.
I applied TFX crosswise, then carefully spread it over the surface of the chip. Obviously, I didn't apply it enough, judging from your follow up photo.
I used Gelid Extrime 1.5mm thermal pads on the VRAM and Gelid Extrime 2.0mm on the VRM. Under the backplate I used Gelid Ultimate 1.5mm gaskets. All in all, I used everything you advised me earlier.

I used to think that thermal paste smoothed out the irregularities of the substrate and crystal, so I applied it in a thin layer. But after seeing you apply thermal paste, I was a bit surprised.
Do you really need that much thermal paste?

Before I answered you, I went to a friend's house. He also has a 3090FE. He doesn't use it for gaming. Four months ago I replaced his stock thermal pads with Thermalright odyssey 1.5mm. The thermal paste used was the simplest, no name. I went in to see what his GPU and GPU Hot Spot temperatures were under load. We ran both CP2077 and FurMark. And this is what I saw after 15 minutes of running and playing, the delta did not exceed 14C. GPU temperature was 75C, hot spot was 89C, memory temperature was 95C. And this despite the fact that the graphics card has been running 24/7 at 100% load for the last four months To say I was surprised is nothing to say. What could it depend on?
The graphics cards are the same. The only thing that confuses me a bit is the issue date of my card (there's nothing wrong with that), but what the Z-V3 engraving means is the question. I highly doubt that Nvidia will make changes to the steam chamber design. Anything is possible, though.
Well, now I'm going to try applying thermal paste like you showed in the picture. I will replace the thermal gaskets with new ones similar to the ones I had.

I read your post somewhere that said not to overtighten the GPU mount plate. Is this true or not?
I tighten it all the way down.

After I get everything assembled, I'll report back.
If you have any other thoughts, post them.

 

Falkentyne, Thank you for answering me. This is very important to me.
I will now try to respond to your messages one by one.
I applied TFX crosswise, then carefully spread it over the surface of the chip. Obviously, I didn't apply it enough, judging from your follow up photo.
I used Gelid Extrime 1.5mm thermal pads on the VRAM and Gelid Extrime 2.0mm on the VRM. Under the backplate I used Gelid Ultimate 1.5mm gaskets. All in all, I used everything you advised me earlier.

I used to think that thermal paste smoothed out the irregularities of the substrate and crystal, so I applied it in a thin layer. But after seeing you apply thermal paste, I was a bit surprised.
Do you really need that much thermal paste?

Before I answered you, I went to a friend's house. He also has a 3090FE. He doesn't use it for gaming. Four months ago I replaced his stock thermal pads with Thermalright odyssey 1.5mm. The thermal paste used was the simplest, no name. I went in to see what his GPU and GPU Hot Spot temperatures were under load. We ran both CP2077 and FurMark. And this is what I saw after 15 minutes of running and playing, the delta did not exceed 14C. GPU temperature was 75C, hot spot was 89C, memory temperature was 95C. And this despite the fact that the graphics card has been running 24/7 at 100% load for the last four months To say I was surprised is nothing to say. What could it depend on?
The graphics cards are the same. The only thing that confuses me a bit is the issue date of my card (there's nothing wrong with that), but what the Z-V3 engraving means is the question. I highly doubt that Nvidia will make changes to the steam chamber design. Anything is possible, though.
Well, now I'm going to try applying thermal paste like you showed in the picture. I will replace the thermal gaskets with new ones similar to the ones I had.

I read your post somewhere that said not to overtighten the GPU mount plate. Is this true or not?
I tighten it all the way down.

After I get everything assembled, I'll report back.
If you have any other thoughts, post them.

Yes you do need that much thermal paste. The core is convex or the mounting is convex.

GeForce RTX 3080 and RTX 3090 with bended package - why water and air coolers have such a hard time | Investigative | igor´sLAB

Of course, you’ll be amazed if, for example, three supposedly identical cards from one manufacturer have completely different temperatures or fan speeds and this is then reflected very similarly in…

www.igorslab.de

But I do not know if that is your problem. I don't know how you spread the paste. But TOO THIN IS BAD. Period.
You must solve this problem yourself. It's too difficult for me to solve it. Because I don't know what you did.

The 'pink paste' Kryonaut Extreme--that person used "water block" (custom loop) with perfect block flatness, so that example is perfect. You can't do better than that.

But this is "close" to how your TFX should look ideally by being pushed "off the core", then what is left over will slowly harden up (cure) and make a solid connection.

I have no idea what Z-V3 is or what this stands for. Mine doesn't have this marking.

I have no idea if the heatsink tolerances were changed. I have no way to find out.

The easiest way to test this is to buy Gelid Extreme 1.0mm and 1.5mm (2 packs).
Then try Gelid Extreme 1.0mm on VRAM and Gelid Extreme 1.5mm on VRM's.
Then full spread TFX on the core.

If the VRAM thermal pad is too thin, your GPU will have 10-12C Core to hotspot delta but VRAM temp will be 100-110C.

I think that is the only possible outcome unless another crazy example is needed, like 1.0mm on front VRAM and 2.0mm on front VRM (instead of guessing 1.0mm front vram, 1.5mm front VRM).
Those are the only three possibilities.
But I find it hard to believe you can spread TFX so thin to make it that bad.
TFX is a very thick paste. How did you spread it so thin? I just do a thick pattern "X + dots" then I spread it manually in a nice thick layer, making sure the entire die is covered. Then I attach the PCB->heatsink->Screws etc. (X-bracket=criss cross pattern, one turn each screw, then diagonally alternate-->1 turn---diagonally alternate-1 turn, diagonally alternate again.

like this. (1, 2, 3, 4). But again I do NOT Know if this is your problem!! It's too hard for me to solve this.

1----3

4----2

 

And apply the paste to the 9 points? It won't need to be smeared?

 

Before you do this, check to see if you have a pressure problem from having the pads too large/badly trimmed or cut wrong.
It's possible that is your problem and that 1.5mm + 2.0mm is still correct, for core side VRAM and VRM's.
Because if they were too thick, you would have good compression all over, but as you can see you have a balance problem. I think this is because of bad cutting, bad application (not straight, not directly on chips)--or both. Again I don't know for sure!

But I think you need to test that first and eliminate that.
You maybe can still re-use the same thermal pads if it has only been 1 day, just make sure you trim them correctly (especially the left side VRM, remember--that part is too short at the bottom left).
so maybe you need to cut a new VRM strip.

You can use Gelid Extremes (2mm) or Gelid Ultimates (2mm) on the front mosfet (VRM's) in the three sections. Either one is fine for 400W users or people without a shunt mod.
Gelid Extremes are probably easiest since they compress the best. Ultimates are better for durability under high heat.

Try to test the proper trim and application first, before you change the thickness. Could just be user error....

For VRAM (GDDR6X) only use Gelid Extreme (1.5mm) on CORE side of the card. For backplate yes you can use Gelid Ultimates on VRAM on that side and i recommend that.

You apply paste to the 9 points then smear it after.
The points will connect each other and then spread straight. If there is gaps or some paste sticks to the spatula, just go in another direction and move back the opposite way.

Also, TFX is easier to spread if you boil water in a pot, remove pot from fire/stove, wait about 5 minutes, then after 5 minutes, put the TFX Syringe inside the water (cap covered of course!!) and let it sit about 15 minutes. Then drain the pot, remove the syringe, and carefully use potholder (just in case it's hot) and dry the syringe tip and side, and then apply.

If you have a hot air dryer (usually only if you have long hair or use this thing) it's much easier...just apply a hot air dryer directly to the paste after you do the 9 dots, then spread it.

 

As I wrote earlier, I will be posting my experiments with gaskets here. Perhaps for someone it will be interesting and informative. Perhaps it will help someone.
Thank you for the tips and communication, Falkentyne!

 

Falkentyne, one more question for you: did you use this example (https://bit.ly/2Q8fcPk pad cutting layout by Matthew Lloyd ) to cut the gaskets, or did you measure them in place?

 

I cut them in place on top of the chips by lining it up and guessing.

 

Hey @alloxa09,

FWIW after taking the advice from @Falkentyne regarding the TFX paste, I have used it many many times now for myself and clients ( I run a PC store), and after struggling to find a less time consuming way to apply it, I have come up with this solution: Leave it under the heat gun on low heat (taking care not to melt the plastic tube, which is surprisingly heat resistant) Do the "X" and spread as much as you can with a hard flat plastic tool (like ones builders use for patching holes in walls) then wrap a piece of plastic wrap around your index finger and "pat" it down with your finger. After doing this to multiple 3090s I've had through the shop with high hot-spots, for I would say over half a year (?) I've yet to have one come back with hotspot issue. All thanks to @Falkentyne. I should start paying him commission

 

Hey @alloxa09,

FWIW after taking the advice from @Falkentyne regarding the TFX paste, I have used it many many times now for myself and clients ( I run a PC store), and after struggling to find a less time consuming way to apply it, I have come up with this solution: Leave it under the heat gun on low heat (taking care not to melt the plastic tube, which is surprisingly heat resistant) Do the "X" and spread as much as you can with a hard flat plastic tool (like ones builders use for patching holes in walls) then wrap a piece of plastic wrap around your index finger and "pat" it down with your finger. After doing this to multiple 3090s I've had through the shop with high hot-spots, for I would say over half a year (?) I've yet to have one come back with hotspot issue. All thanks to @Falkentyne. I should start paying him commission

Hey @Gandyman,

Thanks for the tip
Now it remains to experiment with the thickness of the gaskets and practice cutting them accurately🔪
I think I can do the latter, as well as applying thermal paste.
But now I really want to wait for the gaskets and start combining with them.

 

I did all you said, every single thing, and my delta increase of 2°C in 9 days, i have it now at 13.5-13.8

 

first days it was 11°C

 

Anything from 10 - 15 is perfectly fine and compeltely normal.

Only worry if it starts going over 20. [For eg mine was over 30c! It would get to 106c a few seconds after starting Heaven 4.0 even tho the core temp was 72c - 74c]

 

I have/had the same issues. I also own a 3090 ROG Strix. Had a delta from GPU temp to Hotspot of 30°. It reached almost a 110° on the Hotspot.
I repasted the card back in June 2021. After that, I only had a hotspot delta of 15-18°. Reaching 95° max.

Now after 6 months, I had again a delta of 25 degrees, reaching 105° on the hotspot, almost 80° on the gpu.

So I repasted the card again, using Kryonaut this time (before MX-5). No I have a delta of 13° max. Great temps like never before.
However, I think it will only last a couple months until it goes bad again.

I think there is a problem with the heatsink or mounting pressure. I could see the letters on the heatsink from the die.
Attached a pictures temps before after and one of the heatsink

 

Email or call these guys and ask them if you can buy a sample of this.

IC Contact Test Analysis Kit – Innovation Cooling

It says out of stock, but it is repackaged Fujifilm "Ultra Low Prescale" (the same stuff on sensorprod.com 's website).
Ask them if they can source one for you. They should be able to or at least direct you for a way to buy it, unless you want to go to sensorprod yourself and request a free sample from them (you can use Ultra Low prescale, which is the Innovation Cooling one, or Extreme Low prescale, which is I think 7-28 PSI rather than 28-85 PSI.

Then when you get it, do a test mount with it (you will need thermal pads installed, otherwise the test is completely useless), then post a picture of the contact sheet so people can see the pressure area.

Keep in mind as well you need proper thermal pad thickness and pads (the ones on the VRAM) that are soft enough. I believe that website that sells "Kritical" pads have pads that are as soft as Gelid Extremes (apparently they said the newer batches of Gelid Extremes are a bit harder than older ones), or you can use TG PP10 thermal putty to avoid any pad hardness issues.

 

Hello all,

I took the card apart again, replaced all the thermal pads and did a re-paste with Thermalright TFX.
It was a nightmare to clean off the used pads on the back, otherwise everything went smoothly. I used Gelid Extreme pads.
I had the impression that 2mm pads for the memory on the backplate might be a little bit to think, since one pad seemed to have much less pressure.
Or the OEM pads could just be a little bit too hard.

The TFX paste, I put it in the oven for 30min at 60°C. It was pretty easy to spread and I made sure to put a decent layer on the die.

Memory junction temps are lower now by about 8-12°C.

Hotspot delta is about 12-13°C.

I hope that did it for the long term. Not planning to re-paste that card again.

Thank you all for your help! Especially to Falkentyne and Gandyman for all the research and the support!

 

Best of luck! Sounds like your going to be all sorted!

 

I wish NVIDIA would be less closed lipped on everything..

I know that AMD RDNA cards can easy hit 115c die junction temp, and AMD claimed that to be completely fine. And seeing as 5800xt aren't blowing on en-masse I guess they are correct. Maybe 115c hotspot is within spec for NVIDIA's Samsung silicon? Because the amount of 3090.owners out there who don't even know to look at hotspot aren't all having their 3090s die right? Even miners. I know it's only been a bit over a year since launch but surely some major issues would be popping up and being reported on by now. Who knows I guess.

 

I wish NVIDIA would be less closed lipped on everything..

I know that AMD RDNA cards can easy hit 115c die junction temp, and AMD claimed that to be completely fine. And seeing as 5800xt aren't blowing on en-masse I guess they are correct. Maybe 115c hotspot is within spec for NVIDIA's Samsung silicon? Because the amount of 3090.owners out there who don't even know to look at hotspot aren't all having their 3090s die right? Even miners. I know it's only been a bit over a year since launch but surely some major issues would be popping up and being reported on by now. Who knows I guess.

I agree with you. I also posted this issue on the rog forums when I first noticed. But no reply from ASUS.

I think you are correct with your assumption. There was a lot of discussion about the junction temps of the AMD cards back when they came out.
I remember also when the R9 290X came out and a 95°C core temp was "ok" and "safe". Which seemed impossible at that time.
The hardware can definitely take more beatings than we think.
And, if the cards where not so crazy expensive and almost impossible to get, I would have been less worried.
I took every card I owned apart and did a rep-paste, just to make sure everything is as good as it can be.

I'm sure the engineers knew about this issue before the cards were released. But their calculations showed that 99% of the cards would surpass the warranty period or even beyond that.

 

My hotspot delta is increasing again... significantly... Seems that TFX does not solve this issue. This paste is behaving worse than kryonaut

 

My hotspot delta is increasing again... significantly... Seems that TFX does not solve this issue. This paste is behaving worse than kryonaut
View attachment 2550157

Your die temps are high. No reason why it should be 77C. What fan speed are you running?

Make sure your thermal pads are not too thick and are not interfering with core contact on the heatsink. Weak contact pressure on core to heatsink will cause bad hotspots no matter what paste you use. This would be extremely noticeable if you used Liquid Metal (requires insulation of SMD's around the die, e.g. Kapton tape or conformal coating) on the die instead of regular paste---if your card reached 105C instantly or shut down the instant you tried running a game, then you know 100% it's contact pressure on the die, as LM is much much thinner than any thermal paste.

Also make sure that your VRM (mosfet) thermal pads are not too thin either. Too thin thermal pads on the mosfets also affect the hotspot (found this out from a user who had a FE card, great paste coverage on the core but a missing section of VRM pad completely missing from a small mosfet section, and had 105C hotspot, which dropped 30C when he replaced the thermal pads).

It's very important to get the pad thickness and spacing correct. Memory temps will tell you if the memory pads are too thin. But either memory or mosfet (VRM) pads too thick will cause bad core and hotspot temps due to insufficient die pressure and hotspots on the die. Mosfet (VRM) pads too thin will cause good core temps but bad hotspot temps due to VRM's not being cooled enough (under some circumstances this can cause black screen + 100% fans also).

 

It looks much better than it did but I can follow you investigation. There was always more pressure in the center no matter what I did.
Also, every Rog Strix 3090 teardown shows the same spread pattern...
Since I think the thermal pad dimensions are correct, I could try even softer pads than gelid extremes. Or some thermal puddy.
I cut all the pads for the memory in single pieces now, and slightly smaller that the memory chips so they can maybe compress a bit easier.

I can post a picture from my previous re-paste.

 

That swimming pattern of paste is still indicative of low mounting pressure. Whenever you see something like "mountain ranges from space" when looking at a GPU like that, there's a contact pressure problem between GPU and heatsink.

Here is how it should look.

See what I mean now and what I wrote in my above post?
All the paste is pushed outside the die, between the die and the SMD's, with only a micro film layer remaining on the die.

Now that's perfect mounting pressure and fit. You aren't going to get that without a custom block fit, but still, the paste even on a less perfect, but still "solid" fit, should be pushed off the die between SMD's and die edge.

There is a possible mod you can try to improve this.
If you have some Super 33+ tape or Kapton polyimide tape (i suggest kapton tape, personally), you can apply stacked layers (maybe about 12 total stacks), right where the "square" part of the X-bracket, where it goes in around the back of the GPU die outside of the SMD SP-Cap and MLCC cap area, touches the PCB. Apply them to the PCB, not to the X-bracket, and you would have to do them in a square (so 4 strips cut to size in a square will be "one" stack, but each 'strip' should not be sitting on top of a strip going the other direction). Try to not have the strips overlap each other at the "edges" of each strip however (that causes uneven spacing at the corners!). Then once you get one full square, that's one stack, then you can do that about 12 times.

This 'should' create extra pressure downwards and should help a bit. The leaf spring (x-bracket) is a reverse spring.

 

You should be able to measure the screw's size and pitch and get replacements at your local hardware or online. I assume they are M2 screws, maybe 6-7mm? Though I have no idea about the specifics of the Strix so that's just guesswork on my end.

Ideally if you were to manage to find a replacement screw of similar size and pitch, I would personally recommend getting Torx/hex head ones instead of standard philips so they aren't as easy to strip.

 

Sorry all for the bad picture

They're Philips, countersunk screw heads... M2 or M2.5. With a spring attached to it like the classic heatsink screws from previous generations.
But you cant remove the screw from the bracket without removing the spring from the screw first. That is why I was not able to take the measurements.

They look like this but with a smaller screw head, I think countersunk.

1

 

Hi all, ready for an update.
Ordered some screws and got them today.
My Hotspot delta was on 23°C again... It raised 10°C in the last 10 days.

What I did now, I just removed all screws from the backplate and leaf spring. The screws from the leaf spring have springs on it, which can just be pulled off. (Why did I not try that before)
Pull them off and never put them back on.

So, I did not re-paste the card, just replaced the screws and my Hotspot delta is decreased by almost 10°C.
I assume these stupid springs on the screws just get "stuck" sometimes and don't compress correctly which causes insufficient pressure...

I will re-paste the card again in the coming days. It is fine for now.

Here are the screws you need if you want to upgrade to TORX screws like I did:
M2x8 for the backplate screws
M2.5x6 for the "GeForce" bracket
M2x6 for the leaf spring
There are some original Torx on the metal inserts on the backplate, did not replace those.

Be sure to get 90° countersunk screw heads.

Here a picture. Screw on the right is the original from the leaf spring. Left the 3 sizes you want if you like to replace them

 

My Hotspot delta was on 23°C again... It raised 10°C in the last 10 days.

Are you using Thermalrite TFX ?

 

These are my temps roughly 3 weeks after using the EKWB block with TFX paste + Gelid Pads.
Hotspots are within 1 degree difference all the time for the last 3 weeks depending on the games I play and how much wattage is being used by the GPU.

 

Hello! I did a full custom water loop with my rig a few weeks back, used a EK quantum vector trio 3080/3090 for my gpu since it's a 3090 gaming x trio. Was doing it slowly over time due to unfortunate and impeccable timing of events that hinder my progress but i finally put everything together 2 weeks ago.

I had only use my computer to watch videos or play some quick light games that weren't too intensive and time consuming due to my busy schedule. Got the whole day free to myself today so I decided to try and do some tinkering while stress testing to which i noticed that the gpu temp and hot spot temp had a difference of 29 to 38ºC while under heady load and then having 22ºC difference when in idle.

I used the thermal pads that EK issued for both front and active backplate water block which were 1.0mm and 1.0+2.0mm and also used their thermal paste respectively. I'm not sure if its due to me not measuring the pads correctly or if there's not enough pressure for contact. Do i need to get gelid and TK to help with this issue? If i do need them, which gelid thermal pads should i get for my gpu for both the front and back?

Here are the temps screenshots

Spoiler: heavy load

Spoiler: idle

Really hoping i could fix this without having to ultimately buy and replace my gpu for my rig as this is kind of stressing me out a bit...

 

Hello! I did a full custom water loop with my rig a few weeks back, used a EK quantum vector trio 3080/3090 for my gpu since it's a 3090 gaming x trio. Was doing it slowly over time due to unfortunate and impeccable timing of events that hinder my progress but i finally put everything together 2 weeks ago.

I had only use my computer to watch videos or play some quick light games that weren't too intensive and time consuming due to my busy schedule. Got the whole day free to myself today so I decided to try and do some tinkering while stress testing to which i noticed that the gpu temp and hot spot temp had a difference of 29 to 38ºC while under heady load and then having 22ºC difference when in idle.

I used the thermal pads that EK issued for both front and active backplate water block which were 1.0mm and 1.0+2.0mm and also used their thermal paste respectively. I'm not sure if its due to me not measuring the pads correctly or if there's not enough pressure for contact. Do i need to get gelid and TK to help with this issue? If i do need them, which gelid thermal pads should i get for my gpu for both the front and back?

Here are the temps screenshots

Spoiler: heavy load

View attachment 2555102

Spoiler: idle

View attachment 2555103

Really hoping i could fix this without having to ultimately buy and replace my gpu for my rig as this is kind of stressing me out a bit...

1) Use Thermalright TFX paste.
2) Use Gelid Extreme Pads as per EK sizing given in manuals. Stock EK pads are ****ty.

Your temps looks more like a mounting issue to me that anything else. Did you use the correct thickness of pads? AFAIK, front part of the card is mainly 1mm.

 

Spoiler: Front gpu

Spoiler: Front gpu block

Spoiler: Active backplate gpu block

these are the pictures that I took after draining the gpu waterblocks and disassembled it to remove the ek pads so that I can put on new gelid pads and tk paste. Honestly can't tell from which side is not getting enough pressure from the mount...

 

Just adding to the thread my experience.

I have just done thermal re-paste on the RTX 3090 Surprim X using some paste I had lying around , cheap G40 5.2W/m-K)

Dropped hotspot temperatures 30+°C on small overclock on the AIR cooler.If temperatures creep back up after awhile no big deal to re-paste again.

Went from 104.5 °C down to 73°C

Results after thermal re-paste only. Pads did not need to be changed,they were fine.

 

Just adding to the thread my experience.

I have just done thermal re-paste on the RTX 3090 Surprim X using some paste I had lying around , cheap G40 5.2W/m-K)

Dropped hotspot temperatures 30+°C on small overclock on the AIR cooler.If temperatures creep back up after awhile no big deal to re-paste again.

Went from 104.5 °C down to 73°C

Results after thermal re-paste only. Pads did not need to be changed,they were fine.

In what ambient temperature is that? thanks

 

@larvabom
Thanks for the information

 

So I dissasembly the RTX 3080 12gb Gaming Z to see why hotspot in some benchmarks were to high compared with games.

Games around 10C delta, some benchmarks around 18C delta and aida64 +20C delta.

For me dosent look very bad or Im tottaly wrong ? Where was the issues ?

@Falkentyne can you take a look here pls as I see you helping many of us. Thanks

 

 

VRM pads look perfect.
GPU looks suspect. Looks like low contact pressure. Too much paste is on the GPU core when removed. Most of the paste should be pushed completely off the die, leaving only a very thin layer. I think this is a pressure problem.
Or.....
Why are there gaps in the heatsink????
That is very bad...any gaps on direct die is terrible ( not as important on a IHS covered CPU).