Overclock.net banner
Cancel
Create thread New Forums

3090 GPU Hotspot Temperature - 30c Delta.

2 reading
Jump to Latest
41 - 60 of 146 Posts
Discussion starter · #41 ·
JungJungLLCatta said:
For all of you guys looking for a solution like I have been today. I have solved my issue on the 3090 Suprim X from MSI. The problem was, the hotspot was running 87 to 89 when gaming, even in games it previously had not reached 85, I had noticed in recent fays thatr the delta between the core and the hot spot has risen and now even reached 19 or 20 degrees. Guys: providing your thermal pads are in mint condition the solution is like so: Today I dissassembled my gpu cautiosly and wiped the old paste with some lint free tissue. Next I apllied theremal grizzly cryionaut meticulously so it covered the ehole day. I carefully assembled the card back. Now after 30 minutes of intense gaming I saw the highest core temp of 67 and the hot spot 77, the delta is now 10 degrees so this helped. Hope that helps. :)
Click to expand...
So on mine, after changing the paste like you did, hotspot would go back to 10c delta and look fixed, it would last a week or so like that, but gradually, it would start to get worse, first 12 then 15, then after a few days back to 20, then 25 than 30... I hope yours is fixed but given my experience with them I would have my doubts. I would keep an eye on it if I were you.
 
#42 ·
Gandyman said:
So on mine, after changing the paste like you did, hotspot would go back to 10c delta and look fixed, it would last a week or so like that, but gradually, it would start to get worse, first 12 then 15, then after a few days back to 20, then 25 than 30... I hope yours is fixed but given my experience with them I would have my doubts. I would keep an eye on it if I were you.
Click to expand...
Why were you gone for 2 months?
If your core temp is not increasing but your hotspot delta is increasing (this is important to check), then it's your VRM pads that are either degrading or don't have sufficient contact and are melting right at the outer layer of the pad. I determined since then that hotspot comes from BOTH core AND VRM's together, as the 'hottest' of the hotspots from either section (not average temp on a VRM cluster but the same type of thing as memory junction temp). But there is an absolute minimum delta allowed where it will never go lower than a certain C above the core. (For Founder's Edition 3090 FE cards this is 10C no matter what, seems to be 7C on 3080 Ti FE's). Try increasing the thickness of the VRM pads by 0.5mm and make sure you are using Gelid Extreme pads for best compressibility, but you can also try Gelid Ultimate pads for best durability (your choice). The VRM pads have less of an effect on the core contact because they are outside both the VRAM and core area, but they can still mess things up if they're too thick.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #43 ·
Falkentyne said:
Why were you gone for 2 months?
If your core temp is not increasing but your hotspot delta is increasing (this is important to check), then it's your VRM pads that are either degrading or don't have sufficient contact and are melting right at the outer layer of the pad. Try increasing the thickness of the pads by 0.5mm and make sure you are using Gelid Extreme pads for best compressibility, but you can also try Gelid Ultimate pads for best durability (your choice).
Click to expand...
I meant at the beginning of the year when I was using all the pastes available in my country. Since using the Thermarite paste you suggested, and not over tightening the mounting pressure, my delta hasnt moved. It started at exactly 12.8c delta and even right now [ I just checked ] it's identicle. The delta hasnt moved even 0.1 of a degree
 
#44 · (Edited)
Gandyman said:
I meant at the beginning of the year when I was using all the pastes available in my country. Since using the Thermarite paste you suggested, and not over tightening the mounting pressure, my delta hasnt moved. It started at exactly 12.8c delta and even right now [ I just checked ] it's identicle. The delta hasnt moved even 0.1 of a degree
Click to expand...
I edited my post while you were replying. Check the VRM pads.

I actually found a clever way to determine if it's the thermal paste/die pressure on the core, or the VRM' generating the hotspot issues.

Set your card to a low power limit, like 350W.
Then either turn off the fans or set them to minimum speed while running a paused or static benchmark scene at uncapped FPS, like Superposition 4k, Fortnite (lobby), Heaven, etc, basically anything with a static scene that doesn't change in power usage or FPS. Then have the card slowly heat up to 80C on the core, starting at 50C and going up slowly. (I assume this is easiest if you're air cooled, I don't know about people with water cooling loops, sorry).

If your hotspot delta remains within 0.5C at 40C to 80C, then it's not the thermal paste or die pressure causing problems. If it starts at let's say, 11C (Delta) at 40C, and goes up to 16C delta at 80C, at a low power limit (350W), then it's the thermal paste issue. The VRM's are not stressed that much at a low power limit. If it it starts at 11C delta and rises to only 11.5C to 12C, going from 40C core temp to 80C core temp, then your paste is fine.

Then, set your card to a high power limit, like 500W, and do the same test.
If your hotspot starts at like 11C at 40C and goes up to 16C at 80C, that means it's the VRM's.
One way to especially notice this, during a static scene, to see if the hotspot "jumps" back and forth like 1.5C up and down with the core temp remaining the same (stable ,e.g. 0.2C variance), but hotspot is bouncing all over the place. That's probably VRM's.
 
  • Rep+
Reactions: encrypted11
Save
Reply Quote Rep+ Rep+
#47 ·
Gandyman said:
I haven't really been gone, just been a overwhelemed with the baby and trying to rebuild my business in a post covid world as well as keep food on the table for my family =\ basically just personal stuff. Sorry for not being in touch
Click to expand...
It's fine. I was lucky enough to determine that "Core hotspot "gets its reading from the "hottest" junction section of BOTH the core AND VRM's. I found out because someone with a FE Card had a 25C (!!) hotspot delta but only 60C core temps. When he took apart the card, there was a COMPLETELY MISSING thermal pad section over part of the VRM's. Core paste and VRAM (GDDR6X) was fine on both. When he put on a replacement, it fixed everything.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #48 ·
Falkentyne said:
It's fine. I was lucky enough to determine that "Core hotspot "gets its reading from the "hottest" junction section of BOTH the core AND VRM's. I found out because someone with a FE Card had a 25C (!!) hotspot delta but only 60C core temps. When he took apart the card, there was a COMPLETELY MISSING thermal pad section over part of the VRM's. Core paste and VRAM (GDDR6X) was fine on both. When he put on a replacement, it fixed everything.
Click to expand...
That's very interesting. I'm quite sure I have a very "bended package" like Igor showed in his article. Since using the TFX the hotspot hasnt moved. I don't know wether it's actually fixed it, or wether it's now going to take longer than a less viscus paste to start going bad, but I hope it's fully fixed.

I have a client who has identical gpu to me [3090 strix oc] and 6 months ago or so when he bought it, his hotspot was fine, [stock paste + cooler] but about a month ago his went like mine did, all of a sudden he noticed super loud fans, even tho core temps rated the same, worked out it was the hotspot ramping the fans to 100%, looked for solutions, found me. I put the TFX on his stock cooler with stock pads and his seems to be fixed also [need more time to know for sure]

But seeing as 100% of samples [small sample size but still] I'm curious about the long term health of 3090s, seeing as in my experience dealing with the general public, most of them have bad airflow glass front cases because they look cool, never check hwinfo because they don't care let alone know it exists, and probably wouldn't think the fans pegging to 100% is anything abnormal.

My guess is lots of 3090s will be dead in a year or two.
 
Discussion starter · #49 ·
Falkentyne said:
It's fine. I was lucky enough to determine that "Core hotspot "gets its reading from the "hottest" junction section of BOTH the core AND VRM's. I found out because someone with a FE Card had a 25C (!!) hotspot delta but only 60C core temps. When he took apart the card, there was a COMPLETELY MISSING thermal pad section over part of the VRM's. Core paste and VRAM (GDDR6X) was fine on both. When he put on a replacement, it fixed everything.
Click to expand...
Also If you think 25c is bad I had a 32c delta at its worst with the stock thermal paste. Was 74 gpu temp and 106c hotspot temp after about 5 seconds of heaven benchmark
 
Discussion starter · #50 · (Edited)
@Falkentyne

how do you know it was the lack of pad that caused the hotspot reading? I would bet that his package was bent like mine, and after he found the missing thermal pad he also re applied thermal paste, which fixed the hotspot (at least temporarily) seems to me the finding of the missing thermal pad is a lucky cooincidence

my hotspot was going from 30c to 10c delta after changing thermal paste, but the normal noctua and TG pastes i was using must have lost its seal or something because after a few weeks it would slowly climb back up to the 30ish c delta.
 
#51 ·
Gandyman said:
@Falkentyne

how do you know it was the lack of pad that caused the hotspot reading? I would bet that his package was bent like mine, and after he found the missing thermal pad he also re applied thermal paste, which fixed the hotspot (at least temporarily) seems to me the finding of the missing thermal pad is a lucky cooincidence

my hotspot was going from 30c to 10c delta after changing thermal paste, but the normal noctua and TG pastes i was using must have lost its seal or something because after a few weeks it would slowly climb back up to the 30ish c delta.
Click to expand...
He's using the retail cooler. It's a Founder's Edition card. Not water cooling. And only a small section of pad was missing and there was no problem with core temps so no there was no 'bent package'. Otherwise core temps would not have been 60C on air. All of the FE cards have a 10-12C hotspot delta at stock, 10C delta is minimum by firmware. you're thinking way too much now. Relax.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #52 · (Edited)
Falkentyne said:
He's using the retail cooler. It's a Founder's Edition card. Not water cooling. And only a small section of pad was missing and there was no problem with core temps so no there was no 'bent package'. Otherwise core temps would not have been 60C on air. All of the FE cards have a 10-12C hotspot delta at stock, 10C delta is minimum by firmware. you're thinking way too much now. Relax.
Click to expand...
You don't understand what I'm saying.

I don't know how to explain it better, so maybe read it again.

What has water-cooling got to do with any of it.

It was you who linked me the igors lab article as an explanation to what was going on with my card, so dont pretend this is new info all of a sudden.

I'm not over thinking it, I spent 6 months trying to diagnose and track down and learn what was causing these hotspots that not only me but multiple other people ( friends and clients of mine) are experiencing. With your help included in that, I have been able to fix my hotspot and all of the ones in my local area. I have a good grasp of what is going on with them now due to sheer experience with it.

And yes it was your reccommendation on a viscus paste that isn't even available on my country that was the final solution to the problem. The uneven or bent package as Igor reported is causing normal viscosity pastes to lose contact after a while, causing hotspots. Re applying a less viscus paste will fix it temporarily, but over a few weeks it will get bad again.

I would argue presuming that a die sensor also includes vrm temperature in it's reporting is over thinking it...

Seems quite obvious that the neccassary paste re apply after his thermal pad change is what fixed the die hotspot..
 
#53 · (Edited)
Gandyman said:
You don't understand what I'm saying.

I don't know how to explain it better, so maybe read it again.

What has water-cooling got to do with any of it.

It was you who linked me the igors lab article as an explanation to what was going on with my card, so dont pretend this is new info all of a sudden.

I'm not over thinking it, I spent 6 months trying to diagnose and track down and learn what was causing these hotspots that not only me but multiple other people ( friends and clients of mine) are experiencing. With your help included in that, I have been able to fix my hotspot and all of the ones in my local area. I have a good grasp of what is going on with them now due to sheer experience with it.

And yes it was your reccommendation on a viscus paste that isn't even available on my country that was the final solution to the problem. The uneven or bent package as Igor reported is causing normal viscosity pastes to lose contact after a while, causing hotspots. Re applying a less viscus paste will fix it temporarily, but over a few weeks it will get bad again.

I would argue presuming that a die sensor also includes vrm temperature in it's reporting is over thinking it...

Seems quite obvious that the neccassary paste re apply after his thermal pad change is what fixed the die hotspot..
Click to expand...
I know exactly what you're saying. You're talking about the die.
All the dies are like this more or less. But why would a missing strip of thermal pad far away from the die (farther even than the memory is) make the curvature worse? That has nothing to do with it.
Have you even seen the 3090 FE's mounting system? It's nothing like your Strix (forgot if that's the card you have).
You're talking as if a missing section of pad far away from the X-bracket will make the PCB warp (which was mentioned on igor's article also) which you sound like is happening to you, from tightening, although I don't know.

Igor is water cooling. We all know this.. We had this discussion when Igor posted a picture of EK thermal pads and claimed they were the Aliexpress 120mm * 120mm Gelid Extremes...:/

I don't know why i keep needing to explain the basics though.
If the problem was die contact, his core (the reddit user with a stock FE and defective VRM pad section) wouldn't be ice cool. The average core temp would have risen also. I've seen countless people who used Thermalright Odyssey pads who saw their deltas rise over time due to low compressibility and low contact pressure on the core. But their core temps also rose as well. Someone on reddit just took a picture of bad contact pressure with 100C hotspot and high core temps and barely any paste spread being pushed off the core. The behavior of 'bended package' + low contact pressure + too thin thermal pastes was amazingly predictable over time, especially with how many people are using Kryonaut and MX-4, two of the most unstable pastes for imperfect low pressure mounts out there...

I already determined though spending WAY too much money myself that the VRM temps are also related to the hotspot. its NOT just the core. I spent money on this. And it takes time. When I mean time it takes weeks to months to find this out. It's not quick, it's not pleasant and I'm sorry it doesn't give people the fast results they want. Life isn't always convenient.

I saw my hotspot (but not core!) also get worse after 2 months when I was using 1.5mm gelid extreme pads for the VRM's. I know the original measurement was 1.8mm-2.0mm but they were also highly compressed after and figured ok 1.5mm was enough. I looked at the VRM pads after 2 months and saw this.



Compare the left side.


Nothing at all to do with "bended package".
That's from the pad pressure being too low and the heat collecting on the outer layer and causing melt marks since the entire pad isnt diffusing the heat evenly. The contact pressure on the left pad was quite nice (contact AND some compression!) but on the right, there's contact but barely any compression.
Also you can look at the picture of the 3090 FE review teardown on Igor's page from last September or October and see that all the pads with the melt marks were NVVDD pads :)

So I swapped just those VRM pads with 2mmm Extremes. Kept the 1.5mm for the VRAM. You can see how nice the imprint was on the memory so the size was fine there already.

I know you think I'm full of crap, but I messaged Igor on facebook about the VRM issues I found with the hotspot and he told me he was very interested and to tell him if I found anything else.
As I said you are seriously overthinking this. you are not the only person dealing with issues like this. I've spent way too much time helping people with this problem on their 3080 FE's and the bizarre inconsistent results some people are getting (2mm extremes =bad core temps, but 1.5mm=bad VRAM temps for one user, second user gets perfect results on the exact same card with 1.5mm extremes)...
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #54 ·
I don't doubt your experiences or knowledge, only doubted that you understood what I was saying.

I've had 6 customers with 3090s that all exhibited the same behaviour including one strix oc identical to mine. Making it 7 3090s in my small town that present with identical symptoms. Average temp doesn't go up, only hotspot. (But fan speed reacts to hotspot) It stays fine and consistent for a few months, then all of a sudden over only a few weeks at most the hotspot delta gets worse almost daily. As in for months it will be about 12c, then all of a sudden 18 one day 20 the next 24 the next 29 the next 32 the next. Out of all of them mine is the only one with a waterblock ( my customers are quite average consumers and not enthusiasts ) changing the paste to any of the pastes available in Aus will fix it for a short time, with the hotspot delta being 10c for a week or two at most, then the rapid degredarion occurs again ( without core temp changing) I ordered multiple 6g tubes of the TFX you reccomended, and so far [ fingers crossed] on all the cards (including mine with the warerblock) the hotspot hasnt changed yet. Nothing to do with pads is touched (mine is the only one I tried other pads on all the rest are what came with their respective brands). Changing to the firm paste fixed the problem entirely in 100% of cases so far. Granted none of them are FE as FE aren't available in my country. It's been an assortment of MSI gaming X trio, gigabyte gaming, gigabyte eagle, Asus tuf, and Asus strix. Interestingly, the first 3090 I ever sold in a client build ( it's a ichill X3) doesn't have the problem. Also these are all exclusively 3090s I'm yet to see it on a 3080 (although I see alot less 3080s due to their avaliability, all these people myself included bought 3090s because 3080s weren't available)

Make of it what you will but the TFX paste fixed all hotspot issues I've seen on 3090s. If anyone comes back to me with the hotspot delta increasing again after the change to the good paste I'll be sure to update you.
 
#55 ·
Gandyman said:
I don't doubt your experiences or knowledge, only doubted that you understood what I was saying.

I've had 6 customers with 3090s that all exhibited the same behaviour including one strix oc identical to mine. Making it 7 3090s in my small town that present with identical symptoms. Average temp doesn't go up, only hotspot. (But fan speed reacts to hotspot) It stays fine and consistent for a few months, then all of a sudden over only a few weeks at most the hotspot delta gets worse almost daily. As in for months it will be about 12c, then all of a sudden 18 one day 20 the next 24 the next 29 the next 32 the next. Out of all of them mine is the only one with a waterblock ( my customers are quite average consumers and not enthusiasts ) changing the paste to any of the pastes available in Aus will fix it for a short time, with the hotspot delta being 10c for a week or two at most, then the rapid degredarion occurs again ( without core temp changing) I ordered multiple 6g tubes of the TFX you reccomended, and so far [ fingers crossed] on all the cards (including mine with the warerblock) the hotspot hasnt changed yet. Nothing to do with pads is touched (mine is the only one I tried other pads on all the rest are what came with their respective brands). Changing to the firm paste fixed the problem entirely in 100% of cases so far. Granted none of them are FE as FE aren't available in my country. It's been an assortment of MSI gaming X trio, gigabyte gaming, gigabyte eagle, Asus tuf, and Asus strix. Interestingly, the first 3090 I ever sold in a client build ( it's a ichill X3) doesn't have the problem. Also these are all exclusively 3090s I'm yet to see it on a 3080 (although I see alot less 3080s due to their avaliability, all these people myself included bought 3090s because 3080s weren't available)

Make of it what you will but the TFX paste fixed all hotspot issues I've seen on 3090s. If anyone comes back to me with the hotspot delta increasing again after the change to the good paste I'll be sure to update you.
Click to expand...
TFX fixed my issues as well.
My remaining issues were from VRM's also affecting the hotspot, which I attributed to "TFX issues" originally. Like you, I didn't know what was going on.
However when I took apart the card to test the contact pressure with innovation cooling pressure paper (fujifilm Ultra Low Prescale), the contact pressure was even higher than it was on my Radeon card or my desktop CPU IHS! I didn't put 2 and 2 together until I remembered that reddit post about the ripped VRM pad and the 30-40C hotspot delta.
This then stared to make sense when I realized I was seeing hotspot wild swing oscillations by as much as 1.2C back and forth, while the core temp was sitting at like 0.1C...
That's when I took the card apart and saw the VRM pad on the right side.

then I looked at igor's picture (had to flip the orientation) to see that the melted areas were all on NVVDD VRM's...and guess what feeds the core? Yep.

I put 2mm pads on the VRM's instead and not only did temps drop on everything, power draw dropped also. (I could tell the power draw dropped and wasn't a false reading, because the three GPU 12v rails (not the PCIE power ones) weren't dropping as much either, so the VRM's were getting cooled better=less power wasted=less power used.

That's why I suggested that you check your VRM pads. But I didn't realize your hotspot was already fixed. I thought you had bad hotspot temps again.
I guess it was the thread bump and how you worded the first reply which made me think you were having issues again.

I guess it was this post, I thought you were having problems again. I didn't know it was fixed. That's why I made the comment for you to check the VRM pad contact pressure.

www.overclock.net

3090 GPU Hotspot Temperature - 30c Delta.

Edit: It's a 3090 Strix OC Hey guys With the inclusion of "GPU Hotspot Temperature" in the latest HWinfo, I have noticed my 3090 Strix OC has a about a 30c Delta between load core temp and hotspot temp 72c temp, 102c hotspot temp. Can others with 3090 show their hotspot temp with Heaven...
www.overclock.net www.overclock.net

It's really a mess that both Core AND VRM contact issues get reported to the hotspot.
 
Save
Reply Quote Rep+ Rep+
#56 ·
Falkentyne said:
TFX fixed my issues as well.
My remaining issues were from VRM's also affecting the hotspot, which I attributed to "TFX issues" originally. Like you, I didn't know what was going on.
However when I took apart the card to test the contact pressure with innovation cooling pressure paper (fujifilm Ultra Low Prescale), the contact pressure was even higher than it was on my Radeon card or my desktop CPU IHS! I didn't put 2 and 2 together until I remembered that reddit post about the ripped VRM pad and the 30-40C hotspot delta.
This then stared to make sense when I realized I was seeing hotspot wild swing oscillations by as much as 1.2C back and forth, while the core temp was sitting at like 0.1C...
That's when I took the card apart and saw the VRM pad on the right side.

then I looked at igor's picture (had to flip the orientation) to see that the melted areas were all on NVVDD VRM's...and guess what feeds the core? Yep.

I put 2mm pads on the VRM's instead and not only did temps drop on everything, power draw dropped also. (I could tell the power draw dropped and wasn't a false reading, because the three GPU 12v rails (not the PCIE power ones) weren't dropping as much either, so the VRM's were getting cooled better=less power wasted=less power used.

That's why I suggested that you check your VRM pads. But I didn't realize your hotspot was already fixed. I thought you had bad hotspot temps again.
I guess it was the thread bump and how you worded the first reply which made me think you were having issues again.

I guess it was this post, I thought you were having problems again. I didn't know it was fixed. That's why I made the comment for you to check the VRM pad contact pressure.

www.overclock.net

3090 GPU Hotspot Temperature - 30c Delta.

Edit: It's a 3090 Strix OC Hey guys With the inclusion of "GPU Hotspot Temperature" in the latest HWinfo, I have noticed my 3090 Strix OC has a about a 30c Delta between load core temp and hotspot temp 72c temp, 102c hotspot temp. Can others with 3090 show their hotspot temp with Heaven...
www.overclock.net www.overclock.net

It's really a mess that both Core AND VRM contact issues get reported to the hotspot.
Click to expand...
So which gaskets would you recommend to use on the 3090FE? With thermal paste everything is clear, it's TFX unequivocally. But what gaskets to use on the GPU side (VRM, VRAM) and which ones under the backplate - not so much. You write about using Gelid GP Extreme and Gelid GP-Ultimate, 1.5mm and 2.0mm. Can you tell me more about that. Thank you.
 
#57 · (Edited)
alloxa09 said:
So which gaskets would you recommend to use on the 3090FE? With thermal paste everything is clear, it's TFX unequivocally. But what gaskets to use on the GPU side (VRM, VRAM) and which ones under the backplate - not so much. You write about using Gelid GP Extreme and Gelid GP-Ultimate, 1.5mm and 2.0mm. Can you tell me more about that. Thank you.
Click to expand...
3090 FE with stock heatsink?
Front side: 1.5mm Gelid Extremes VRAM (memory), 2.0mm Gelid Extremes VRM's (VRM's are in three sections, third section is a single isolated small chip). NOTE: Gelid Ultimate 2.0mm may work on VRM's also, I'll test this when I move my system to a new build.

Back plate side: Gelid Ultimate 1.5mm everywhere.

With a custom loop? Consult the instructions. But I would probably still use Gelid Extremes on the front VRAM.
 
Save
Reply Quote Rep+ Rep+
#58 · (Edited)
Falkentyne said:
3090 FE with stock heatsink?
Front side: 1.5mm Gelid Extremes VRAM (memory), 2.0mm Gelid Extremes VRM's (VRM's are in three sections, third section is a single isolated small chip). NOTE: Gelid Ultimate 2.0mm may work on VRM's also, I'll test this when I move my system to a new build.

Back plate side: Gelid Ultimate 1.5mm everywhere.

With a custom loop? Consult the instructions. But I would probably still use Gelid Extremes on the front VRAM.
Click to expand...
Yes, with the stock heatsink.
Five days ago I replaced the stock pads with thermalright odyssey. I used MX-4 thermal paste. I did what was shown on one of the YouTube channels :) I used 1.5mm gaskets everywhere.
I didn't like the imprint on the VRM and VRAM where the GPU is. While it was still a little noticeable on the VRAM, the VRM didn't have any press-fit prints at all. But I put the video card together anyway.
And what I started to observe after a day of running the graphics card. At idle, the temperature on the GPU's hot spot would occasionally rise to 75C. At the same time, the temperature of the GPU itself did not exceed 36-40C. The fans were running and spinning at 800 RPM. Next, I decided to turn on the CP2077 game. The fans immediately worked at 100% GPU temperature was no more than 65C (peak), on the GPU hot spot it reached 106C (peak). Subsequent attempts to turn on any 3D application ended with an increase in temperature at the GPU hot spot to 106C. And this was happening in a matter of seconds.
Next I found this forum and your discussions in it. So I decided to ask a few questions.
I have now removed the video card, but have not disassembled it yet.
I ordered THX thermal paste. It's easier to buy. I will have to wait for the gaskets. But it's better to wait for the gaskets than to wait for the graphics card to fail.
That's my story.
 
#60 ·
Falkentyne said:
3090 FE with stock heatsink?
Front side: 1.5mm Gelid Extremes VRAM (memory), 2.0mm Gelid Extremes VRM's (VRM's are in three sections, third section is a single isolated small chip). NOTE: Gelid Ultimate 2.0mm may work on VRM's also, I'll test this when I move my system to a new build.
Back plate side: Gelid Ultimate 1.5mm everywhere.
Click to expand...
Yes... in this case the Thermalright odyssey pads disappointed me a lot. When I took the video card apart, I found no print on the VRM, there were still some prints on the VRAM. After your advice I ordered a set of 1.5mm Gelid Extremes VRAM (memory) and 2.0mm Gelid Extremes VRM. For the back panel I ordered Gelid Ultimate 1.5mm. I ordered TFX thermal paste. And what I got after the third day of my 3090FE graphics card. When playing CP2077, where the core temperature used to be 65-70C and the hot spot reached 106C, the picture now looks rosier. The core temperature did not rise above 53.4C and the hot spot temperature was 63.8C, the memory temperature was exactly 58C. The fans could not be heard at all, as they were running at 800 RPM. And that was the temperature range for 3 hours of play. Well, what can I say after all that, let's see what happens next.
But most importantly, thank you for your help in this situation.
You really helped me!






 
41 - 60 of 146 Posts
About this Discussion
145 Replies
30 Participants
Last post:
gamervivek
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling