3080 FE Thermal Pad change: Mem Temps still 100c!! Help!

[uscool]

One thing I would expect the CPU getting warmer , if you have an AIO at the top exhausting , the back fan on the 3080 Fe you can feel the heat of airflow pushing upwards

 

[SentineL-X]

With how much heat mine produces in the case, I have the rad at the front, with 2 fans on it (120mmx2) and 2 exhaust fans, one at the top of the case and one at the rear. I replace the ones that came with the case as the RPM wasn't great, so now I have a custom fan curve on the exhaust fans to run at max settings just to get the hot air of the case, keeping case internals cooler.

Seems to work much better having made loads of changes post receiving the 3080.

 

[Falkentyne]

Yes the screw head was rounded off, but wasn't threaded. Sorry my bad, at the time couldn't think of the correct words for what had happened.

That screw has now been removed and just waiting to get a replacement for it.

Case wise I have the BeQuiet Darkbase 2 intake at front 1 exhaust rear. Using the fans that come with the case

Wait you were the guy with the stripped screw on the X-bracket or something?

 

[uscool]

Wait you were the guy with the stripped screw on the X-bracket or something?

No screw underneath magnetic cover

 

[Skynet5]

As another reference, I'm currently playing COD:WZ (on max graphic settings) with no mods to me 3080 FE
I'm getting the following temps:
GPU: 68
Hot Spot: 82
Mem Junction: 84

This is with a room ambient temp of around 22c.

I was getting avg 106 on mem junction and 100 on hot spot, but changed my 3600XT stock wraith spire cooler for an Artic freezer 2 AIO which has had a MASSIVE difference to the ambient case fan temp, resulting in a drop of around 20c on overage for the card itself. I'm not bothering with any pad mods for the time being given how this has helped...

Sorry to hear your troubles, but hope you sort!

That's not bad actually on the memory. Be interested in knowing temps with quake 2 rtx.

 

[Rhaos]

Wait you were the guy with the stripped screw on the X-bracket or something?

One of the screws the ones under the magnetic cover was rounded off so I couldn't unscrew it to redo the Thermal pads. Managed to get it out in the end

I red the reddit thread you linked about the TFX Fluu spread. Is 2mg tube not enough to do the job as I see you recommend 6.2g?

 

[RavenXXX2]

You aint gonna see much difference with different paste, you gone wrong somewhere else. A 3080 I repasted with mx4 has same temps as stock, throwing money away.

 

[Skynet5]

You aint gonna see much difference with different paste, you gone wrong somewhere else. A 3080 I repasted with mx4 has same temps as stock, throwing money away.

Yeah. I was thinking the same. There can only be very minor differences between top tier pastes.

Most will be spend on pads and paste than the card at this point

 

[koeiwong]

One of the screws the ones under the magnetic cover was rounded off so I couldn't unscrew it to redo the Thermal pads. Managed to get it out in the end

I red the reddit thread you linked about the TFX Fluu spread. Is 2mg tube not enough to do the job as I see you recommend 6.2g?

TFX is very hard to spread, and better to have another paste just in case.

 

[Falkentyne]

One of the screws the ones under the magnetic cover was rounded off so I couldn't unscrew it to redo the Thermal pads. Managed to get it out in the end

I red the reddit thread you linked about the TFX Fluu spread. Is 2mg tube not enough to do the job as I see you recommend 6.2g?

It's not about how much you can apply on the GPU (though you will be using most of that 2g tube!). Too many 1g/2g tubes of anything are way too inconsistent. I've seen dried out paste reports of 1g Kryonaut tubes, I have several Thermagic ZF-EX, 2g (seems to be a TFX clone) purchased from the same supplier at once (I think I bought 6?), where some were as "easy" (satire, maybe almost as easy to spread as SYY-157) to spread as TFX and some felt more like moist clay.
I've almost never seen reports of dried out paste in larger syringes, like 6g, 11.1g, 50g, and so on.

Here's how much TFX you need to apply to the GPU just to avoid it all sticking to the spatula! I started with a squiggly snake pattern then some dots to get areas that were starting to go back to the spatula :/

Got that idea from this guy on notebookreview who did this on his laptop to make TFX spread.

You may have to preheat the TFX syringe in extremely hot water or if you have one (this is faster), use a hairdryer at >60C on the entire mess after you squeeze it on the GPU, and spread it (80C seems best).

 

[Falkentyne]

You aint gonna see much difference with different paste, you gone wrong somewhere else. A 3080 I repasted with mx4 has same temps as stock, throwing money away.

Not true actually. I think it was koeiwong who said he had degrading temps fast with MX-4 and had it slower with MX-5 until he decided to yeet the paste application.
(TFX, SYY-157 and FuzeIce Plus or ALSEYE T9+ Platinum would have helped here).

Pastes don't matter much when the surfaces are nice and flush, flat and have strong contact pressure.
They *DO* matter immensely when you have weak pressure or convex or concave dies or heatsinks! There are so many paste reviews where they use Ryzens or Cores and MX-4 ends up like 2C worse than MX-5 and MX-5 is 1C worse than SYY-157 or something, with all the decent pastes within 4C of each other. There are abnormally few reviews about these pastes on weak pressure, convex laptop chips or heatsinks or on video cards.

Please read this thread. There's a goldmine of information about this issue there.
There's as much as a **10C** difference between MX-5 and (TFX / SYY-157 / FuzeIce Plus) all due to the integrity of the surfaces! And MX-4 is even worse.
On my laptop video card, MX-5 is only like 2C worse than SYY-157. But look at the below thread to see how this changes with a convex die, and a concave, weak fitting laptop direct die heatsink!

Kryonaut starts off great then pumps out/dries out fast. (Kryonaut Extreme "seems" to be more stable--it doesn't appear to dry out at all but STILL isn't thick enough to avoid pump out from expansion/contraction).

http://forum.notebookreview.com/thr...nd-apply-traditional-and-liquid-metal.806840/

I've been testing the "new" series of Shenzhen pastes (of which TFX, the thickest of them, is in this group). All of these pastes seem to cure in a strange way. Basically they actually -do- dry out intentionally, but the paste components don't "separate" or turn "hard"/brittle like we've seen from MX-4 or Kryonaut. Rather they seem to change into a harder, firm layer that coats the imperfections on the die, almost like some sort of phase change. I looked at this closely on my IHS CPUs. They all seem to mold into this equivalent dryer (but not brittle!) form that ends up stable long term, which is probably the resistance to pump out. But naturally this requires stable mounting pressure, as if the paste evolves this way and the pressure then decreases due to contraction when cooling or something, your temps go up and there's no more fluid paste to shift around anymore. So you want to get a layer that "cures" and doesn't start shifting around anymore when it hardens into a more thick form.

 

[Skynet5]

The issue here is less to do with paste quality and more to do with the ability to mount the cold plate to specification once the end user has ****** about with pads.

 

[koeiwong]

Not true actually. I think it was koeiwong who said he had degrading temps fast with MX-4 and had it slower with MX-5 until he decided to yeet the paste application.
(TFX, SYY-157 and FuzeIce Plus or ALSEYE T9+ Platinum would have helped here).

I didn’t use MX-4, pastes I used are MX-5, tfx and syy.

I can think of two reasons for degarding temps, 1) not enough paste because I only applied a thin layer, 2) I used Ultimate pads which maybe to hard causing the gpu not making enough contact with the cold plate.

 

[Major774]

Reference point for you, as I have done this mod to my 3080 FE.

• Gelid Extreme pads (2mm front and 3mm back).
• MX-4 paste.
• 3440 x 1440 - Cyberpunk 2077 - RTX gaming
• The delta between the GPU temp. and hotspot temp is 10/11C.
• The hotspot temps. runs in the 80C's.
• VRAM temps. runs in the 80C's.
• Undervolted: 1,850Mhz at 850mV.

Prior to the mod, I can only remember my fans running at higher RPM and VRAM temps. in the 100C's.

 

[Falkentyne]

Reference point for you, as I have done this mod to my 3080 FE.

• Gelid Extreme pads (2mm front and 3mm back).
• MX-4 paste.
• 3440 x 1440 - Cyberpunk 2077 - RTX gaming
• The delta between the GPU temp. and hotspot temp is 10/11C.
• The hotspot temps. runs in the 80C's.
• VRAM temps. runs in the 80C's.
• Undervolted: 1,850Mhz at 850mV.

Prior to the mod, I can only remember my fans running at higher RPM and VRAM temps. in the 100C's.

The problem is there were TWO people this week on the Nvidia sub who did 2mm on front, 2+3mm on backplate and had *black screen + 100% fans* (critical core shutdown) either when windows loaded or when trying to run a 3d app, and at least three other people who had 90C-100C hotspot temps. They had to remove the 2mm pads and replace them with 1.5mm, which kept their VRAM temps maybe 1-2C higher but dropped hotspot temps by 20C...

So this advice cannot be taken as gospel. It was this exact post very early this year which caused this problem.
Someone posted that 2mm Gelid Extremes worked fine and then suddenly all hell breaks loose.


I mentioned this before. I have NO idea if there are two revisions of the 3080 sink floating around. I don't work for Nvidia. I have no way to determine this and I sure as hell am not going to buy a card to find out.
Yes, you have to do whatever works obviously. For your particular system. But there have been way too many people in the last full week-2 weeks who have been complaining about this.

(this is a 3080 Ti fe thread here but there were just as many 3080 FE threads complaining about the exact same thing over there. Many were moderator deleted though).

https://www.reddit.com/r/nvidia/comments/ppz07q/repaded_3080ti_fe_worse_temps_few_questions/
https://i.imgur.com/7Qyn50M.jpg

The 3080 Ti FE heatsink is the same as the 3080 heatsink. People have posted date codes for the 3080 Ti FE heatsink going all the way back to December 2020.
But the user's HS in that thread is May 6, 2021.

Here was a 3080 FE's heatsink who was having high hotspots after 2mm Gelid extreme pad repad. He PM'd me thanking me when 1.5mm fixed his hotspot temps.

https://i.imgur.com/nGnXkFT.jpg

Here was one of the the original threads which got deleted. The OP PM'd me thanking me when he switched to 1.5mm pads (2mm on VRM's) which fixed his problem.
https://www.reddit.com/r/nvidia/comments/pq6pra/temp_issues_following_3080_fe_thermal_pad_mod/

Here was an even older thread. Read the comments. Black screen+100% fan on 2mm Extremes simply did not work on these guys 3080 FE's.
https://www.reddit.com/r/nvidia/comments/pptc9u/extreme_thermal_throttling_after_3080_fe_thermal/

kingartos

OP·6d

Possibly, but the disconnect for me is that a lot of people were able to get good results with the 2 mm. Something is weird, and i can remember when i opened it up at the first failed event, the 2 mm pads on the PCB were not even compressed . there was hardly any imprint of the memory chips. Is it possible the pads are old and less compressible as they age?

 

[SentineL-X]

That's not bad actually on the memory. Be interested in knowing temps with quake 2 rtx.

GPU: 74
Hotspot 85.9
Mem Junction: 96

I'll take that!

 

[Skynet5]

GPU: 74
Hotspot 85.9
Mem Junction: 96

I'll take that!

You likely have the best card built!

Mine are after 20mins of running around quake 2:

GPU: 71c
Hotspot 81c
Mem Junction: 100c

If I had 1 or 2 c off the mem junction I would have been much happier. But I think its all good. Certainly better than some peoples, especially after they tried to repad.

 

[uscool]

The problem is there were TWO people this week on the Nvidia sub who did 2mm on front, 2+3mm on backplate and had *black screen + 100% fans* (critical core shutdown) either when windows loaded or when trying to run a 3d app, and at least three other people who had 90C-100C hotspot temps. They had to remove the 2mm pads and replace them with 1.5mm, which kept their VRAM temps maybe 1-2C higher but dropped hotspot temps by 20C...

So this advice cannot be taken as gospel. It was this exact post very early this year which caused this problem.
Someone posted that 2mm Gelid Extremes worked fine and then suddenly all hell breaks loose.


I mentioned this before. I have NO idea if there are two revisions of the 3080 sink floating around. I don't work for Nvidia. I have no way to determine this and I sure as hell am not going to buy a card to find out.
Yes, you have to do whatever works obviously. For your particular system. But there have been way too many people in the last full week-2 weeks who have been complaining about this.

(this is a 3080 Ti fe thread here but there were just as many 3080 FE threads complaining about the exact same thing over there. Many were moderator deleted though).

https://www.reddit.com/r/nvidia/comments/ppz07q/repaded_3080ti_fe_worse_temps_few_questions/
https://i.imgur.com/7Qyn50M.jpg

The 3080 Ti FE heatsink is the same as the 3080 heatsink. People have posted date codes for the 3080 Ti FE heatsink going all the way back to December 2020.
But the user's HS in that thread is May 6, 2021.

Here was a 3080 FE's heatsink who was having high hotspots after 2mm Gelid extreme pad repad. He PM'd me thanking me when 1.5mm fixed his hotspot temps.

https://i.imgur.com/nGnXkFT.jpg

Here was one of the the original threads which got deleted. The OP PM'd me thanking me when he switched to 1.5mm pads (2mm on VRM's) which fixed his problem.
https://www.reddit.com/r/nvidia/comments/pq6pra/temp_issues_following_3080_fe_thermal_pad_mod/

Here was an even older thread. Read the comments. Black screen+100% fan on 2mm Extremes simply did not work on these guys 3080 FE's.
https://www.reddit.com/r/nvidia/comments/pptc9u/extreme_thermal_throttling_after_3080_fe_thermal/

I have 3080fe 2mm gelid Xtreme die side and 3mm on the back ( I think you could even use 2mm extreme gelid for the back ) and I don't have such problems how do you explain that ? Different heatsinks ?

My temps at highest at stock been running for 2 weeks 75c GPU with hotspot Delta 11-12 . My temps are lower with undervolt of 1905@ 893mv which mostly boosts to 1920 and max just barely hits 70c

Vram has had massive difference before would go high into the 90s now stays in the 80s , mining before instantly hit 110c now 82-86c

I used thermal grizzly I will keep an eye out if temps worsen if they do I will try TFX but for pads I will use the same as I've had no issues on what you reported

 

[SentineL-X]

You likely have the best card built!

Mine are after 20mins of running around quake 2:

GPU: 71c
Hotspot 81c
Mem Junction: 100c

If I had 1 or 2 c off the mem junction I would have been much happier. But I think its all good. Certainly better than some peoples, especially after they tried to repad.

I'm starting to think that myself reading so much about pad changes/issues with temps!

I had the same concerns before sorting out my case airflow/ambient temp, swapping to an AIO was one of the cheapest/easiest/best upgrades I've done

 

[uscool]

You likely have the best card built!

Mine are after 20mins of running around quake 2:

GPU: 71c
Hotspot 81c
Mem Junction: 100c

If I had 1 or 2 c off the mem junction I would have been much happier. But I think its all good. Certainly better than some peoples, especially after they tried to repad.

Thing is each person setup that could be purely down to case airflow and if using custom fan curve on the card and not the actual card being better airflow in case makes big difference the Fe cooling design is unique

I can set mine at 1905@893mv beats stock benchmarks and with more aggressive fan curve ( highest I set fans are 53%) the card will struggle to hit 70c and vram stays 82-86c pad mod