Jump to content

RTX 3080 GDDR6X temp concerns

Stahlmann
Go to solution Solved by DalekDoc,
On 4/12/2021 at 2:37 AM, Stahlmann said:

I did take my card apart and reassembled it one more time. But in my stupidity i forgot to make more pictures of the thermal pads. They were contacting the memory, but there was no real "dent" in the pads, so there was no real pressure. That at least explains why my VRAM was "just" hot and not overheating to the point of throttling. Reassembling my card didn't fix this issue. VRAM temps still sit at 95-100°C when playing demanding games at 4K.

 

They did ask for more pictures of my problem. But i haven't had time to pull apart my loop/GPU again since then. So the conversation with alphacool didn't really come to any conclusion yet. As i don't play any demanding game atm it's not really a priority for me at the time.

 

Did you contact them about this?

No, I never contacted them but noticed high memory temps. Initially I thought it was just the wrong backplate thermal pads (mine was initial batch where they sent all 2mm thermal pads, they've now updated it to send 3mm pads for 3080 on backplate vram area and 2mm for 3090) but was still having memory temps of 94-98c in nice hash.

 

Today I took the block apart and noted the thermal pads on the vram (on gpu chip side) were making contact but not a very strong impression. I replaced the stock 1mm pads with 1.5, did a test fit and noticed the thermal paste wasn't spreading too well so was worried the chip wasn't making contact with the coldplate. I just pressed down really firmly on the pads to thin them out a bit and then cranked the screws very tight (tighter than I should have probably). Noticed a very slight bow in the pcb with this but after getting my loop put all together, my gpu core temp is now 3-4c higher (likely the memory heat sinking into the coldplate properly) but memory temps are 30-35c cooler!

 

IMO this needs like 1.2-1.3mm thermal pads but no one makes those but the 1mm is clearly not enough.

19 hours ago, DalekDoc said:

Did Alphacool ever get back to you and was your vram temp issue resolved. I'm having similar issues with my Zotac 3080 + alphacool block..

I did take my card apart and reassembled it one more time. But in my stupidity i forgot to make more pictures of the thermal pads. They were contacting the memory, but there was no real "dent" in the pads, so there was no real pressure. That at least explains why my VRAM was "just" hot and not overheating to the point of throttling. Reassembling my card didn't fix this issue. VRAM temps still sit at 95-100°C when playing demanding games at 4K.

 

They did ask for more pictures of my problem. But i haven't had time to pull apart my loop/GPU again since then. So the conversation with alphacool didn't really come to any conclusion yet. As i don't play any demanding game atm it's not really a priority for me at the time.

 

Did you contact them about this?

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

On 4/12/2021 at 2:37 AM, Stahlmann said:

I did take my card apart and reassembled it one more time. But in my stupidity i forgot to make more pictures of the thermal pads. They were contacting the memory, but there was no real "dent" in the pads, so there was no real pressure. That at least explains why my VRAM was "just" hot and not overheating to the point of throttling. Reassembling my card didn't fix this issue. VRAM temps still sit at 95-100°C when playing demanding games at 4K.

 

They did ask for more pictures of my problem. But i haven't had time to pull apart my loop/GPU again since then. So the conversation with alphacool didn't really come to any conclusion yet. As i don't play any demanding game atm it's not really a priority for me at the time.

 

Did you contact them about this?

No, I never contacted them but noticed high memory temps. Initially I thought it was just the wrong backplate thermal pads (mine was initial batch where they sent all 2mm thermal pads, they've now updated it to send 3mm pads for 3080 on backplate vram area and 2mm for 3090) but was still having memory temps of 94-98c in nice hash.

 

Today I took the block apart and noted the thermal pads on the vram (on gpu chip side) were making contact but not a very strong impression. I replaced the stock 1mm pads with 1.5, did a test fit and noticed the thermal paste wasn't spreading too well so was worried the chip wasn't making contact with the coldplate. I just pressed down really firmly on the pads to thin them out a bit and then cranked the screws very tight (tighter than I should have probably). Noticed a very slight bow in the pcb with this but after getting my loop put all together, my gpu core temp is now 3-4c higher (likely the memory heat sinking into the coldplate properly) but memory temps are 30-35c cooler!

 

IMO this needs like 1.2-1.3mm thermal pads but no one makes those but the 1mm is clearly not enough.

Link to comment
Share on other sites

Link to post
Share on other sites

4 hours ago, DalekDoc said:

No, I never contacted them but noticed high memory temps. Initially I thought it was just the wrong backplate thermal pads (mine was initial batch where they sent all 2mm thermal pads, they've now updated it to send 3mm pads for 3080 on backplate vram area and 2mm for 3090) but was still having memory temps of 94-98c in nice hash.

 

Today I took the block apart and noted the thermal pads on the vram (on gpu chip side) were making contact but not a very strong impression. I replaced the stock 1mm pads with 1.5, did a test fit and noticed the thermal paste wasn't spreading too well so was worried the chip wasn't making contact with the coldplate. I just pressed down really firmly on the pads to thin them out a bit and then cranked the screws very tight (tighter than I should have probably). Noticed a very slight bow in the pcb with this but after getting my loop put all together, my gpu core temp is now 3-4c higher (likely the memory heat sinking into the coldplate properly) but memory temps are 30-35c cooler!

 

IMO this needs like 1.2-1.3mm thermal pads but no one makes those but the 1mm is clearly not enough.

Thanks for the info! I'm gonna order some 1.5mm thermal pads and see how it works for me. I'll get back with my results in a few days.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

8 hours ago, Stahlmann said:

Thanks for the info! I'm gonna order some 1.5mm thermal pads and see how it works for me. I'll get back with my results in a few days.

You're welcome. FYI this is the thermal pad I used - https://www.amazon.com/gp/product/B08CGVZ4YG/

This size was just enough for the VRAM and one of the mosfet/VRM pads but not both so I left the mosfet and VRM pads the stock original and just changed the VRAM. I was nervous that lifting up the center of the block by 0.5mm may affect the contact at those sides, there are not temp sensors so I dont know but ran 3 hour stress test with GPU power maxed out and had complete stability so I think it's all good.

 

You could use the same thermalright pads as I did but if you find something a bit softer/more compressible, it might be better.

 

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, DalekDoc said:

You're welcome. FYI this is the thermal pad I used - https://www.amazon.com/gp/product/B08CGVZ4YG/

This size was just enough for the VRAM and one of the mosfet/VRM pads but not both so I left the mosfet and VRM pads the stock original and just changed the VRAM. I was nervous that lifting up the center of the block by 0.5mm may affect the contact at those sides, there are not temp sensors so I dont know but ran 3 hour stress test with GPU power maxed out and had complete stability so I think it's all good.

 

You could use the same thermalright pads as I did but if you find something a bit softer/more compressible, it might be better.

 

I ordered some Arctic pad. I always had a good experience with Arctic stuff in the past and it's not that expensive. For the start i'll only do the VRAM and see if it works. I'll probably leave the VRM stuff etc. alone for now.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

@Stahlmann

 

I have the same GPU and currently using the Bykski cover waterblock. Almost same temperature with yours in terms of gaming. I tried undervolt tonight and managed to get 48C core, 65 hotspot and 67 VRAM. 

 

 

Link to comment
Share on other sites

Link to post
Share on other sites

On 4/14/2021 at 4:58 PM, DalekDoc said:

You're welcome. FYI this is the thermal pad I used - https://www.amazon.com/gp/product/B08CGVZ4YG/

This size was just enough for the VRAM and one of the mosfet/VRM pads but not both so I left the mosfet and VRM pads the stock original and just changed the VRAM. I was nervous that lifting up the center of the block by 0.5mm may affect the contact at those sides, there are not temp sensors so I dont know but ran 3 hour stress test with GPU power maxed out and had complete stability so I think it's all good.

 

You could use the same thermalright pads as I did but if you find something a bit softer/more compressible, it might be better.

 

 

On 4/14/2021 at 5:20 PM, Imannudein said:

@Stahlmann

 

I have the same GPU and currently using the Bykski cover waterblock. Almost same temperature with yours in terms of gaming. I tried undervolt tonight and managed to get 48C core, 65 hotspot and 67 VRAM. 

 

 

 

So i've applied the 1.5mm pads now (only to VRAM, left all other pads alone) and now the VRAM get's to 62°C under full load (short term load). It worked. Only thing is i had to remount the cooler 3 times because i got shutdowns due to GPU overheating. Turns out the DIE did not make contact with the waterblock at all. Manually flattened the VRAM pads some more and remounted. Now everything is fine. Thanks again for the tip!

 

Simpler solution: Don't go with the alphacool block to begin with 😄 But glad i could fix it without buying a new block.

 

@DalekDoc

Edit: Played Cyberpunk for a few hours on the weekend. So now i have a picture where my temps end up under long load and when my water loop completely heats up.

 

GPU ends up at around 60°C, compared to the 54°C i've had before. (+6°C)

VRAM now ends up at around 72°C compared to the 92°C i've had before. (-20°C)

 

So again it's like in your case with higher GPU core temps. Idk if it's because of weaker contact between the DIE and waterblock due to the thicker thermal pads, or if it's because the heat from the VRAM is now properly going into the waterblock and thus additionaly heating the DIE a bit more. But as it only rises to 60°C now under several hours of continuous load it's completely fine and still better than most air coolers.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

15 hours ago, Stahlmann said:

 

 

So i've applied the 1.5mm pads now (only to VRAM, left all other pads alone) and now the VRAM get's to 62°C under full load. It worked. Only thing is i had to remount the cooler 3 times because i got shutdowns due to GPU overheating. Turns out the DIE did not make contact with the waterblock at all. Manually flattened the VRAM pads some more and remounted. Now everything is fine. Thanks again for the tip!

 

Simpler solution: Don't go with the alphacool block to begin with 😄 But glad i could fix it without buying a new block.

 

Cool, glad it worked out. Yeah alphacool really messed up with this.. could have easily fixed it by providing the right thickness thermal pads in the box...!

Since you're already in communication with them about the topic maybe you want to point them to this thread..

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 weeks later...
On 4/22/2021 at 2:06 PM, DalekDoc said:

Cool, glad it worked out. Yeah alphacool really messed up with this.. could have easily fixed it by providing the right thickness thermal pads in the box...!

Since you're already in communication with them about the topic maybe you want to point them to this thread..

So after a few days my GPU core temps are now at ~75°C at longer gaming sessions. While this is not very high by itself, it is very hot for water cooling. My guess is that the thermal paste that is already making suboptimal contact hardened out making the contact even worse.

 

So i went back and contacted alphacool again, this time including some of the pictures i took back when disassembling the card with only alphacool's thermal pads. Unsurprisingly they also don't have access to 1.2 or 1.3 mm thermal pads. But they'll send me some ultra-soft 1mm and 1.5mm pads that i should stack. Due to the softness they should help me getting better contact to the die while also contacting the memory. Also the customer support told me that this sadly is a known issue, but not entirely the fault of their waterblock and also linked me to a post from igorslab talking about how RTX 3000 GPUs suffer from uneven Dies, which makes aftermarket cooling more of a challenge.

 

I'm gonna wait for the new thermal pads to arrive and I'll get back to you with my results.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

  • 2 weeks later...
On 4/22/2021 at 2:06 PM, DalekDoc said:

Cool, glad it worked out. Yeah alphacool really messed up with this.. could have easily fixed it by providing the right thickness thermal pads in the box...!

Since you're already in communication with them about the topic maybe you want to point them to this thread..

So i got the special extra-soft thermal pads and now i'm finally getting the results i want. The GPU is properly contacting once again and the VRAM temps are also greatly improved compared to my starting point.

 

I was playing Cyberpunk over the last week or so and had HWinfo64 always running in the background. Even after hours and hours of gaming the max reported temps were 58°C core and 75°C VRAM. These are temps i can live with. So basically the same VRAM temps but 17°C lower core temps just by switching to softer thermal pads that had the same thickness.

 

I'm glad Alphacool reacted so quick and sent me the right thermal pads so fast. While they released a somewhat flawed product, the support was fast and effective. So this will not stop me from buying Alphacool products in the future.

 

I strongly suggest contacting Alphacool when you have VRAM temp issues with one of their waterblocks. They will quickly help you to resolve the problem.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

  • 10 months later...

Hi every one

 

I hope every one is doing good

 

I bough RTX 3080 FE & mining eth

 

in binging my memory temp was 95-100

 

i changed thermal pads & got 

GPU 0 3080 Speed 102.37 MH/s Shares 1095/0/0 3.852T Power 237 W Efficiency 431.94 KH/W

GPU 3080 core temp 50/ Memory temp 74 Fans 75 % Core 1440 Memory 10652 Kren 6

 

Explaining the method

 

I replaced front side thermal pads with NBA Thermal pads 15 W/mK of 2mm

As we known the recommended is 1.5 mm front side thermal pad but accidental did this 

in the back 2.0 mm & 1 mm

After that when Ming eth core temp was 70 memory was 50 -60

 

i undusted that my GPU is not contacting with heatsink properly so i open the card again & press all the thermal pads with my thumb in the front side & i get size around 1.7mm NBA Thermal pads 15 W/mK are little bit harder then Gelid thermal pads 

The Picture of GPU 0 Thermal Pads Front Side is old i forget to get Pic after parsing the pads 

 

The is my results is there HiveOs

GPU RTX 3080 core temp 50 Memory temp 74 Fans 75

Core 1440 Memory 10652 Power Power 237 W 

 

 

GPU 0
01:00.0
GeForce RTX 3080 10017 MB · NVIDIA
Micron GDDR6X · 94.02.42.80.77 · PL 100 W, 320 W, 370 W
102.4 MH 
fb74b3b4.svg237 W
Core1450
Memory 2800
PL 240
 

GPU 0 Thermal Pads B Side.jpeg

GPU 0 Thermal Past.jpeg

GPU 0 Thermal Pads F Side.jpeg

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×