Vram temps reaching 108-110°C on watercooled gpu while mining ethereum

[RemiRanch]

When using phoenix miner, I am reaching vram temps of 108-110 C on my watercooled gpu, even with my fans and pump set to max speed. My GPU core temperature is fine though, as it only reaches temps of about 35-40 C when my fans are set to max speed.

Are these vram temps normal when mining? For reference, I am using a EVGA 3080 XC3 ultra, cooled with a heatkiller waterblock and a passive backplate.

[vf1000ride]

42 minutes ago, RemiRanch said:

When using phoenix miner, I am reaching vram temps of 108-110 C on my watercooled gpu, even with my fans and pump set to max speed. My GPU core temperature is fine though, as it only reaches temps of about 35-40 C when my fans are set to max speed.

Are these vram temps normal when mining? For reference, I am using a EVGA 3080 XC3 ultra, cooled with a heatkiller waterblock and a passive backplate.

That's pretty much maxed out for memory temps.  I would say it's time to pull the cooler off and figure out if there is a thermal pad out of place or that the cooler isn't mounted correctly.

[AnonymousGuy]

I sit at 65C right now with 0.1V increase to the memory and +1150 overclock while mining.  Pretty much as hard as you can run the card.  I use a Bykski block with Thermalright pads though so it's not the same thing.

 

The backplate shouldn't matter on the 3080, only the 3090.  What thickness pads did you use?  The VRAM sits 1.0mm off the PCB so depending on how tall the standoffs are on the waterblock, and how tall the memory section of the block is, you'd have to calculate if the pad thickness is correct or not.  And the thermal conductivity of the pads matters too.

 

You would have to disassemble the card to look at he pads and see the situation.  Pads that are too thick would probably not be an issue if you're seeing good core temps (thick pads can cause poor die contact).  Pads that are too thin would 100% be a problem.  Or pads that are low thermal conductivity.

[Kilrah]

It's normal, it'll be the temps of the chips on the back of the card. You can cool the front all you want, if you still have a passive backplate it's not going to change anything for those chips.

[RemiRanch]

1 hour ago, Kilrah said:

It's normal, it'll be the temps of the chips on the back of the card. You can cool the front all you want, if you still have a passive backplate it's not going to change anything for those chips.

Aren't the vram modules only on the front of a 3080 if I recall correctly?

[RemiRanch]

1 hour ago, AnonymousGuy said:

I sit at 65C right now with 0.1V increase to the memory and +1150 overclock while mining.  Pretty much as hard as you can run the card.  I use a Bykski block with Thermalright pads though so it's not the same thing.

 

The backplate shouldn't matter on the 3080, only the 3090.  What thickness pads did you use?  The VRAM sits 1.0mm off the PCB so depending on how tall the standoffs are on the waterblock, and how tall the memory section of the block is, you'd have to calculate if the pad thickness is correct or not.  And the thermal conductivity of the pads matters too.

 

You would have to disassemble the card to look at he pads and see the situation.  Pads that are too thick would probably not be an issue if you're seeing good core temps (thick pads can cause poor die contact).  Pads that are too thin would 100% be a problem.  Or pads that are low thermal conductivity.

I used the thermal pads that came with my heatkiller waterblock, which came in 0.5mm and 1mm thickness.

Could the high vram temp problem be caused by the included heatkiller pads being too thin for the memory modules?

[Kilrah]

8 minutes ago, RemiRanch said:

Aren't the vram modules only on the front of a 3080 if I recall correctly?

Oops indeed.

 

5 minutes ago, RemiRanch said:

I used the thermal pads that came with my heatkiller waterblock, which came in 0.5mm and 1mm thickness.

Could the high vram temp problem be caused by the included heatkiller pads being too thin for the memory modules?

Well the manual is pretty clear on what should go where, with 0.5 on the memory chips.

 

Probably not much to do then, either the fact pads are used at all is the limiting factor, or it's the chips themselves that have such thermal resistance that nothing can do much better. GDDR6X might have been the time to start making them bare dies like the GPU or HBM instead of plastic packaging...

[RemiRanch]

Could the issue have been caused by a general poor installation of the thermal pads, such as not putting enough compression between the memory, pad, and block? Or something like forgetting to peel the plastic on the thermal pads?

[Kilrah]

Well if installed incorrectly it can cause issues of course, yes...

Compression you apparently can't do it wrong unless you didn't tighten the screws, as to forgetting the peel - well yeah that would be a problem. 

[RemiRanch]

Yeah, I'm sure I've tightened the screws on my waterblock.

It's so confusing trying to find the cause of these vram temps...

[Guest]

1 hour ago, RemiRanch said:

Yeah, I'm sure I've tightened the screws on my waterblock.

It's so confusing trying to find the cause of these vram temps...

The thermal pads may not be making good contact or slid when assembling. 

Double check the thickness too.

If you have to pick up new thermal pads, look for high thermal conductivity cause most pads are poor. 

[AnonymousGuy]

You would have to disassemble the card to look at how much indentation is in the pads to see if its making good contact...or that you didn't goof forgetting to take off the plastic peel.  I would assume that Watercool spec'd the right thickness and it would be something I could easily verify in my workshop via measurement but absent that you have to look at them and see.  There's a comment on OCN by a Watercool guy that their pads are 3W/mK which is acceptable so I don't think the pad material would cause issue.  It's especially less important on thin pads like 0.5mm.  

 

It's pretty cheap and easy to just buy a spare set of pads (https://www.amazon.com/Thermalright-85x45x2mm-Conductive-Resistance-Silicone/dp/B08CGR62QX/ref=sr_1_4?dchild=1&keywords=thermalright&qid=1631231978&sr=8-4&th=1) and cut them to fit, if you do end up disassembling the card.

 

Also...are you sure you're looking at VRAM temperature and not VRM?

[RemiRanch]

1 hour ago, AnonymousGuy said:

You would have to disassemble the card to look at how much indentation is in the pads to see if its making good contact...or that you didn't goof forgetting to take off the plastic peel.  I would assume that Watercool spec'd the right thickness and it would be something I could easily verify in my workshop via measurement but absent that you have to look at them and see.  There's a comment on OCN by a Watercool guy that their pads are 3W/mK which is acceptable so I don't think the pad material would cause issue.  It's especially less important on thin pads like 0.5mm.  

 

It's pretty cheap and easy to just buy a spare set of pads (https://www.amazon.com/Thermalright-85x45x2mm-Conductive-Resistance-Silicone/dp/B08CGR62QX/ref=sr_1_4?dchild=1&keywords=thermalright&qid=1631231978&sr=8-4&th=1) and cut them to fit, if you do end up disassembling the card.

 

Also...are you sure you're looking at VRAM temperature and not VRM?

I've looked at the memory junction temperatures on hwinfo64, which list them at 110C, so I think its the vram temps that are the issue