RTX 3090FE overheating VRAM

[oz555]

I bought an RTX 3090 Founders Edition last December from Scan computers UK.

I have noticed the cooling fans have been running rather loud even though the GPU temps generally sit around 60 degrees while playing DCS. An internet search revealed many with the same issue and pointed towards the VRAM being the issue and a replacement of thermal pads, requiring some card disassembly to be a potential cure. 

I downloaded HW monitor and ran it while playing DCS. The memory temp quickly rose to 102-104 degrees, which is as I understand it, above the recommended temp limit.

I opened an online chat with an NVIDIA employee yesterday and was told that temps around 100 degrees for the memory was normal, however this is the first card I have owned which has spun up its fans so readily. Makes me a bit uneasy considering the current state of affairs with GPU's. 

Can anyone with knowledge confirm that this behaviour is indeed normal and not damaging the card?

[Burt]

96-100c with DCS seems to be the norm for the 3090. I use msi afterburner’s custom fan curve. Back when I was tweaking my settings that was the normal vram temps for me. I have also heard it was normal under load.

 

Edited October 27, 2021 by Burt

[Hector45]

As Burt said, 96-100C is supposed to be within spec for DDR6X memory. That being said it would definitely make me nervous what those temps will do to the lifespan of the modules. For contrast my hottest memory module sits around 67C at load.

I've seen some mods that have dramatically reduced the VRAM temps on the back of the 3090FE, which include better thermal pads like you said (proper thickness being most important) and also adding small heatsinks to the top of the backplate over where the memory sits. Though that kind of ruins the clean look of the FE card.

[Bossco82]

HI Hector45,

Have a look on Youtube for the JayzTwoCents video, 

might help give you some more info on your problem. If I had a 3090 I would want to do everything I could to assure its longevity. 

[Hector45]

5 minutes ago, Bossco82 said:

HI Hector45,

Have a look on Youtube for the JayzTwoCents video, 

might help give you some more info on your problem. If I had a 3090 I would want to do everything I could to assure its longevity. 

I think you meant to reply to the OP, not me. I've seen that same video and was referencing that mod in my post.

[Bossco82]

Hi Hector45,

yeah sorry mate I thought I had fully scrolled up for the OP's name. My bad!

[oz555]

Thanks for replies and video link. 

Have to say I'm not at all thrilled at the thought of stripping down my 3090. 

[Bossco82]

I get what you mean oz555. I have done similar before and fitted waterblocks to GPU's too. Its not that difficult. As long as you check temps afterwards your good.

Scan UK are a very reasonable retailer, as it is such an expensive GPU drop them an email and see what there attitude would be to the warranty etc if you were to do the thermal pad modification. Its the route I would take if I bought one. 

[javelina1]

I under volt mine, works great.  I have VRAM temps in the low to mid 80's for my gaming VR profile.  (GPU is usually around 47-50C).  For my daily "desktop" profile, I have VRAM temps in the middle 60's. 

[oz555]

Is there a DCS performance hit if I under volt? 

Edited October 28, 2021 by oz555

[Thinder]

Check your case venting and internal temperature.

If you want your internals to cool, you need the most efficient case cooling possible, and room temperature plays a role in this too.

I use a set of NOCTUA NF-F12 120mm, one for exhaust and one for feeding the GPU, the two other fans are standard 120mm coming with the case, the high pressure of the NOCTUA provide plenty of cooler airflow to the GPU without the deficit of too high airflow speed which can starve the GPU fans from air (boundary layer speed).

In short, if you fit a fan in front of the GPU with a too high speed airflow, it can "suck" the air away from the GPU fans, weither high pressure/low speed will feed them instead.

Edited October 28, 2021 by Thinder

[Burt]

On 10/27/2021 at 9:28 PM, javelina1 said:

I under volt mine, works great.  I have VRAM temps in the low to mid 80's for my gaming VR profile.  (GPU is usually around 47-50C).  For my daily "desktop" profile, I have VRAM temps in the middle 60's. 

That’s what I’m thinking of doing too. I may try to drop the power limit down by 10-20 % to see the difference. 

[Burt]

I watched this video and Im interested in Locking the frequency. 

 

[Bossco82]

I'm a big fan of undervolting. A custom curve in MSI Afterburner is a good route. 

I don't see it doing any harm in DCS. See what frequency you GPU maxes out at in DCS then match that with the voltage once you run the OC scanner. Not handled a 3090 personally but it worked on a 3080 and 3080ti. 

Plenty of guides on YouTube.