Extreme Undervolting RTX 3090 bad for VRM?

[adm0n]

Hi,

I was looking to buy a RTX 4080 this generation, but seeing that it costs above 1500€ here, I instead opted to buy a used RTX 3090 for around 850€. I don't need that much performance, but I wanted to have above 12 GB of VRAM for productivity reasons. So I ended up with a Palit GeForce RTX 3090 GameRock. Since I don't need that much performance, I undervolted it. Currently I'm running it at 0.725V at 1500MHz with a power draw of around 220W.

 

My question is: If I play simple games like Rocket League, the card can go for a long time, without the fan even kicking in. Will this hurt the VRM in the long run? I sadly can't find any teardowns of the card, so I don't even know if the components on the card expect airflow, or if all important components are connected to the cooler as well.

 

Also, is there a tool to link the power consumption of the card to the fan speed? Ideally in combination with the temperature? Something like, start the fans if the power consumption is above 80W consistently for 10 seconds, but everything after that is controlled via the GPU temperature.

 

Thanks in advance

[podkall]

I'm not an expert, but,

 

VRM along other things is an electric component, such components are designed for a certain amount of power to run through them,

 

the components get hot because the power running through them, however, if the power or the tempature isn't too high it's perfectly fine.

 

RTX 3090 is a high end gaming card, the manufacturers keep in mind that the card components need to handle high power draw,

 

because who would buy a GPU that would die after couple of games of Rocket League? Nobody.

 

So basicaly, you should be fine, you can check temps of the card if you're worried, but I'm sure you're fine.

[IkeaGnome]

14 minutes ago, adm0n said:

so I don't even know if the components on the card expect airflow, or if all important components are connected to the cooler as well.

The VRAM on the back of the board use thermal pads to touch the backplate and use the backplate as the cooler for them.

https://www.kitguru.net/components/graphic-cards/dominic-moass/palit-rtx-3090-gamerock-oc-review/2/

As long as you're not crashing or over heating, you'll be fine.

[adm0n]

1 minute ago, podkall said:

I'm not an expert, but,

 

VRM along other things is an electric component, such components are designed for a certain amount of power to run through them,

 

the components get hot because the power running through them, however, if the power or the tempature isn't too high it's perfectly fine.

 

RTX 3090 is a high end gaming card, the manufacturers keep in mind that the card components need to handle high power draw,

 

because who would buy a GPU that would die after couple of games of Rocket League? Nobody.

 

So basicaly, you should be fine, you can check temps of the card if you're worried, but I'm sure you're fine.

I've just heard about cases, where exactly this was a problem. I think even Linus talked about it in a recent WAN show, But if I remember correctly it was about the 1000series. 

So I'm not sure if this is still a problem. Sadly there are no temperature sensors for the VRM, so I cannot check their temperature

[adm0n]

3 minutes ago, IkeaGnome said:

The VRAM on the back of the board use thermal pads to touch the backplate and use the backplate as the cooler for them.

https://www.kitguru.net/components/graphic-cards/dominic-moass/palit-rtx-3090-gamerock-oc-review/2/

As long as you're not crashing or over heating, you'll be fine.

Thanks for the Link, I tried looking for something like this
 

But I was talking about the VRM, not the VRAM.

[IkeaGnome]

1 minute ago, adm0n said:

But I was talking about the VRM, not the VRAM.

VRMs are only on the front and use thermal pads to contact the black base plate of the cooler. 

Blue is VRM and red is VRAM. All 22 phases of VRM are on the front of the card.

[adm0n]

3 minutes ago, IkeaGnome said:

VRMs are only on the front and use thermal pads to contact the black base plate of the cooler. 

Blue is VRM and red is VRAM. All 22 phases of VRM are on the front of the card.

This I understand. My problem is, that with the default fan curve, the fan only kicks in once in a while, cools down the card down to 30 degrees and then doesn't spin for the next 10 minutes. I'm scared that some of the components, that aren't connected to the heat sink, that expect there to be air flow, heat up too much under medium loads. I'm just not sure if this is a problem or not.

[IkeaGnome]

Just now, adm0n said:

This I understand. My problem is, that with the default fan curve, the fan only kicks in once in a while, cools down the card down to 30 degrees and then doesn't spin for the next 10 minutes. I'm scared that some of the components, that aren't connected to the heat sink, that expect there to be air flow, heat up too much under medium loads. I'm just not sure if this is a problem or not.

 

16 minutes ago, IkeaGnome said:

As long as you're not crashing or over heating, you'll be fine.

They are cooled even when the fans aren't spinning though. You're over thinking it. 

 

1 minute ago, adm0n said:

I'm scared that some of the components, that aren't connected to the heat sink

What components are those that you're worried about AND aren't connected to the heat sink?

[adm0n]

8 hours ago, IkeaGnome said:

What components are those that you're worried about AND aren't connected to the heat sink?

Now that I have a picture for reference, I'd say all the capacitors, the inductors labeled 2R2 (one is at the lower right of the card) and the one chip in the top right. There seem to be specific holes in the cooler, that allow air to pass to those components.

 

From what I found out from googling, thanks to the pictures you provided is that:
Capacitors are usually rated to operate at ~105°C for around 2000h of operation. For every 10°C they operate cooler, their lifetime doubles approximately.
Inductors don't really need to be cooled, but since they are close to the powerstage, they usually can conduct quite a bit of heat away from it, which is why they are often cooled.

 

So I don't need to worry about the inductors, but should I still worry about the capacitors? I would like them to last quite a bit

[TheDankKoosh]

1 hour ago, adm0n said:

Now that I have a picture for reference, I'd say all the capacitors, the inductors labeled 2R2 (one is at the lower right of the card) and the one chip in the top right. There seem to be specific holes in the cooler, that allow air to pass to those components.

 

From what I found out from googling, thanks to the pictures you provided is that:
Capacitors are usually rated to operate at ~105°C for around 2000h of operation. For every 10°C they operate cooler, their lifetime doubles approximately.
Inductors don't really need to be cooled, but since they are close to the powerstage, they usually can conduct quite a bit of heat away from it, which is why they are often cooled.

 

So I don't need to worry about the inductors, but should I still worry about the capacitors? I would like them to last quite a bit

Running the card with that aggressive of an undervolt won't have any of the vrm components going above 60c ever, high end gpus like this will typically get 105-125c 5k-10k caps which would probably run this card for at least 20 years with occasional maintenance, I'd wager something else would die on the card first with heavy use. 

[adm0n]

16 minutes ago, TheDankKoosh said:

Running the card with that aggressive of an undervolt won't have any of the vrm components going above 60c ever, high end gpus like this will typically get 105-125c 5k-10k caps which would probably run this card for at least 20 years with occasional maintenance, I'd wager something else would die on the card first with heavy use. 

Alright. Thank you for clarifying!