Question Cooling for an RTX 3060 Ti ?

[Robox]

For the cooling of a Gigabyte Eagle OC RTX 3060TI I have a water block available to use on the GPU (connected to a custom system) and I should dissipate the VRMs and the Memories with heat sinks and fans placed on it. I have some doubts about it:



- 1. Regarding the VRMs, should only those related to the GPU be dissipated or also the others present on the card? (for example the VRMs of the vrams)


(In fact, in addition to the 6 vrms next to the gpu on the left side, looking at the card from the front, there are two on the right side, possibly vrm of the vrams, and maybe one on the upper left side of the gpu)


(Also, the thermal pads that allow stock heatsink contact are only present on GPU vrms, as well as most video cards)


(if there are other components to cool besides vram and vrm, could you indicate them to me on the board?)



- 2. Only the VRMs or even the CHOKEs (the cube-shaped chips marked R15 next to the VRMs) need to be cooled (some boards also have the thermal pad on these, but I don't know if it's for safety when placing the heatsink or it's because they really heat up and need to be cooled down)



- 3. Which of these options is more efficient in terms of dissipation?



A. small dissipators mounted directly on the chips (vrm and vram) glued with a thermoconductive double-sided adhesive)



B. self-built aluminum plate, cut to size, to increase the dissipating surface, in contact with VRM, RAM, or both, using thermal paste (or a pad, for safety).


The plate (or several plate if it will be used only for the VRM andr for the RAM) will be fixed with screws, using the holes in the video card.


Heatsinks will be glued to the plate using thermoconductive double-sided tape.


(is it worth having this greater dissipating surface and exploiting more heatsinks glued on top or does the mere fact of using pastes (or pads), double-sided tape lose the advantage? Furthermore, is it risky for any short circuits due to contact?)



Thank you

 

[Eximo]

Generally not worth it to water cool mid-range cards. But if you already have the components, I don't see why not.

Doesn't sound like you have a full coverage block.

The reason for covering all the VRMs and Chokes is to get as much heat out of the board as possible. Choke generate a little heat, but they are also large coils attached to the PCB by solder, so they soak up heat. Once saturated they can be used to extra additional heat.

Stacking is generally frowned upon, but more surface area always helps. So building your own spreader and then putting heat sinks on it isn't terrible, but heatsinks of the appropriate size are likely more efficient if you are going to have a fan pointed at them.

Of course you can short out stuff, you just have to pay attention about components that are taller than the things you are attaching metal to. Thermal pads and pastes are generally not electrically conductive, at least not enough to matter.

 

[Robox]

Generally not worth it to water cool mid-range cards. But if you already have the components, I don't see why not.

Doesn't sound like you have a full coverage block.

The reason for covering all the VRMs and Chokes is to get as much heat out of the board as possible. Choke generate a little heat, but they are also large coils attached to the PCB by solder, so they soak up heat. Once saturated they can be used to extra additional heat.

Stacking is generally frowned upon, but more surface area always helps. So building your own spreader and then putting heat sinks on it isn't terrible, but heatsinks of the appropriate size are likely more efficient if you are going to have a fan pointed at them.

Of course you can short out stuff, you just have to pay attention about components that are taller than the things you are attaching metal to. Thermal pads and pastes are generally not electrically conductive, at least not enough to matter.

I have done the undervolt (1950Mhz and 0.925v) and the temperatures have decreased by 2°-4°C on the GPU and on the hotspot, in benchmarks such as heaven, timespy and firestrike, while they have decreased by 10°C in cyberpunk 2077 benchmark.

Better if I can improve the situation more, at no cost.

So, I put a water block (not fullcover) on the GPU and heatsinks on the memories and mosfets and I placed two 80mm fans at low speed, one on the mosfets and one blowing on memoirs (all things I already had). The temperatures of the GPU and the hotspot have now decreased by about 20-30°C.


I would just like to understand if heatsinks plus fans are enough to cool the mosfets and memories.

 

[Eximo]

Generally yes, that is how many of them get cooling on lower end models. Either with no heatsinks at all or passive sinks with the GPU axial fan(s) blowing down on them.

Maybe 3 or 4 generations ago they started including heatspreaders/mid-plates or directly cooling the VRM, memory, and chokes with the main heatsink.