TIM on GPU's

[MagR]

Hi

I'm about to try watercooling for the first time and wanted some advice on applying TIM. I use the single blob method with CPU's and it works well for me on Sandy Bridge CPU's (single smallish blob on middle then let pressure spread it out).

I was wondering if it is the same with gpu's? Are GPU's larger than CPU's (have not got stock air cooler off yet)? I saw a video where the chap used a central blob of TIM then four more towards the corners of the GPU. This seemed a lot of TIM to me but perhaps GPU's need more. I'm aiming to get this right first time as refitting a waterblock to change the amount of TIM later looks very time consuming.

Any thoughts would be appreciated.

Regards

Mag

[kalniel]

It's exactly the same as CPUs. They vary in size depending on model, but are in the same kind of range as CPUs.

[CAT-THE-FIFTH]

Make sure you have adequate cooling for the VRMs on the graphics card. In many cases it is the VRMs overheating which limits overclocking of a GPU or even its stable operation.

[piggin23]

Most manufacturers but more TIM on their chips than required.
It also varies by chip type ie the 680GTX is very small & needs only a very small amount & the 580GTX is a bigger chip so requires a little more.
Best to take the cooler off, clean everything up & put a little grain size blob on the chip. Then test fit the cooler, remove and check for good spread across the chip.

Also make sure you have the correct screwdrivers to remove the cooler.
nVidea now use 'thread lock' to stop the screws coming undone.
This makes things more difficult (not impossible) to remove but you must have the correct tools.

If you are doing a 680GTX, then look at the last page of my build, it may help
http://forums.hexus.net/chassis-mods...d-build-9.html

[Bugbait]

With full cover GPU blocks a little too much TIM is argubaly better than too little. Very few (none that I can think of) full cover GPU blocks use an active pressure mounting system (screws + springs) unlike most, if not all CPU blocks. This makes it harder to ensure the pressure is distributed evenly without risk of over-tightening. Sure, there are spacers but that assumes all sides are perfectly even and you're tightening the screws with a torque enabled screwdriver. Doesn't quite give the same reassurance as a set of springs on each corner IMO. I'd prefer a loss in efficiency from a layer that's too thick compared to having air. At the end of the day heat transfer from TIM is still greater than air.

EK full cover blocks recommend a cross pattern of TIM running from each diagonal edge.

[MagR]

It's exactly the same as CPUs. They vary in size depending on model, but are in the same kind of range as CPUs.

Thanks for the reply. It is the Fermi 110 GPU which looks abit big compared to the new 600 series so I might have to use a little more.

Regards

Mag

[MagR]

Make sure you have adequate cooling for the VRMs on the graphics card. In many cases it is the VRMs overheating which limits overclocking of a GPU or even its stable operation.

Thanks for your reply. I'm using full cover blocks from EK. I read a review on the older 400 series block by Skinnee and the EK block cooled the VRMs well. I hope it is the same for their 500 series block I am using.

Regards

Mag

[MagR]

Most manufacturers but more TIM on their chips than required.
It also varies by chip type ie the 680GTX is very small & needs only a very small amount & the 580GTX is a bigger chip so requires a little more.
Best to take the cooler off, clean everything up & put a little grain size blob on the chip. Then test fit the cooler, remove and check for good spread across the chip.

Also make sure you have the correct screwdrivers to remove the cooler.
nVidea now use 'thread lock' to stop the screws coming undone.
This makes things more difficult (not impossible) to remove but you must have the correct tools.

If you are doing a 680GTX, then look at the last page of my build, it may help
http://forums.hexus.net/chassis-mods...d-build-9.html

Thanks for your reply. I'm using EK waterblocks for two Pallit 500 series cards. The pictures I have seen show this is quite a large GPU compared to the 600 series so presumably will use more TIM. The Pallit uses a custom pcb with straightforward Phillips screws so I think I will be fine taking it off. I like the idea of doing a test first - I'll be doing it on a bench rather than in my chassis with a CPU already fitted to a motherboard as usual so it should be easy - why didn't I think of that?

As a follow up question do you need TIM over the whole GPU surface. When I use the blob method on a CPU the spread is circular with the corners exposed but this doesn't affect coverage of the cores so is fine. Is it the same with GPU's?

Thanks again for you help

Mag

[MagR]

With full cover GPU blocks a little too much TIM is argubaly better than too little. Very few (none that I can think of) full cover GPU blocks use an active pressure mounting system (screws + springs) unlike most, if not all CPU blocks. This makes it harder to ensure the pressure is distributed evenly without risk of over-tightening. Sure, there are spacers but that assumes all sides are perfectly even and you're tightening the screws with a torque enabled screwdriver. Doesn't quite give the same reassurance as a set of springs on each corner IMO. I'd prefer a loss in efficiency from a layer that's too thick compared to having air. At the end of the day heat transfer from TIM is still greater than air.

EK full cover blocks recommend a cross pattern of TIM running from each diagonal edge.

Thanks for your reply. I've never had problems getting good results on CPU's with air coolers without the springs you mention so perhaps the quantities of TIM I used were correct. I'm going to do a test fit following one chap's suggestion wihout actually using screws or thermal pads for VRM's etc so that may help me guage the amount. When I get my CPU block I'll make sure it has this active pressure mounting you mention as it looks like a great idea. I'm leaning towards an XSPC Raystorm at present so hope the mounting is as you suggest.

Thanks again for your help

Mag

[piggin23]

The die on the 580GTX is a lot bigger than the 680GTX.
I also show two 580 GTX's in my build
http://forums.hexus.net/chassis-mods...d-build-4.html
Just put a little more on the chip.
You could always put the TIM on & spread it around like a CPU

[MagR]

The die on the 580GTX is a lot bigger than the 680GTX.
I also show two 580 GTX's in my build
http://forums.hexus.net/chassis-mods...d-build-4.html
Just put a little more on the chip.
You could always put the TIM on & spread it around like a CPU

Thanks again. I'm feelin more confident now.

Mag

[Greecean]

with large chips (nvidia, larger ati, 2011) i usually use a bit of plastic to spread it so i know there is coverage all the way out

on smaller chips, i use the pea-spread method

[GoNz0]

with anything you need to do a test application 1st, you have to check your using the right amount. i tend to do a square of paste and put a cross through it, seems to work best for me on my 480's

also best to use none conductive paste on a GPU.

[MagR]

Thanks for your help

Mag

[MagR]

Thanks for your help. I'm going to do a test run without actually screwing down the block or using the thermal pads to test the spread.

Regards

Mag

[GoNz0]

you will need to screw down around the GPU to check the spread m8!

the pressure pushes it out and spreads it.