Asus 780 Ti needs VRM cooling

bathrobehero · 16 Apr 2014 at 01:36

Hi, I have half a dozen ASUS GTX 780 Ti DirectCU II cards and I'm running them stressed 0-24h with a very conservative GPU temp target of 65°C without overclock. But lately I noticed that GPU-Z reports that even at that GPU temp the average Voltage Regulation Module temperatures are hovering over 90°C. Speeding up the fans doesn't have any effect on the VRM temperatures so I guess the heatsinks are not in touch with the VRM's. Based on this site that VRM temperature would kill the cards in 2-3 years on average but I plan on using them much longer, besides I'm not comfortable with those temperatures anyway. The cards are very far from each other (1m PCI-E extenders) sitting on a frame so airflow is not the issue.

I contacted Asus and asked if there was anything I could do to reduce the temperature without voiding the warranty and they said no.

So now I'm willing to void the warranty and do whatever you good people suggest me to do to bring down the temps by at least 20 °C.

Tealc · 16 Apr 2014 at 23:09

The electrolytic capacitors have only a limited function in a voltage regulator module and shouldn't be getting that hot, unless theyve dried out after several years of use. The VRM itself should easily be able to cope with more than 100 degrees. VRM are typically heatsunk so would be under a slab of aluminium.

bathrobehero · 17 Apr 2014 at 00:42

Thank you for the reply, the cards are only a few weeks old and I'm kind of decided that I want to cool them down, although first I might grab an infrared thermometer from ebay (any recommendations?) just to confirm. I also have a few Gigabyte Windforce 3 cards (I hate them though, poor quality fans and coilwhine) which doesn't seem to have VRM temp sensors built in so I guess a thermometer would be the best start.

Make · 17 Apr 2014 at 04:09

Well, the only way to get the temps down is too change the cooler really.. and that means going into watercooling. It will cost you a lot. Or, somehow blowing even cooler air into the cards, for example from outside with AC

dfour · 17 Apr 2014 at 12:15

Make said:
Well, the only way to get the temps down is too change the cooler really.. and that means going into watercooling. It will cost you a lot. Or, somehow blowing even cooler air into the cards, for example from outside with AC

I dont think cost is going to bother someone who has spent over £3000 on graphics cards alone :eek:

so depending upon your setup your looking at 6 ti blocks, 2 pumps, 2 res 4 or 5 meters of tubing (guessing here as we dont know how this is all connected) fittings throughout, 2 or 3* 140.4 rads (560) and fans to cool them

bathrobehero · 17 Apr 2014 at 13:42

It's an investment so I do care very much about the costs. Spending 2 cards worth on watercooling is not the way to go. Maybe if I would have problem with the core temps too, but those are just fine with the stock fan/heatsink.

I was thinking more like filling the gap between the VRM'S and the heatsink safely and I found out that I could probably use thermal conductive pads so now I'm looking into that.

Edit:
http://promos.asus.com/US/GTX700Series/DC2/
This site perfectly shows how things look like, except I don't think that tiny black pad is there because there's a 1-2mm gap. I checked it with a piece of paper which had plenty of space so there's no contact at all.

Tealc · 17 Apr 2014 at 14:32

Judging by the lack of contact plate to those VRM modules (they are the SMD components between the chokes and the electrolytic capacitors) cooling is a non-issue for the designer of the thermal solution. They are likey relying on airflow from the fan to stop any thermal runaway. Either that or a contact plate and sink is missing from those images.

Cards have become more efficient and better efficiency means less waste heat, and one of those improvements has been made with DC-DC converters which is what the VRMs are.

bathrobehero · 20 Apr 2014 at 14:46

Tealc said:
Either that or a contact plate and sink is missing from those images.

Thanks, and you're right, the plate was missing from the images:

So I ordered a few Phobya Thermal PadXT's (1 and 1.5mm) and did some experiments. First off I replaced the thermal pad under the heatsink that was missing from the pictures but it had no effect on the temperatures at all. I also cleaned the liquified thermal paste from where it didn't belong.
What worked quite well though, is to put a few layers of Phobya on top of the inductors(?) until it contacted the heatsink:

The reported max VRM temperature is now 81°C (~9 years) instead of 91°C (~2.8 years) and I'm satisfied with that. Obviously the cards won't last that long, it's not a gurantee in any way, but I'm happy to know they'll most likely die to something other than VRM.

I plan to use thicker pads on the rest of the cards instead of using more layers of the thinner ones. Then, I'll try to figure out the temps on the Gigabyte cards. Any recommendations for an infrared thermometer?

8 Pack · 20 Apr 2014 at 14:56

The area you have covered in blue are not what needs cooling. A heatsink already exists on the VRM and the contact looks fine from the pic.

I am sure your VRM are fine and the reported temps are incorrect. I push 1.6v on that VRM and my cards are running fine.

This area is the only one needing a heatsink:

bathrobehero · 20 Apr 2014 at 15:07

That area has a heatsink.
If the reported temps are incorrect I don't see why it would hover around 80°C. I also wouldn't mind overvoltage or 90°C VRM if I weren't using the cards 0-24.

The thing is, most of this is located under the fan's belly therefore there's no air going down there, only the blades are pushing air down on the sides which leaves me to believe that even as bad of a contact as I did with multiple layers of thermal pads is beneficial.

bathrobehero · 23 Apr 2014 at 22:11

Update:
I recieved the second batch of the thermal pads and decided to do the rest of the cards. While cleaning the liquified thermal paste off of them, I realized I was quite wrong about the position/size of the belly of the fan as it is barely covering anything. Obviously the cards are not designed to work on full load non stop, but even then I found it shocking that one of the cards were literally dripping thermal paste and they are not even 2 months old. Either way, I ended up with the hottest card hovering around 84°C from around 91°C. Not that great, but I take it. One of the cards and GPU-Z didn't get along well as the reported values for VDDC/VRM were all over the place (it's showing the presumably correct values like one third of the time), and it is showing lower temperatures, but it's still jumping around.

My guess is the thermal sensor is located somewhere in the middle and the thermal pads are just helping to distribute the temperature across the components.

Pictures: