Water cooling for graphics cards

[Trigmaster15]

Hey guys,

I just recently built a water cooler onto my AMD Radeon 7950 3GB gpu, and I'm trying to figure out if some results I'm seeing are normal. I have an ideal situation, in which my wife's computer is running an AMD 7850 2GB (not identical, but similar) card, so I can *somewhat* compare my measurements. Please note - her card is NOT watercooled (just uses two stock fans).

One of the things I notice is that her card idles lower than mine. My card is always around 32-35 C, but hers never goes above 30 C unless she's running a game or something. That made me think that I had set something up wrong, so I decided to run FurMark on each (using identical settings) and compare the results. After only a few minutes, her card heated all the way up to 80 C, while mine slowly climbed to about 60 C. I ran a few trials and noticed the same results - her card consistently heated up much faster than mine every time.

Thinking this cemented my card's performance, I then noticed something else - when cooling down, her card consistently outperformed mine every single time. Her card could calm all the way down into the 30's C, while mine lagged behind in the 40's every time. Is this just an inherent nature of water cooling? Better performance under pressure, but slower recover time? I pulled my GPU apart and made sure all my connections were on tight, and thermal grease was well applied. What do you guys think?

*Also note - after 45 minutes of running FurMark, I topped my card out at 80 C.
* I am using an Arctic 7970 Hybrid Accelero water cooler.

 

[Trigmaster15]

Sorry one other thing - both of us are using identical cases (NZXT Phantoms) with the exact same amount of fans in the exact same places. Our rigs are basically identical in every way, except for my GPU, and my core is at 3.5 GHz while hers is at 3.4 GHz.

 

[Spaniard United]

Few ideas off the top of my head ...

- Too much/little paste?
- Try push-pull config on radiator fan(s)

For all intents and purposes it sounds like it's doing what it should. The single fan versus a push-pull configuration might be why it takes it longer to cool down when tested against a fan blowing it straight off.

Bit surprised it made it to 80 degrees after 45 minutes ... that doesn't sound right. Unless it's really hot where you live or your ambient temps are high, it shouldn't have hit 80.

 

[DRosencraft]

I would note that you're using a non-ideal water-cooling solution, in that it's a closed loop. As most will tell you, a closed loop system like yours is not going to provide a great deal of cooling performance relative to a custom open-loop system, which will cost a deal more than what you've already spent, and in some cases not be any better than a quality air-cooled setup. Personally I'm not surprised that your performance is so similar to your wife's, most of the difference probably attributed to your card's higher specs. I would keep looking around the net though to see if you can find anyone who has been using the same cooler and what performance they're getting. Really the only way to be certain if your experience is normal or not is to compare it to someone else using the same components or benchmarks for the system. The best recommendation I can give without having to shell out a ton of cash is try what Spaniard said and get a second fan for a push-pull setup on the radiator, or see about getting a fan with better performance.

 

[Trigmaster15]

Right, thanks for your responses. I get what you're saying - it's sort of just like car, factory performance vs custom will always be limited. However, I do have relative confidence that my gpu water cooler setup IS working, because I know from talking to others that when it's not, its very noticeable (e.g. idle temps 40/50+ C), and also the fact that mine does so much better than hers on a burn test. I actually flipped one of my 120mm fans to create a push pull setup, but I didn't see too much variance in the results.

As far as the paste goes, I just used exactly the amount that came in the syringe in the box. But you're right, it could be a possibility. I want to collect some more data and do some more research online; maybe someone else can give me a benchmark to compare against. Thanks for your responses,

Jon

 

[Spaniard United]

I Google-searched your cooler upon my initial reply and found several benchmarks. These indicated (more or less) similar performance to what you saw, short of the 80 degree part. None that I found saw temps get that high.

That being said, if I'm looking at the pictures and descriptions correctly, the full amount of paste applied may in fact have been too much and could be part of the reason for your less than stellar temperature performance. Too much thermal paste will effectively dampen the heat transfer, meaning the heat is struggling to move through the paste and into the heatsink/pipes/radiator/etc. Conversely too little paste will struggle to transfer the heat from the GPU to the heatsink effectively. You basically want a thin layer that covers the GPU. The old adage for paste is "Less is more".

The push-pull config for the radiator should also help to cool the water faster. It may not drop your operating temps by more than a degree or two if that, but I would think you'd see an improvement in the time it takes to cool back down when idle.

 

[coozie7]

@ Trigmaster15: Watercooling always takes longer to reach its stable temperature than air cooling, the mass of coolant takes time to heat up as does the radiator. When cooling down, that mass has to shed its heat, which again takes some time.
Where have you placed the radiator? Placement can have a big impact on performance.
If you used the whole syringefull you've used far too much thermal compound, a drop about the size of a grain of rice is plenty-just spread it out gently using a finger to form a fairly even film, pressure from the cooler baseplate will do the rest.
Excess thermal compound can be removed using paper towels moistened in either cleaning fluid-most computer stores now sell it-or PURE alcohol, Isoprplyl alcohol or surgical alcohol will both do the trick.
If you have used the whole syringe, first GENTLY CARFULLY scrape the excess off using something non metallic, strips of card from a breakfast cereal packet will do but do n't use anything stiffer or you may damage the card.