Replacing the thermal paste helps temps?

[Patrick Wolf]

I have the 9800 GX2 OCed to 715/1050 (stock 600/1000). My room temp is 72F and right now it's idle at 69C-GPU1/65C-GPU2. If I position my 90mm PCI slot fan blower where my floppy drive sits and blow cool air over the top of it, GPU2 can get down to 55C.

I'm going to add another 120mm fan to my CM 690 to blow cool air toward the GPU's intake, but I wonder if taking it apart and applying some AC5 would help bring down the temps?

 

[WaTaGuMp]

IF it did I doubt its in danger of being too hot with whats already on it.

 

[yh125d]

I'd be hesitant about taking that sucker apart, it's pretty complex

http://www.coolaler.com.tw/ima...ws/tnews_9800gx2_1.jpg

 

[Hauk]

Originally posted by: Patrick Wolf
I have the 9800 GX2 OCed to 715/1050 (stock 600/1000). My room temp is 72F and right now it's idle at 69C-GPU1/65C-GPU2. If I position my 90mm PCI slot fan blower where my floppy drive sits and blow cool air over the top of it, GPU2 can get down to 55C.

I'm going to add another 120mm fan to my CM 690 to blow cool air toward the GPU's intake, but I wonder if taking it apart and applying some AC5 would help bring down the temps?

You've set the card's fan speed to what? Card fan speed bump will yield measurable results. Blow all the air on it you want, move air through it instead! Are you bumping card's fan speed? With that room temp, bump fan speed to at least 55%, take a side panel off case, and report back on idle and load temps.

And no, do not take apart. Long story short, AS5 will yield marginal results. All other contact points use a thermal paste/patch that is not designed to be re-used, you'll never estabilish a factory bond again.

 

[Painman]

The thermal paste that most mfgrs currently use on GPUs is already pretty good. As stated above, don't crack open the factory thermal bonds unless you really need to. The pads on the VRMs are especially fragile and may not take re-application very well.

Just turn up the fans with a utility prog -- that will do a lot.

 

[chizow]

I've seen reports of 3-5C drops from re-applying thermal paste to the core and VRMs. Most reviewers or people who bother with this complain about too much thermal paste on the GPU, and similar to CPU, this can actually insulate the GPU. I'd agree that going through with this on a GX2 seems pretty complicated though, if you decide to do this I'd recommend a 2nd PC with detailed pics to help walk you through what to expect.

 

[Painman]

AS5 or other quality aftermarket TIM will do a better job -- no argument there. But I disagree about assertions made by reviewers, etc. that the stock TIM insulates the GPUs, provided that the board/cooler were assembled properly. Those stock coolers get HOT, even too hot to touch on some power hungry cards throughout GPU history (8800GTX comes to mind), and the reason they get so hot is because the TIM is doing what it's supposed to. Also note what happens when you give these coolers more fresh air -- core temps drop immediately and sharply. More evidence that the TIM is doing what it's supposed to.

If I had any suggestions to make to the guys/gals who engineer these things, it wouldn't necessarily be to use better/thinner TIM. They have their reasons to use what they use, and I can well imagine that those reasons have to do with mass manufacturing considerations (thicker, drier TIM is more stable, won't drool off of the coolers while they sit in a hot warehouse in Guangzhou, etc). I'd just tell them to set the cooling fans so that they ramp up before the cards get so damn hot to begin with.

So... if an end user wants the best thermal performance they can get, and they're comfortable taking the board/cooler apart and reapplying TIM, go for it... but I think the stock TIM is generally OK. Like I said, I just wish they'd be more liberal with fan speed settings so that end users had less to feel anxious about when they loaded up GpuZ and looked at temps.

 

[chizow]

Well you can't really make a conclusion based on the cooler getting hot, the stock paste is definitely working, but if it were working better, the cooler would get hotter as more heat would be drawn away from the GPU itself, which is what you'd want with ideal heat transfer.

My comments aren't just based on what reviewers have found though, its also from experience with applying too much thermal compound, where re-applying can yield better results. I'm sure many others have had similar experiences when applying CPU paste, where too much actually results in worst temps.

 

[Patrick Wolf]

Originally posted by: yh125d
I'd be hesitant about taking that sucker apart, it's pretty complex

http://www.coolaler.com.tw/ima...ws/tnews_9800gx2_1.jpg

I saw that before considering taking it apart, and it is a bit intimidating. I believe I could do it, but as this would be my 1st attempt at something like this AND I would be dealing with not 1, but 2 GPU's... I may be over my head.

I don't want to shell out the $$$ for a new GPU if I break the thing so I'll just add 1 or even 2 more 120mm fans to blow cool air at it. That should work fine. I mean the card works fine as is, but I wish it ran cooler.

Also to the person who asked, I'm using EVGA Precision to OC it and have the fan speed set to Auto. default is 45%.

When I run 3DMark Vantage (High Preset) the fan cranks up to like 80% towards the end of the 2nd graphics test.

 

[chizow]

Case cooling can help a lot with these cards. If you have a side case fan, you may want to actually reverse flow to exhaust air rather than intake.

On the Antec 900 and 1200, the exhaust fan is almost directly above the GX2's exhaust port. You can also look at some slot/cyclone blowers on either side of the GX2 to help exhaust hot air from either side of the card.

For intake you can use more 120mm for sure along with some Antec Spot Fans. They go on sale a lot from MicroCenter for $9.99 no MIR and are great for pushing air to specific areas.

 

[WaTaGuMp]

Originally posted by: Patrick Wolf

Originally posted by: yh125d
I'd be hesitant about taking that sucker apart, it's pretty complex

http://www.coolaler.com.tw/ima...ws/tnews_9800gx2_1.jpg

I saw that before considering taking it apart, and it is a bit intimidating. I believe I could do it, but as this would be my 1st attempt at something like this AND I would be dealing with not 1, but 2 GPU's... I may be over my head.

I don't want to shell out the $$$ for a new GPU if I break the thing so I'll just add 1 or even 2 more 120mm fans to blow cool air at it. That should work fine. I mean the card works fine as is, but I wish it ran cooler.

Also to the person who asked, I'm using EVGA Precision to OC it and have the fan speed set to Auto. default is 45%.

When I run 3DMark Vantage (High Preset) the fan cranks up to like 80% towards the end of the 2nd graphics test.

I can say with pretty much 100% accuracy that any aftermarket cooling I have gotten for my video cards has been a total waste of money. The people designing the cards fully understand the temps they are going to run at, why would they sell something that ran so hot that it either died or caused issues. People wanna get better temps to feel better about it and I understand that also, cause like I said I have bought aftermarket cooling. I probably never will again though. I run the fan on my 4870 at 37% to lower temps, but I don't really HAVE to, a little safety doesn't hurt of course and that's fine for me as long as its free.

 

[Painman]

Originally posted by: chizow
Well you can't really make a conclusion based on the cooler getting hot, the stock paste is definitely working, but if it were working better, the cooler would get hotter as more heat would be drawn away from the GPU itself, which is what you'd want with ideal heat transfer.

My comments aren't just based on what reviewers have found though, its also from experience with applying too much thermal compound, where re-applying can yield better results. I'm sure many others have had similar experiences when applying CPU paste, where too much actually results in worst temps.

There are some TIMs out there that are designed to be effective when the application layer is relatively thick -- it's quite likely that what we typically see on GPU and PIB heatsinks is one of these types of materials. The unevenness of the IHSs that get slapped on top of processors is something that the folks in the thermal management business are attempting to address.

Have a look at this -- Shin Etsu sheet based phase change TIM

It's thick on purpose, but TBH, if its thermal conductivity is competitive with something like AS5, I'd be quite willing to give it a try... it just sounds cool.

The world of TIM is a diverse and wondrous place.

 

[chizow]

Originally posted by: Painman
There are some TIMs out there that are designed to be effective when the application layer is relatively thick -- it's quite likely that what we typically see on GPU and PIB heatsinks is one of these types of materials. The unevenness of the IHSs that get slapped on top of processors is something that the folks in the thermal management business are attempting to address.

Have a look at this -- Shin Etsu sheet based phase change TIM

It's thick on purpose, but TBH, if its thermal conductivity is competitive with something like AS5, I'd be quite willing to give it a try... it just sounds cool.

The world of TIM is a diverse and wondrous place.

Actually from what I've read, most agree copper/nickel/aluminum are all better at transferring heat than the TIM itself, so the goal would always be to have as little TIM as possible. The difference is TIM is better at transferring heat than air, so you'd want enough to ensure there was no air between the CPU and HS, which is where TIM comes into play.

Here's a link someone referenced from Cases/Cooling, couldn't find the original link thread though:
Dan's Data Goop Round-Up

He basically found anything works, as long as there's not too much and it provides good contact. This seems to confirm what reviewers and people replacing the TIM have found, there's just way too much TIM on the GPUs, which is worst at transferring heat than the IHS or copper heat sink.

Also, the link to Shin-Etsu you provided, that doesn't look particularly thick or excessive, it looks like the neat little squares of TIM found pre-applied to some heat sinks. Now compare it to this from a GTX 280:

Check out how much they apply, about 1/2 way down

 

[Painman]

Chizow, I'm not here for a pissing match. You seem like a pretty cool guy, I'm just trying to point out that product engineers for Intel, AMD, nV, etc. make the choices of TIMs that they make for very good reasons.

Some of them have to do with commodity logistics (TIMs need to be easy to apply, ship and store under various conditions).

Some of them have to do with cost/benefit ratios (why spend extra $ per unit ensuring perfectly flat IHSs when a company like Shin Etsu will sell you TIM that performs @ acceptable levels while tolerating uneven junction thicknesses).

All the while, these companies take thermal management VERY seriously, make no mistake about that. Their products are designed to operate faultlessly under conditions that the average ATer would consider criminally neglectful (think of how often the average PC user checks their CPU/GPU temps, let alone blows the dust out of their fans/case, and you are approaching the "norm" operating condition that is likely to be assumed by a Big Tech, Inc. thermal management engineer). The goal is to achieve thermal management solutions that hold up under less than ideal conditions, resulting in as few RMAs as possible due to overheat failure.

We here at AT freak out when our GPUs break 70º C. Truth is, we're nowhere near the tjmax for these GPUs... they're doing fine. nV pays people good money to make sure their products don't melt -- these people are professionals, they are well versed in their disciplines, their future employment depends on results, and I trust their judgements more than those of the crew @ Dan's Data. Or the casual observations of laypeople looking at a pic of Gray Goop on top of an IHS.

For the record, that Gray Goop is pretty good -- I've re-applied it while converting cards from WC back to AC, and it performs just fine, even after sitting in an anti-static bag for 2 years. I'm not surprised, TBH. Whoever made it, I'm sure it was qualified by nV to hold up for quite a number of years.

Bottom line: Can we stop second guessing the judgement of people whose job security depends on making sure nV, AMD, Intel, etc. don't look like a pack of amateur fools? Can we place a bit of faith in the OEM TIM suppliers they choose? These are deeply serious businesspeople we're talking about.

 

[Hauk]

Hey OP, yea regarding the card's fan speed, leaving set to auto isn't a bad thing, idle temps will be higher than if running at say 60% minimum.

Chizow mentioned case cooling and that's where I was headed with the suggestion to remove a side panel and monitor temps. Moving a larger amount of air through the case is indeed necessary. Follow the Antec designs over the last couple years, from dual 120mm to triple 120mm + 200mm exaust fans. Moving air in and out of the case helps bigtime. A case should be effective that ambient temps inside nearly match your room temp. That's proper cooling.

Regarding TIM, despite the discussions I'd still say don't attempt it. If you've never worked on removing and applying new TIM, and sandwich card will be the giant you wish you never took on. I've done many and was going to put together a pictorial. I will someday. I use 99.9% pure alcohol, AS5 for GPU, and Arctic Ceramique for everything else. I do it mainly to check the quality of the factory work, mild temp drops are a bonus..

 

[Zap]

Originally posted by: Patrick Wolf
I have the 9800 GX2 OCed to 715/1050 (stock 600/1000). My room temp is 72F and right now it's idle at 69C-GPU1/65C-GPU2. If I position my 90mm PCI slot fan blower where my floppy drive sits and blow cool air over the top of it, GPU2 can get down to 55C.

My thoughts... first, that's a terrific overclock... and the overclock adds to the heat.

Second, your experiment with the slot fan cooler indicates that your system can benefit from better case ventilation. I don't think you'd get a whopping 10ºC difference from changing thermal compounds.

 

[Patrick Wolf]

I have an update. Someone on the EVGA forums removed the plastic casing on his 9800 GX2 and said it helped tremendously. He also said that replacing the thermal paste/pads helps even more.

I thought it made sense to remove the casing as it seems like it would do nothing but trap heat inside the card. Well I didn't want to mess with the thermal goop so I just removed the case and here's the results.

I have a 120mm side fan as intake, I know people suggest outake works best for this card, but from my tests that's definetly not true. (maybe it depends on case and room temp.) And another 120mm fan sitting on the floppy rack blowing cool air over the top of the card. Chassis is the CM 690. Air temp outside the case is normally 72F. Fan speed Auto.

Idle Temps w/ case: GPU1 63C / GPU2 59C
Idle Temps w/o case: GPU1 54C / GPU2 50C

Damn. A 9 degree drop. Now this I can live with. By the way removing the case was really quite simple for anyone willing to try it. The top cover is held together by some tiny screws and two larger ones on the end that are hidden under rubber caps. The bottom cover was a little trickier because it was getting caught up on something, may have been the power jack.