7970 TIM replacement.

haste

Supreme [H]ardness
Joined
Sep 27, 2005
Messages
7,994
Saw something about a few cards from the factory seems to have a sloppy TIM application causing the cards to run a tad warmer than necessary. Anyone that has gotten a 7970 changed the stock TIM yet? If so, did you notice a temp difference?

Do we even have a good baseline as to where these cards temperatures should run at stock?


I should be receiving a Sapphire 7970 on Thursday and I'm just trying to get my ducks in a row. If I need to, I will take apart the card before I even install it if people have been seeing a significant drop in temps with a new TIM application on the reference 7970 cards...
 
Don't you have to redo the thermal pads for the heatsinks if you pull off the shroud and heatsink ?
 
Don't you have to redo the thermal pads for the heatsinks if you pull off the shroud and heatsink ?

That usually depends on whether they are stuck to the heatsink and/or the chips on the pcb. If the card is new, from my experience, i've found the pads stay on the card and the heatsink can be reapplied as normal after GPU TIM replacement is done.

If the card is older, the pads tend to be more brittle and less malleable therefore making it easier to damage them. Either way, utmost care needs to be taken not to rip/damage the thermal pads. I've never had to order any new ones and I'd like to keep it that way.
 
if the cards new the pads will usually stick to the heatsink so the pads really aren't an issue since they haven't gone through enough heatcycles to stick to the vrm's/ram chips yet..
 
I don't remember if the 4890 used pads, alhough I think it did, but there are no issues at all with pads on the 4890, 6970 and 7970 should follow the same basic design. Unless you have feet for hands, you should have no problem disturbing the pads by using a little common sense when removing the heatsink. The only issue you would ever have is if you tried to sand the heatsink or something...
 
One of my cards in crossfire runs ~10C hotter than the other. I am suspecting poor TIM application. Even reversing the cards the one stays much hotter so it is not a placement/air flow problem.

I suspect I should do a TIM replacement one of these days.
 
I'm scared to hell to try and change anything on a video card personally. The handful of times I've fooled around with 'em and this was done extremely carefully, the cards had artifacts. It's extremely easy to damage those memory chips, even taking a hair off the corner of those brittle GPU's can prove fatal. To be honest, I'm not even sure what I did wrong with the 2 or 3 cards I added 3rd party cooling solutions to but they were jacked up when I got done. Bad tim or not, I'm not touching my new Gigabyte 7970 when I receive it later today other than to install it and even then, I use anti-static gloves. Hell, I'm careful not to even look at my new computer parts wrong out of fear lol.
 
I don't know what you doing wrong when your taking them apart and putting them back together but your doing somthing. I have put at least 10 water blocks on video cards and after market coolers and never had an issue. Would I take the card apart just to replace the TIM probaly not unless I thought it would drop it a good 10c. Then I would probaly do it. Just my 2 cents.
 
I always found it easier to work with them after heating the card up a little - stops the sink from sticking as much and makes everything more pliable. Maybe try that in the future?
 
Ok an updated on my TIM replacement. I replaced the original TIM on my 7970s. It was poorly applied. It was a big blob of white thermal paste. It coverd pretty much the entire GPU and some on the entire socket. Cleaning the residue on the entire GPU plus socket took me 45 minutes. It was caked on the little transistors. Eventually with enough isopropyl alcohol 99% I eventually got it all off. I applied the tuniq TX-4. After some gaming I got a reported reading of 90C at full load on both GPUs in starwars the old republic. So I removed the Tuniq TX-4 and replaced it with MX-3. Now my GPU load is at 81C (in crossfire). Now the fan isn't as loud as it was with TX-4. Did it help me? Sure. I assume if everything is working fine for you stick with it. But if you're a neat freak and want the GPU to have the "correct" amount of thermal paste, this will drive you nuts. It was a sloppy job on both my GPUs.

found this in another thread, figured id throw it in here. this is tempting me to change out the TIM on my card...
 
I changed the Tim for some MX-4 and it made no difference. If anything, maybe temps were 1C worse. I guess I already had a good application from stock, since the vapor chamber was hot when I removed the card (had been running some heaven and furmark.
 
My stock Diamond 7970 was hitting 91c in an FT02 case with a fan blowing directly on it. I removed the shroud and heatsink (which by the way was the easiest I have done in a long time) and finally the vapor cooler. The thing is, 4 of the thermal pads on the memory tore in half. They are this white woven sticky stuff. Very thin and fragile. The die was caked in a hard grey thermal material. In fact, so miuch that the transistors on the die were completely covered. The contact zone on the heatsink is anything but polished as well. Very rough and visible milling on it. I cleaned it all with 99% Iso and reapplied AS5. The temps were exactly the same after the process. No benefit.

I returned the card yesterday for an exchange. The new card (same brand) runs 10c cooler and is able to overclock MUCH higher. In fact, the old card could not go 50mhz higher on core w/o a voltage increase. The new card can max out in CCC with no voltage and even leaving the +20% on zero. The guy at Frys did mention that he has had 3 returns in 4 days since they stocked the cards with people exchanging them for a new, same, card. He did not say why the customers had been returning them though. I'm wondering if this is a bigger issue than reported of poor chips being sent out due to insufficient supply at AMD. (As a note, the old card, even when not overclocked would overheat reaching 95c and shutting off, in addition to constant driver not responding errors)
 
Technically AMD is making chips so they work as advertised at 925/1300. We wantt he best bang for buck so we overclock them, they aren't warrantied to overclock and thats why those are being returned.

I hope my 7970 that im still waiting for OCs don't wanna send it back to california from toronto.
 
My stock Diamond 7970 was hitting 91c in an FT02 case with a fan blowing directly on it.

Well, you did get a not so great card this time, but do remember in case for the future that vapor chamber coolers aren't always so keen on being used in non standard orientations, such as that of the FT02.

http://forums.anandtech.com/showthread.php?t=2204972

You may know this, but just in case felt that it could be good to remember it to the [H] community :)
 
7970 Black Edition o/c to 1.125MHz GPU / 1.575 MHz Memory. Furmark burnin test:

82 C
42% fan speed

TIM seems ok on this one.
 
Well, you did get a not so great card this time, but do remember in case for the future that vapor chamber coolers aren't always so keen on being used in non standard orientations, such as that of the FT02.

http://forums.anandtech.com/showthread.php?t=2204972

You may know this, but just in case felt that it could be good to remember it to the [H] community :)

I had heard about this, but not seen an article. Thank you for that link. Curious if the 7970's design would qualify for this concern.

7970_cooler2.jpg


So this being a "vapor change cooler" with no visible heatpipes, would it be affected by vertical orientation?
 
Last edited:
Well, you can't see it but it is there. Best bet is to ask directly on the tech forum of a hardware maker.

I tried to look a bit on google, and didn't seem to find a pic or explanation about the 7970's :s so yeah, asking the card maker's tech support seems to be in order.

EDIT TO ADD:
Actually, since your new card (the replacement) was running 10ºC lower, i would think that it is working... but in the future will have to keep that in mind.
 
Well, you can't see it but it is there. Best bet is to ask directly on the tech forum of a hardware maker.

I tried to look a bit on google, and didn't seem to find a pic or explanation about the 7970's :s so yeah, asking the card maker's tech support seems to be in order.

EDIT TO ADD:
Actually, since your new card (the replacement) was running 10ºC lower, i would think that it is working... but in the future will have to keep that in mind.

Just fired off an email to Diamond MM for clarification.
 
Technically AMD is making chips so they work as advertised at 925/1300. We wantt he best bang for buck so we overclock them, they aren't warrantied to overclock and thats why those are being returned.

I hope my 7970 that im still waiting for OCs don't wanna send it back to california from toronto.

Why would you buy your 7970 in California if you are in Toronto?
They have been in stock in Toronto since day one in 5 retails stores at least that i checked, and more then 5-10 in some of them.. CanadaComputers, Infonec, Ncix etc...
 
May have spoken too soon, with a custom fan profile my 7970 stays at 80-82c in Furmarks 15 minute test and it stays stable. Thats with fans cranking up to about 85%.
 
May have spoken too soon, with a custom fan profile my 7970 stays at 80-82c in Furmarks 15 minute test and it stays stable. Thats with fans cranking up to about 85%.

Sounds pretty high... I'm running auto fan (which maxes at about 42-44% during long gaming sessions) with 1.2v, 1135core, 6000mem, and I only hit 75-77c max. This is of course with stock cooling (I'm on a reference Sapphire card).
 
My stock Diamond 7970 was hitting 91c in an FT02 case with a fan blowing directly on it. I removed the shroud and heatsink (which by the way was the easiest I have done in a long time) and finally the vapor cooler. The thing is, 4 of the thermal pads on the memory tore in half. They are this white woven sticky stuff. Very thin and fragile. The die was caked in a hard grey thermal material. In fact, so miuch that the transistors on the die were completely covered. The contact zone on the heatsink is anything but polished as well. Very rough and visible milling on it. I cleaned it all with 99% Iso and reapplied AS5. The temps were exactly the same after the process. No benefit.

I returned the card yesterday for an exchange. The new card (same brand) runs 10c cooler and is able to overclock MUCH higher. In fact, the old card could not go 50mhz higher on core w/o a voltage increase. The new card can max out in CCC with no voltage and even leaving the +20% on zero. The guy at Frys did mention that he has had 3 returns in 4 days since they stocked the cards with people exchanging them for a new, same, card. He did not say why the customers had been returning them though. I'm wondering if this is a bigger issue than reported of poor chips being sent out due to insufficient supply at AMD. (As a note, the old card, even when not overclocked would overheat reaching 95c and shutting off, in addition to constant driver not responding errors)

I think since mine seems to be in a fairly good place temp-wise that I won't bother trying to re-do the TIM :), thanks for the report.
 
Sounds pretty high... I'm running auto fan (which maxes at about 42-44% during long gaming sessions) with 1.2v, 1135core, 6000mem, and I only hit 75-77c max. This is of course with stock cooling (I'm on a reference Sapphire card).

My temps during long BF3 gaming sessions stay in the 60s.
 
Yea, running the fan high is kind of unbearable, although useful to see how temperature dependent a good oc is here. For me, it pretty much convinced me of dropping blocks on them, as there seems to be a fair amount of oc headroom to be gained.

And just to clarify unbearable, I mean it sounds like an industrial blower at 80%+. I'm sure there'll be some that will run the fan ramped up all the time, but I don't envy anyone living with them.
 
In my Level 10 GT case all 3 of my cards get to around 80*C and stay there while playing BF3 at Ultra settings. I added a 120mm fan to just push extra air into the area of the cards from the front, combined with the side fan built into the case and it's nice and quiet. Can't wait for my EK blocks to get here so I can water them up!
 
In my Level 10 GT case all 3 of my cards get to around 80*C and stay there while playing BF3 at Ultra settings. I added a 120mm fan to just push extra air into the area of the cards from the front, combined with the side fan built into the case and it's nice and quiet. Can't wait for my EK blocks to get here so I can water them up!

Jealous that you got a full set of blocks before they went oos. Which ones did you end up going with? Gong to get in by weekend so we can hopefully see pics of it soon :)?
 
Yea, running the fan high is kind of unbearable, although useful to see how temperature dependent a good oc is here. For me, it pretty much convinced me of dropping blocks on them, as there seems to be a fair amount of oc headroom to be gained.

And just to clarify unbearable, I mean it sounds like an industrial blower at 80%+. I'm sure there'll be some that will run the fan ramped up all the time, but I don't envy anyone living with them.

Yeah, it definitely seems these things like to be cool to stay artifact-free... once the temp hits a certain point at any given speed/volt for me basically it starts artifacting, hence why I'm running these lower speeds/voltages... fast, and quiet enough, but water would definitely let me get higher. I'd agree there's headroom to be filled by doing so for people.
 
Back
Top