Replaced TIM on 285: Results and Questions

M

Mnx4

[H]ard|Gawd
Joined
Jul 29, 2005
Messages
1,039
Decided to make a new thread as the other one was getting cluttered.

So I decided to do some before and after tests on an EVGA GTX 285 vanilla. Wanted to see if there was a difference in temps between normal settings and overclock settings, then replace the TIM and see what if there was a change. Used Furmark, EVGA Precision, and GPU-z.

Before replacing TIM
Normal settings-
Core: 648 MHz
Shader: 1476 MHz
Memory: 1242 MHz
Tested at Fan Auto, 80%

Furmark 1:
Stability Test
30 Minutes (1800s)
Xtreme Burning Mode
1280*1024, 0xMSAA
Ambient: ~68F
Idle (Auto, 80%): 32C
Auto Fan (got to 57%): 81C
80% Fan: 70C

3dMark Vantage (default): 13307


Started messing around with the clock speeds on the 285. The shader seemed to be the most finicky. Used EVGA precision for clocks and fan and ran Furmark on stability test, increasing the different clocks in incremenets every 15 mins to get a feel for what it could do.
The card would not run past (instant blue screed):
Core: 775MHz
Shader: >1573MHz
Mem: ~1500MHz

Max stable OC:
Core: 745MHz
Shader: 1573MHz
Mem: 1462MHz

29x6v6f.gif


Furmark 2:
Stability Test
30 Minutes (1800s)
Xtreme Burning Mode
1280*1024, 0xMSAA
Ambient: ~68F
Idle (Auto, 80%): 33C
Temp: 70-71C

3D Mark Vantage: 14565

Tried leaving the fan on auto to see where it'd go. Wound up getting to about 60% speed (81C), and I figured it got so close to my normal speed, 80%, that there wasn't a point in trying the auto fan speed, so the final oc was at 80% fan speed.
The OC was stable for those two programs but not the games I currently play. Tried Fallout 3 and Left4Dead. Both felt faster/smoother than normal, but only lasted about 15-25 minutes before loop-crashing. The temperatures didn't get past 55-60C for L4D, 50-56C for Fallout 3.


So I got some Arctic Cooling MX-2 and decided to pry open the 285. Except for the screws, there was nothing actually holding the HSF to the card besides the suction from the TIM.

First opened
6eqe82.jpg


Cleaned with Q-Tips and 91% Isopropyl Alcohol.
24z9k0j.jpg


New TIM
qmx5dx.jpg


Aside from the copper portion, the other white silicon pads left no residue on the HSF. The components (ram, chips, etc) had some residue on them which was cleaned up fairly easily.

I used the dollop method as outlined by the user guide from Arctic Cooling. I used a bit more per area on the RAM and small chips than the core. I sat the HSF down on the card once I had applied TIM to all areas that had it before then pulled it back up to make sure that there was enough TIM to contact both surfaces.

Took about 2 hours from beginning to end.

The results?
Normal settings-
Core: 648 MHz
Shader: 1476 MHz
Memory: 1242 MHz
Tested at Fan Auto, 80%

Furmark 1:
Stability Test
30 Minutes (1800s)
Xtreme Burning Mode
1280*1024, 0xMSAA
Ambient: ~70F
Idle (Auto, 80%): 32C
Auto Fan (got to 57%): 81C
80% Fan: 69-70C

With the fan on auto, I started seeing the temps rise to the same level as pre-TIM replacement numbers and remain there for a few minutes, I went right to 80% Fan and let the test finish out to see if it got any lower than before. No dice.

Tried the max OC I got:
Furmark 2:
Stability Test
30 Minutes (1800s)
Xtreme Burning Mode
1280*1024, 0xMSAA
Ambient: ~68F
Idle (80%): 32C
Temp: 69-71C


So apparently there was no difference, which really surprises me. Those generic thermal pads and TIM looked absolutely caked on there and inefficient, so I figured at least having it a little neater would work better, regardless of TIM.

So some questions:
-The max OC didn't work with games but did with the stress test and Vantage. I'm guessing I couldn't keep it stable unless I do some volt modding? Any reason it'd work fine then crap out after 15-25 mins each time?
-From the pics, did it look like the application look fine?

I'll probably open up the card tomorrow after some game playing tonight, see if it looks alright. If anyone wants pics of the tests (Stability tests, Vantage), lemme know and I'll post em. Just didn't want to clutter this original post up a bunch.

I know this isn't a very robust review/test.I still have to find the max stable OC on those games. Anyway, any questions, comments or critiques are welcome.
 
Hopefully the heatsink is a better fit than the 8800GTX one. The pads you removed from the heatsink left a gap on the 8800.
 
The problem is there isnt direct to die contact. You are still dealing with a heat spreader and your temps / contact is only as good as what is underneath the IHS. Changing the TIM out on top of the heatspreader isnt going to make up for lousy contact underneath.
 
Risk to see if there's space? Gonna have to anyway or there's gonna be some long term damage, right?
 
Furmark is suppose to be a bad to the bone stress test. You'd think if it lasted that, then gaming wouldn't be a problem. However, it was. Maybe try different GPU stress test. Give ATI Tool's "scan for artifact" a go, and run it for an hour or so. http://www.techpowerup.com/atitool/
 
I did a search. The 285 supposedly has the IHS soldered on. You'll need to cut into the sides and then heat it up to remove the IHS.
 
yeah, my brother replaced the thermal pads on one of my 8800gts 320's and noticed the gap between the stock heatsink and the mem chips. to solve the problem, he placed copper shims between the heatsink and the chips and placed thermal paste on both sides of them. the video card dropped 3-5 degrees in Fahrenheit.
 
I tried using ATI tool but it won't recognize my card. I think I was using a 2006 version though, but I couldn't find and update.

And lucky777, the IHS being the metal plate on the core with the Nvidia logo and other information? Or that square bracket around the core?

Also, currently taking it apart, seeing if there's a space and if there's sufficient contact. I've checked out other sites and I haven't seen any complaints about there being a space for the 285 stock hsf. Then again, there aren't many people who've taken off the 285 hsf and reported it.
 
Well, good news: There's great contact with the core.
Bad news: There isn't contact with anything else....

ehmaa1.jpg


So, these tests like Furmark also heat up the ram? Or just the core?

Hope Fry's or Microcenter has some replacement pads... Petra's is currently OOS.
 
Thanks lucky777. So by removing the IHS, the rest of the stock HSF should fit better? Or were you just suggesting to remove the IHS, apply the new TM, then put it back on?
 
It should probably fit better. I don't really know though, I haven't tried it yet. If they don't, you could try taping 16 pennies in electrical tape and put those over the ram chips, see if they fit. If they're too big, try dimes.
 
first of all, why would you remove the stock pads if you were not ready to replace them with new pads? obviously they used those for a reason - thermal paste by itself does not fill the gaps, they needed something thicker.

second of all, I urge you not to attempt taking the IHS off the core just to have the hsf fit better (or whatever other reason you had in mind). You are either going to fuck up the core, or more likely, damage surrounding components while working on the IHS. Removing an IHS is difficult enough on a CPU, and that's a chip you can actually take out and move around.

edit: I also think you may have done some damage to your memory/VRMs by leaving them with no surface to transfer heat to while stress testing. If your card still works, that's good, but you never know what the consequences may end up being.
 
Removing the IHS is quite risky. Try the penny/dime thing I mentioned if you're only trying to make contact with the memory chips. If you're looking to lower core temps, removing the IHS is pretty much the only option (water cooling might help, but I've heard the IHS is the bottleneck).
 
Well, I had no idea there was a space between the HSF and the ram. I tried getting information like that from my other thread, but that got derailed fast...
I'm not going to run the card until I get replacement pads. Wasn't on for too long save for the stress tests and some web surfing, but still, that's long enough.
I'm probably not going to remove the IHS. From this thread, removing the IHS makes the stock HSF incompatible in terms of contact. The only reason to do that is applying a custom cooling solution.
I'm probably not going to try the penny trick. That seems too shady for me to even consider.
 
Mnx4 said:
Hope Fry's or Microcenter has some replacement pads... Petra's is currently OOS.
Click to expand...
Contact EVGA through their support. They've sent out replacement pads. Some have reported they sent them at no cost.
 
Well, no dice with EVGA. Called this morning and he said I'd have to buy them from somewhere.
How thick are these pads anyway? I don't have something accurate enough but they look like 0.5-1.0mm. Performance-PCs has some replacements for an 8800 (1mm) and 8800GT (0.5)mm
Gonna check out Frys and Microcenter, maybe even home depot for some thick copper strips...
Thanks for the help guys.
 
Kowan said:
Contact EVGA through their support. They've sent out replacement pads. Some have reported they sent them at no cost.
Click to expand...

They sent me some replacement pads when I owned an 8800GTS. Didn't seem too happy about it though.
 
You must log in or register to reply here.
Back
Top