Learning to seat the heat sink
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Posts: 28
- Joined: Sat Feb 16, 2008 3:00 pm
- Location: Southeastern USA
Learning to seat the heat sink
So far things are going well with my system, but I've found seating the heat sink to be a bit of a problem. (Scythe Ninja Rev B on a E8400, IP35-e mobo and AS5 paste.) Right now I'm running small FFTs in Prime with a very small overclock (3.2ghz at 1.104V) and both core temps are tracking together at 63C. Obviously, this is kinda high. But it's better than after my first attempt (temps heading toward 80C before I stopped the test after all of a few seconds) and even a bit better than my second attempt (66-68C).
The stupid pin mounting seems to be working OK (though I ordered a bolt-through kit just in case). The only thing that's really changing is that I'm using less and less thermal paste each time. Following the Arctic Silver directions for dual-core CPUs seems like an awful lot relative to the pictures I've seen. My coating is thick enough to create peaks upon removal without creating bare spots on either surface.
I have half a mind to shut down again and wipe off half of what's there just to see what the temps will do. I know that this CPU **should** be capable of idling in the 30s and staying in the 40s when loaded at stock speeds and voltages. (I also know that wolfdale temp readings have been whacky, but I've decided to believe mine since they seem reasonable and behave rationally.)
Anybody else have a similar experience on their first system build? Using thermal paste seems to be kind of a black art.
The stupid pin mounting seems to be working OK (though I ordered a bolt-through kit just in case). The only thing that's really changing is that I'm using less and less thermal paste each time. Following the Arctic Silver directions for dual-core CPUs seems like an awful lot relative to the pictures I've seen. My coating is thick enough to create peaks upon removal without creating bare spots on either surface.
I have half a mind to shut down again and wipe off half of what's there just to see what the temps will do. I know that this CPU **should** be capable of idling in the 30s and staying in the 40s when loaded at stock speeds and voltages. (I also know that wolfdale temp readings have been whacky, but I've decided to believe mine since they seem reasonable and behave rationally.)
Anybody else have a similar experience on their first system build? Using thermal paste seems to be kind of a black art.
-
- Friend of SPCR
- Posts: 356
- Joined: Wed Jan 16, 2008 8:56 pm
- Location: Council Bluffs, Iowa
- Contact:
Thermal paste is indeed a black art, but you've given yourself a great start by reading the Arctic Silver application guide. Just bear in mind that you want the smallest amount of paste that will prevent air gaps, and no more. Don't worry too much about using too little paste - your processor's thermal protection will prevent short-term damage. Just put some on (a very thin, even coat) and boot to see what your temperatures are. If they turn out to be scary, shut down and try again.
I'm not sure how many mount/dismount cycles the plastic pins are meant to take, but a dozen tries isn't going to hurt anything if you're careful.
I'm not sure how many mount/dismount cycles the plastic pins are meant to take, but a dozen tries isn't going to hurt anything if you're careful.
-
- Posts: 28
- Joined: Sat Feb 16, 2008 3:00 pm
- Location: Southeastern USA
Thanks for the feedback, crazy bob. I went ahead and re-seated one more time, wiping away the thermal paste that came away with the heat sink and spreading the rest around by sliding the assembly around a bit. That appears to have knocked another 7C off the core temps, as they're now hovering around 56C.
Your way of explaining it is especially helpful, because it sounds to me like I still have too much paste on the chip. There definitely aren't any air bubbles (surfaces seem pretty good) and both come away with a healthy coating when separated. I don't think I have it in me to take it apart again tonight, but I may revisit this again tomorrow.
If I do wind up breaking a pin, I guess that will mean it's time to install the backing plate and spring screws. It should be here this weekend.
Your way of explaining it is especially helpful, because it sounds to me like I still have too much paste on the chip. There definitely aren't any air bubbles (surfaces seem pretty good) and both come away with a healthy coating when separated. I don't think I have it in me to take it apart again tonight, but I may revisit this again tomorrow.
If I do wind up breaking a pin, I guess that will mean it's time to install the backing plate and spring screws. It should be here this weekend.
Your repeated use of the word "spread" makes me nervous.
AS5 spreads itself under pressure and heat; you should either put a very small blob in the center of the CPU or a very thin line across the center, then let the pressure of the springs in the heat sink squeeze the TIM across the CPU/HS area.
It's really hard to use too little AS5, although I guess if you made a mission of it you could.
Regardless of whether you use the blob or line approach, the total volume of AS5 you should apply ought to be about 1/2 the size of a single short-grain rice grain. With the line method, this is a truly surprisingly small amount.
Regardless of line/blob, you should wiggle the heat sink a bit after installation, just to help the spreading.
AS5 spreads itself under pressure and heat; you should either put a very small blob in the center of the CPU or a very thin line across the center, then let the pressure of the springs in the heat sink squeeze the TIM across the CPU/HS area.
It's really hard to use too little AS5, although I guess if you made a mission of it you could.
Regardless of whether you use the blob or line approach, the total volume of AS5 you should apply ought to be about 1/2 the size of a single short-grain rice grain. With the line method, this is a truly surprisingly small amount.
Regardless of line/blob, you should wiggle the heat sink a bit after installation, just to help the spreading.
-
- Posts: 28
- Joined: Sat Feb 16, 2008 3:00 pm
- Location: Southeastern USA
That's basically what I did, didn't use any tools or anything. And I think the need for extra wiggling stems from the fact that I started with waaaaaaay too much, because as you say it's a "surprisingly small amount." I applied a line about as thick as is shown on the AS5 application photo, then wiggled a few degrees to each side per the instructions and called it done. (That's what got me loaded temps pushing 80C.) From what I'm hearing (and seeing in my own temps) I should have divided that by a factor of 5, making the line more like a strand of dog hair.cmthomson wrote: Regardless of line/blob, you should wiggle the heat sink a bit after installation, just to help the spreading.
On the bright side, one little tube of that stuff ought to to last an incredible number of applications.
-
- Posts: 28
- Joined: Sat Feb 16, 2008 3:00 pm
- Location: Southeastern USA
I declare the bolt-through kit to be the winner. I had increased to overclock to 3.65GHz (loaded Vcore=1.208 in CPU-Z) which was producing 63-64C in coretemp. (Passed 12 hours of small FFTs in Prime, BTW.) Not dangerous based on my reading, but a little warm. After installing the backing plate and spring screws, and freshly applying the AS5, it's settled at 57C. That should get a bit better as the compound breaks in.
Thanks for helping me along!
Thanks for helping me along!
-
- Posts: 28
- Joined: Sat Feb 16, 2008 3:00 pm
- Location: Southeastern USA