Proper heatsink prep

[synapse46]

I hand ground the bottom of the heatsink flat, switched from white thermal grease to aluminum-oxide base grease and now my 955BE is stable at 4ghz @ 1.425v with temps around 52C after a few hours of prime95. I'm sure I'm doing plenty of other things wrong but good grease and prepping the heatsink base made a significant difference for me.

 

[Elixer]

Which heatsink was this ?

 

[alkemyst]

Also if you aren't hand sanding with glass you are probably causing more problems than solving. I am finding more and more premium heatsinks are pretty decent right off the bat.

 

[Andle Riddum]

As said above, the rule is to make it flat not shiny

 

[KingFatty]

As said above, the rule is to make it flat not shiny

Perfectly flat would be shiny. If the surface is roughened, technically it would not be flat?

 

[synapse46]

It is a cheap rosewill heatsink, I used 600grit paper on a large marble tile for maybe 20mins of sanding. The base to start with wasn't flat, the center was deep and I sanded until it began making contact with the paper. I would expect a quality heatsink to be flat from the factory.

 

[Andle Riddum]

Perfectly flat would be shiny. If the surface is roughened, technically it would not be flat?

Not true, a rough surface means it's finishing grit size is not smooth enough to reflect the light. All it takes to make a surface shiny is a high grit sandpaper, whether the surface is flat or not.

@synapse46

Respected brands like Thermalright are notorious for having a convex bases on their HSF. Reason being that Intel procs are concave in nature.

 

[synapse46]

I hand ground the bottom of the heatsink flat, switched from white thermal grease to aluminum-oxide base grease and now my 955BE is stable at 4ghz @ 1.425v with temps around 52C after a few hours of prime95. I'm sure I'm doing plenty of other things wrong but good grease and prepping the heatsink base made a significant difference for me.

For whatever reason the system wasn't stable in dragon age for more than 15mins, ended up falling back to 3825mhz(225x17). Maybe if I increased voltage past 1.425 it would work, but I don't think I have the cooling for that.

 

[PrincessFrosty]

Difference this kind of thing makes is pretty trivial to be honest, if the difference between doing it and not doing it causes any kind of problems then chances are your cooling is not appropriate in the first place.

 

[synapse46]

Difference this kind of thing makes is pretty trivial to be honest, if the difference between doing it and not doing it causes any kind of problems then chances are your cooling is not appropriate in the first place.

I'm using a stand-up rosewill heatsink w/ 92mm fan, so I would agree it is not great. I noticed a significant increase in temp around 4.0ghz and above 1.425v.

 

[Rvenger]

I bet you are using this heatsink

Thats actually a pretty good heatsink. If your sanding made a difference in your temps, then I guess you have accomplished something.

 

[Hacp]

So many people with high tech heatsink prep . I had to reinstall my heat sink recently but only had some cheap TIM and didn't even have rubbing alcohol . I ended up cleaning the Heatsink with 40% alcohol aftershave, soaking it in water for a day, then installing it. Haven't noticed a difference in temperature so far after a week.

 

[Revolution 11]

Yeah, getting shiny and flat confused is a common mistake. A shiny heatsink may be flat on the molecular level but could be warped or bulging in a particular zone. This can cause massive air bubbles or voids. A rough heatsink might have many microwells and pits but if the overall surface is flat, it will do a better job.

Shiny is the cherry on top, but the heatsink should be flat or in a shape that fits the CPU with no large voids. I am sure IDC had perfectly explained this in a previous post of his.

 

[Idontcare]

Shiny is function of what is called "surface roughness".

You can have shiny surfaces that are very far from being flat.

^ Very shiny, but not flat by any means. Warning: Do not use this coffee kettle as a HSF!

Flat relates to the radius of curvature at the macroscale, surface roughness relates to the radius of curvature at the microscale (relative to the size of the overall item itself).

A flat shiny HSF is not a problem per se, but neither is a flat rough HSF. In theory (the academic kind), a flat shiny HSF is superior to a flat rough HSF because of the the phonon manifold that is available for phonon-to-phonon coupling at the surface.

A rough surface results in a limited phonon manifold at the surface (all the low-frequency phonon cannot exists close to the surface) which limits the phonon coupling necessary for heat transfer. (heat exists as phonons in solid materials)

But it is not a practical concern for the devices we are using here with our CPU's.

 

[synapse46]

For whatever reason the system wasn't stable in dragon age for more than 15mins, ended up falling back to 3825mhz(225x17). Maybe if I increased voltage past 1.425 it would work, but I don't think I have the cooling for that.

After more testing and playing with voltages determined the instability issue was the RAM speed was too high. The ram errors seem to be more inconsistent than CPU erroring out, was basically the last setting I tried changing. So far looks like I maybe able to run 3.8ghz at 1.4v, for $79 from tiger this 955 has been satisfying for me.