• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Optimizing the AMD Stock Heatsink

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

futura2001

Member
Joined
Mar 5, 2002
Location
Bellevue, WA
whuh?!?!1 WHY?!? you ask?
Well, frankly I am going to be doing this as a proof of concept and because I am curious to see how efficient I can make the poor thing... then when I make it is as efficient as I can, I might use it to cool a NB or GPU and use the same ideas to optimize the best heatsinks further than they already are.
For starters I am going to lap the thing and see what kind of temps that gets me.
Fan replacement is an obvious choice as well. I have a ton of fans left over from the various systems I have had over the years so that is no problem.
After these two very obvious means of optimizing the heatsink, it gets a bit more dicey.
I am thinking of making a custom shroud (out of duct tape and cardboard if need be) to optimally rout airflow.
Then there is the issue of surface area. More importantly, how to expand it. I had to put on my thinking cap to do the math for this one, but I am thinking about punching precisely sized holes through the fins of the heatsink to enlarge the surface area and create more turbulent flow. The equation is as follows: The radius of the hole must be less than the width of the fin the hole goes through because, where the surface area being removed by the hole equals 2πR², the surface area added by the circumference of the hole multiplied by the width of the fin equals 2πR*W, where W is the width of the fin.
Either that, or I could crosscut the heatsink to convert it to a pin array, but then the width of each cut has to be less than the width of each fin...
Whatever the case, I am wondering what else I could do to the stock heatsink (aside from throwing it out) to improve its performance.
 
futura2001 said:
The radius of the hole must be less than the width of the fin the hole goes through because, where the surface area being removed by the hole equals 2πR², the surface area added by the circumference of the hole multiplied by the width of the fin equals 2πR*W, where W is the width of the fin.
Do the math! Radius will be thousandths!

futura2001 said:
I could crosscut the heatsink to convert it to a pin array,
Wrong...HS's efficiency is based on contact area. Less contact area, less heat transfer. Lapping will help 1-3 degrees C. Shine is less inportant than flat. Ducting (cold air from outside the case) to the fan will help, You will get the greatest improvement from higher airflows. Airflow is only limited by how nuch noise you can stand.
 
billb said:
Do the math! Radius will be thousandths!
Only if you try to multiply two fractions together...
Each fin on the heatsink will be somewhere around 1mm wide, give or take, so as long as the hole is less than 2mm in diameter there will be an increase in total surface area.

Wrong...HS's efficiency is based on contact area. Less contact area, less heat transfer. Lapping will help 1-3 degrees C. Shine is less inportant than flat. Ducting (cold air from outside the case) to the fan will help, You will get the greatest improvement from higher airflows. Airflow is only limited by how nuch noise you can stand.
No, the total cooling power of a heatsink is (in a nutshell) a combination of the contact area, the thermal coefficient of the material it is made from, the surface area of the heatsink, and the temperature differential between what needs to be cooled and the ambient temperature. Airflow and surface area are codependent, the more you have of either or both, the lower your temps will be, and for the same reason: both increase the amount of heat that can be dissipated by the cooling medium (the air).
So yeah, I could just strap two massive fans on the sink and call it good, but I really wouldn't learn anything. Instead, I have no qualms with cutting up the crappy stock heatsink and I would like to see if there is anything I can do to the heatsink
 
Fan replacement is an obvious choice as well. I have a ton of fans left over from the various systems I have had over the years so that is no problem.
After these two very obvious means of optimizing the heatsink, it gets a bit more dicey.
I am thinking of making a custom shroud (out of duct tape and cardboard if need be) to optimally rout airflow.

1 fan on top blowing down into HS.
1 fan blowing in to the side, attached to a shroud on the other side.
- to a fan at the back of the case pulling hot air out through the shroud, from the HS.

:D
Been there, done that. Good for up to 10C if your lucky. ;)

:cool:

Read both pages.
 
I was going to duct the thing, but my focus is on the heatsink itself and if I am going to be modifying the thing repeatedly, a complex duct system would only make things more difficult. Not to mention the fact that this is only going to be a temporary solution and I will most likely just set the motherboard on top of its cardboard box and test from there. Maybe I will install everything into the case and attach a duct for the very last test to see what kind of a drop I get with it on...
My 148 is supposed to arrive today, so I will have to calculate the total surface area of the stock heatsink and decide what exactly I am going to be able to do to the thing.
 
futura2001 said:
a complex duct system would only make things more difficult..


youd probably see better results if you just cut a hole in the side of your tower and put a fan on it and a modified pringles can directing air from outisde the tower onto the cpus hsf.

its alot less complex than calculating the surface area of the hsf or drilling alot of holes into the sink and alot less time consuming than lapping the stock hsf.

just a couple thoughts dont get me wrong id love to see the results of your project here but i think youre going to see the biggest gains temp wise from ducting in some outside ambient air instead of all the mods to the hsf you had planned which would still have to deal with the case temps of the mosfets/caps chipset, hdd, memory and videocard all dumping extra heat into the tower.

now if you did your planned mods to the stock hsf coupled with a duct that could only be more effective id think.
 
You can still duct it even on a cardboard box.

You partially missed my point.

You can increase the air flow ( velocity ) over the cpu & get better cooling.

(kinda like more fans, but not, if you work it right.)
 
Since the motherboard, heatsink and all will already be sitting in ambient air, that won't be a problem. So if I want to increase the air velocity over the CPU, a properly sized venturi tube would be the best solution. Not only do I get more velocity, but I also get increased air pressure as well. In the somewhat unlikely event that I will be able to find out the best air pressure/airflow rate for the fan I am using, it is a trivial matter to calculate the right sized venturi tube for my applications. Coupled with a custom shroud to rout air out of the heatsink, this could dramatically improve temps.
In response to a debate I was having, for optimal cooling efficiency, would laminar or turbulent airflow be the best for air cooling?
I would think turbulent airflow would be better because air has a ridiculously low c/w and the low velocity boundary layer produced by laminar flow would cripple cooling efficiency, where turbulent flow would constantly wax and wane the size of the boundary layer at any given point on the surface.
 
I too LOVE experiments like these...i dont know what it is but theres just something thats invigorating about taking something crappy and using math to make it better...math is a beautiful thing
 
futura2001 said:
>>>>>>>>>
No, the total cooling power of a heatsink is (in a nutshell) a combination of the contact area, the thermal coefficient of the material it is made from, the surface area of the heatsink, and the temperature differential between what needs to be cooled and the ambient temperature. Airflow and surface area are codependent, the more you have of either or both, the lower your temps will be, and for the same reason: both increase the amount of heat that can be dissipated by the cooling medium (the air).
So yeah, I could just strap two massive fans on the sink and call it good, but I really wouldn't learn anything. Instead, I have no qualms with cutting up the crappy stock heatsink and I would like to see if there is anything I can do to the heatsink
Well said. There is much more to heatsinks than most realize.

As some of you know custom heatsink design is an interest of mine.
To that end I have spent a fair amount of time studying industrial & academic information sources.

We can think of a heatsink as a "thermal network". Put simply, the heat
should flow through this "network" from the heatsource to the surrounding environment.
Resistance at any key point below will result in a reduction in thermal efficiency of the overall design.
As you'd imagine some have more effect on efficiency than others.

Here is a diagram I put together which displays each important consideration :

Susquehannock-sink2.jpg


A] = Contact resistance.
B] = Spreading resistance
C] = Material resistance
D] = Fin resistance
E] = Film Resistance

Since some mentioned cross-cutting the fins on a stock AMD sink, I would like
to mention that "D" can be greatly reduced with the advent of rounded base
fin/pins since a similar fin/pin with a square base has less contact with the base block.
Makes sense when you think about it.

Happy modding! :)
 
I see what you are saying, I am leaning away from crosscutting the fins, however. To get a sizable gain in surface area, one would have to make very thin cuts, and the more narrow the cuts get, the greater the impact of the low velocity boundary layer. No matter what I do, rounding any hard edges will be a good idea. Thanks, I hadn't thought of that.
Right now I am leaning towards drilling indentations vertically down each fin with a drill press, and drilling holes through the fins themselves to add a small amount of surface area and create some crossflow.
In addition, I might put the heatsink in a rock tumbler or sandblast it to rough the metal up and create an uneven surface (prior to lapping, of course). But, the uneven surface might increase the boundary layer and create tiny dead spots, which would nullify any gains in surface area or turbulence.
 
With the fins only being ~1 mm thick, I'm not sure how you expect to drill tiny holes through them without deforming the aluminum in the areas surrounding the holes. With the fins being so closely spaced, I would think that even a small deformation of the the metal would create some "dead spots" that you speak of. In theory, what you propose makes sense. I just have a feeling that you won't be able to execute these ideas well enough due to limitations in tooling precision. Also, how do you plan to achieve turbulent flow? According to my calculations, you will need an insane flow rate (~200 CFM) to achieve a Reynold's Number even close to 2000.
 
I'm going to immerse the heatsink in melted parafin wax, let it harden, and then do my cutting/drilling. The wax should be enough to keep the metal from deforming, and when I am done I will just have to drop the heatsink in boiling water to melt away all the wax.
As for achieving turbulent flow, my current running model involves running two fans at either side of the heatsink, blowing in at each other. One would have cool, ambient air ducted directly to the fan, and the other one would be pushing warmer air from inside the case. Air would exit out the top of the heatsink through a duct going out the side panel of the case and also through the holes drilled in the sides of the fins. Gonna have to run a few simulations to see if it will work.
If it doesn't, well, a 192mm comair rotron and a properly sized venturi tube will work wonders. Not that I would actually use a 192mm comair rotron in the real world and risk hearing loss, this is just a test.
 
Re: parafin wax .......

Good idea - was going to suggest that but saw your post first.
I was also going to suggest an end mill cutter instead of a drill bit for making
the vertical indents but not sure if that is applicable in your case.

Sounds like a nice project. Please keep us posted how it goes.
 
dude, i don't care if others say this is pointless, or you will do more harm than good.

this is awesome. i love when people do little projects like this!

you may find a way to help us get our heatsinks to be more efficient. and since a majorit of us are still using air cooling, this could be a help to us.

and even if we don't want to cut up our heatsinks and go through the trouble of all this, its AWESOME to see someone else doing it, just to see what happens, and see how well he can polish a turd. :D

also, you going to use AS5? just wondering.
 
I am actually trying to find someone who would be willing machine the heatsink to specification with a CNC mill. Not having much luck with that one, but if all else fails it will probably be carbide bits on the drill press. If anyone knows who to talk to or where I could get it machined in the greater Seattle area, I would appreciate it.
Pretty busy with work right now, but I will keep updating this thread when I have new information.
I will be using AS3 for the majority of testing because a) I have a nearly full tube of it, and b) I don't want to waste the AS5 I have. The current method of testing will be using the white goop provided with the unmodified sink, then seeing how much of a difference AS3 makes, then when the heatsink is as optimized as possible I will see how big a difference AS5 makes over AS3.
Making heatsinks more efficient is what this is all about, depending on the results, I may follow this up with modifying one of the best heatsinks available and seeing if I can make a noticeable improvement... but I will worry about that when I get there.
 
A month later and nothing new to report.
Mainly because ZZF sent me a used motherboard (by used, I mean the box had been opened and the motherboard in the box didn't match the serial number) and they have been exceedingly slow in processing my RMA.
No motherboard means no initial testing with a completely unmodified HSF, and so I really haven't been able to do much other than watch the unopened box with my Opty 148 inside collect dust.
I haven't forgotten about this project, and when I get the replacement mobo, I will get to work on it...
 
Back