Sign in with
Sign up | Sign in
Your question

Heatsink mod idea: cutting grooves in the base

Last response: in Overclocking
Share
June 17, 2013 10:21:51 PM

If this idea is any good, I'd be surprised if people aren't already doing it. I'm sure I'm not the first to think of it, but I haven't run across it while reading about lapping before I tried it myself.

Here's the story: I've lapped my last two heatsinks, and noticed a bigger benefit from the first. I thought that was just from it having a rougher finish, but it occurred to me that I was able to feel plenty of metal-to-metal contact, when mashing it down. It's a heatpipe-direct-touch design, whereas my second heatsink has a solid base. I did not feel the same kind of sensation, with the second heat sink, even though I used the same TIM compound and plenty of rubbing, twisting, and pressure.

So, it occurred to me that perhaps the grooves between the heat pipes provided somewhere for the excess TIM to go, as I worked the heatsink into place. Whereas the unfinished base of the second heatsink (pre-lapping) may have partially fulfilled this function, the smoother & flatter base (post-lapping) essentially trapped it in place.

Now, I'm thinking of cutting a few grooves into the base of my second heatsink, to replicate the effect. I think it should be well worth giving up 5-10% of the contact area, in orer to achieve significantly more metal-to-metal contact. The groves should also help any trapped air escape.

My current plan is to use the corner of a metal file to cut about 3 grooves across the base, slightly off-center. About 1mm in depth. Then, re-lap to ensure the edges of the grooves are completly flush.

What do you think? Has anyone tried this? Is there a better way? Have I just reinvented the wheel?
June 17, 2013 11:06:53 PM

If you are doing this to have a 'play around', go for it. I think with these things, my main thoughts are...

-Is your theory, which isn't based on any science, better than the efforts of teams of engineers with qualifications in thermal dynamics?

-Sure, it could be a case of "this would provide an improvement, but it would cost significantly more to manufacture, and won't provide any user-measurable results"

Either way, I can't see it making a difference that will mean anything. Maybe worth a go if you were insanely OCing a Pentium D which could burst into flames at any moment, but even a stock cooler is more than enough for todays chips, even on a solid OC.

If it makes a difference, it will be so small that any variability would throw off your results. Control temperature, pressure, humidity, TIM type and amount, clamping force etc etc. Average the numbers over at least 2 hours of testing (once ~2 hours have passed to allow temps to stabilise), then perform the mod and repeat.
m
0
l
June 18, 2013 5:23:27 AM

bit_user said:
If this idea is any good, I'd be surprised if people aren't already doing it. I'm sure I'm not the first to think of it, but I haven't run across it while reading about lapping before I tried it myself.

Here's the story: I've lapped my last two heatsinks, and noticed a bigger benefit from the first. I thought that was just from it having a rougher finish, but it occurred to me that I was able to feel plenty of metal-to-metal contact, when mashing it down. It's a heatpipe-direct-touch design, whereas my second heatsink has a solid base. I did not feel the same kind of sensation, with the second heat sink, even though I used the same TIM compound and plenty of rubbing, twisting, and pressure.

So, it occurred to me that perhaps the grooves between the heat pipes provided somewhere for the excess TIM to go, as I worked the heatsink into place. Whereas the unfinished base of the second heatsink (pre-lapping) may have partially fulfilled this function, the smoother & flatter base (post-lapping) essentially trapped it in place.

Now, I'm thinking of cutting a few grooves into the base of my second heatsink, to replicate the effect. I think it should be well worth giving up 5-10% of the contact area, in orer to achieve significantly more metal-to-metal contact. The groves should also help any trapped air escape.

My current plan is to use the corner of a metal file to cut about 3 grooves across the base, slightly off-center. About 1mm in depth. Then, re-lap to ensure the edges of the grooves are completly flush.

What do you think? Has anyone tried this? Is there a better way? Have I just reinvented the wheel?


Lapping to a mirror finish allows the use of almost zero TIM, it also allows the use of thinner TIMs like Artic Coolings MX-4, that will under pressure spread to almost nothing.

As the finer lapped finish you have the less TIM is required, the purpose of TIM is only to fill the microscopic imperfections between the two mating surfaces, you're never after applying a solid layer of TIM, too much TIM acts as an Insulator instead of conducting the heat transfer.

The difference in the feel of the 2 heat sink bases may not have a thing to do with the actual cooling.

It may simply be that you've mounted the HDT heat sink in alignment with the CPUs core under the heat spreader, aligning the cooling pipes with the cores alignment itself.

m
0
l
June 18, 2013 8:32:12 PM

First, thanks for the reply. I (mostly) agree with your point of view.

mrmez said:
-Is your theory, which isn't based on any science
*ahem*, I think it's a well-established (scientifically, even) fact that the best TIM still has thermal conductivity a couple orders of magnitude worse than copper. So, my guiding principle is to minimize the quantity of TIM, without risking airbubbles.

mrmez said:
teams of engineers with qualifications in thermal dynamics?

-Sure, it could be a case of "this would provide an improvement, but it would cost significantly more to manufacture, and won't provide any user-measurable results"
I'm sympathetic to this line of reasoning, but here's the catch. My thinking is that the proposed mod is only significantly beneficial when the heatsink base has been lapped and the CPU heat spreader is sufficiently flat. Furthermore, to do any good, it requires just the right amount of TIM to be applied and then for the heatsink to be worked into place.

Now, as we all know, heatsink manufacturers don't lap for a few reasons. First, most err on the cautious side by making heatsink bases slightly convex. Secondly, any decent lapping job costs money and there are usually cheaper ways for them to deliver equivalent gains.


mrmez said:
even a stock cooler is more than enough for todays chips, even on a solid OC.
Well, I have a 130 W TDP E5-series Xeon, which did not include a stock cooler. The cooler I'm using provides sufficient cooling to keep me out of thermal clock throttling territory, at stock speeds, but I'm still rather dissatisfied with my temps under load.


mrmez said:
If it makes a difference, it will be so small that any variability would throw off your results. Control temperature, pressure, humidity, TIM type and amount, clamping force etc etc. Average the numbers over at least 2 hours of testing (once ~2 hours have passed to allow temps to stabilise), then perform the mod and repeat.
Okay, I guess it's clear where you stand. I agree that there's no way I'll have enough data or of sufficient quality to form any firm conclusions.
m
0
l
June 18, 2013 11:19:58 PM

If you do need to drop a few degrees, why not try a cheap option like getting more air through the sink?
I've done various things over the years, but probably the cheapest and most successful was creating a clean air Venturi through the sink. The only thing that dropped more degrees was water, and then I still used the Venturi idea to cool a 120mm radiator.

Take a front panel off ~CPu height, and use cardboard to create a 'wind tunnel' around the sink. A large opening, small around the sink, then large exit with a 140mm fan etc sucking the air out, preferably as far behend the sink as you can.
Sucking always works better than blowing ;) 

The 3 important elements here are:

-'Clean' air in. Not hot air recirculating around the case.
-Venturi creating high speed, turbulent air.
-Rear 'sucking' fan will provide more flow and pressure, a decent gap will eliminate dead spots around the sink.

Difference case/sink designs require different airflow paths, but you get the idea.
m
0
l
June 19, 2013 10:00:31 PM

Thanks for the reply.

4Ryan6 said:
Lapping to a mirror finish allows the use of almost zero TIM, it also allows the use of thinner TIMs like Artic Coolings MX-4, that will under pressure spread to almost nothing.
In fact, I am using MX-4.

4Ryan6 said:
As the finer lapped finish you have the less TIM is required, the purpose of TIM is only to fill the microscopic imperfections between the two mating surfaces, you're never after applying a solid layer of TIM, too much TIM acts as an Insulator instead of conducting the heat transfer.

I worry about air bubbles, so the way I apply it is by placing a dab in the center and then working the heatsink into place. I've never been a fan of the razor blade method, because one big airbubble could result in far worse performance than a little excess TIM.
m
0
l
June 20, 2013 9:43:15 PM

mrmez said:
If you do need to drop a few degrees, why not try a cheap option like getting more air through the sink?

This is actually a great reply. Specific, detailed, and not just saying what to do but why and how. Thank you, sir. Though your first reply left me with some negative feelings, I think you're a great asset to the forum!

That said, I'm not quite so desperate that I want to significantly increase the amount of noise my PC makes. I already have a 105 CFM 140 mm front fan and a 80 CFM 120 mm rear exhaust fan right behind the CPU (and then a 120 mm fan on the heatsink, itself). The front fan is at the bottom, and blows across the drives (typical mid-tower configuraion Lian Li case).

If I do decide to push beyond the gains I do (or don't) get from my mod, I will probably look at water cooling.

BTW, I actually thought that laminar airflow reduced the boundary layer effect, but all the information I've found on the subject of heatsink airflow seems to support your advice of increasing turbulence.
m
0
l
!