Hows yer heatsink doin` these days?
A year or so ago, most forums were well into lapping the heatsinks to get the best fit possible for the CPU.
I have read very little about heatsink lapping or even checking them for flatness recently.
Because I have proven to myself that flatness and a well seated heatsink is more than idle chit-chat, I would like to pass along (again) some good pointers.
To begin with, all aluminum heatsinks are coated with a finish called anodizing.
Industrial engineers who work in the area of industrial electronics, sometimes huge electrical DC power supplies and DC welders. These engineers know that placing a solid state device on some aluminum heatsink without first cleaning off the anodized finish will cost them in heating and loss of solid state devices. Their "stuff" runs considerably hotter than the average CPU..but, we want our CPUs to run as cool as possible.
an·o·dize (²n“…-dºz”) tr.v. an·o·dized, an·o·diz·ing, an·o·diz·es. To coat (a metallic surface) electrolytically with a protective or decorative oxide. --an”o·di·za“tion (-d¹-z³“sh…n) n.
So, we should always check the bottom of any heatsink with a good straightedge (like a metal ruler etc.) and place the heatsink with the ruler across the bottom between you and a desk lamp, etc. If you see more than a tiny crack of light, the bottom needs to be sanded (or lapped).
Copper heatsinks do not get off clean either..many are shipped out of flat..concave or rounded. Those that are being shipped with the thermal pad seem to be exceptionally bad sometimes...I have not had the opportunity to check many, but most all of those I have checked were in need of lapping..including a high priced Swiftech bolt down type.
I recently purchased a T-bird and a heatsink...when it came, I wanted to check it out naturally, so I installed the CPU, then looking at the heatsink I saw the remains of a thermal pad. I cleaned all of the bottom good, and found that the bottom of the heatsink was very course..looks like it had been finished with maybe 100 grit sandpaper.
I put some artic silver on the CPU and slapped the heatsink on anyway..ran the system a couple of hours..then checked the CPU temps.. I was surprised to find it was near 56 degrees C.
I removed the heatsink, cleaned off the AS and lapped it down flat (it was also rounded on the bottom).
Once I got the bottom flat and smooth, I reattached the sink...after about two hours, the CPU temp was 48 degrees C.
This is a copper bottom CPU, and the proof is here, but you will have to take my word for it because I forgot to make any screen shots for the before and after..not even figuring I would see that much difference.
If you are having any kind of over heating of the CPU, more than what you think it should be...next time you remove the heatsink, check it for flat..and sand that anodize off..if it is an all aluminum sink.
The average CPU temps are about 12 to 18 degrees hotter than the ambient case temps..if you have a CPU running 20 degrees centigrade hotter than your case air, you have a problem. Unless, the case temps are simply too hot. The heatsink fan can only do so much..and when it is pushing hot case air over the heatsink fins, its loosing the ability to cool it down. Case temps must be kept as cool as possible.
I think people don't talk about it as much as it was well covered for quite awhile and poorly finished heatsinks were being called out for it as some companies had started to produce high quality bases. Now most produce a base of at least decent quality that lapping isn't going to gain you much. Stock heatsinks though, they are still garbage and probably will continue to be until they NEED to ship a CPU with a quality sink.
I agree 100%...but by agreeing, I also realize that it has been at least a year since I have read anything concerning lapping heatsinks...and how many new comers have we had at SysOpt in the span of one year...that may not know doodly about heatsinks in general.
You are not 'right on' (if I am understanding you correctly) concerning the quality of finish received on the better heatsinks.
At the time I bought my Swiftech copper/aluminum combo heatsink for about $50.00 W/O any fan, I figured it was possibly the best on the market..certainly the most costly at the time. The heatsink DID have a mirror finished bottom...but!!... it was not flat. It was concave...and when I put a metal machinests rule across the botton, I could get a .001 feeler guage to slide underneith. I don't believe Artic Silver will take up that much slack.
We buy and use the higher priced Artic Silver (and now, other brands with a metal content in the thermal paste) to assist making a good...solid metal to core contact to the CPU. If the bottom of any heatsink is concave or convex (rounded) surface, there is no way to get that heatsink to sit flat down on the core..if that happens, the temps will rise.
The heatsink that I just sanded, and prompted this thread was...simply awful. Whoever manufactured it surmised that the thermal pad would conceal the rough bottom surface and it would work..it didn't!...not until I had lapped it flat and smooth...and also, (I have not discovered the make of this heatsink yet) the sink looks great..well made and heavy duty..it just was not finished out good enough.
You're not complaining because you could get a .001 feeler in there are you? You know that a human hair is between .003 and .005 thick? Come on, what more do you want? Where I work at we do a lot of precision maching for government subcontracts and what have you. Our tollerances are usually around .003. If three thousandths of an inch is good enough for the military then I'm sure its good enough for a heatsink.
I don't complain. I teach perfection anytime I can. I see many threads here at these forums looking for assistance when if the user had applied just a little more interest in getting his/her system setup and running good, they would have used more consideration in the project.
Originally posted by Johnny Fist
You're not complaining because you could get a .001 feeler in there are you?
If you happen to be one of those that just throws stuff together and hope it works, have no desire to learn the correct ways to checkout and setup a computer so that it will perform at its best..then, you don't need to be critical of me..or anyone else who has taken the time to learn how to make it work to its maximum perfection.
Besides, I seriously doubt the powdered granules of silver in the A/S paste is thicker than .001 inch, and makes good contact between a heatsink and CPU core.
Another thing, you are comparing apples and oranges when describing the differences between a hair and a feeler gauge between a straight edge and a miss-shaped heatsink.
Given that you're attempting to achieve perfection I'm sure you went about the proper procedure of grinding the heatsinks' base to within a tenth, correct? Afterwards did you begin a stoning process starting with a 400 fine grit and work your way up to a number 1000 stone before attempting to begin a lapping process with a number 9 micron industrial grade diamond until you had worked the surface to a number one micron optical finish? That, my dear friend is perfection. Thats the sort of work I do for a living.
I can understand the part about perfection, but there really is a point where you're over doing it. The guides you see on the internet showing how you use a little bit of sandpaper and water are trivial at best. But, for what their doing they're more than enough. One thousandth of an inch is not cause for concern.
You have made your point..now, lets just drop it.
I don't care for your attitude and I try to stay clear of any threads you happen to be in..so, before we go any further, lets agree to just leave one another be.
Fair enough, Bovon. I wasn't trying to start trouble with you.
Personalities aside... I lap all my sinks and my north bridges too. I agree extreme perfection in this aplication is not, probably, going to buy much of an increasee in performance. (I would like to see some tests with truely precise lapping and no or super thin TIM (possibly assemble the parts then draw a hard vacuum on them to minimise the air film, or assemble them in a vacuum then let the air pressure help hold the assembly together).
With the sandpaper and glass technique I start to see decreasing returns ( in terms of temperature reduction vs time spent at about 800 or 1000 grit. (Many times I am out for the best bang for the buck not the last 0.0001%)
While 0.001" is relatively small, I would have to relap if I saw a defect (under magnification) that big. With minimal care sandpaper and glass should get within a couple of tenthousandths. (Assuming that the glass is supported by somthing relativly stiff/flat and that excessive down force is not applied). After all, telescope mirrors and lenses have been made this way for centuries.
Hi Happy Joe, thanks for the comments.
I rarely go to any extreme when lapping (or anything else probably) but I do learn why, and to understand the reasons for doing anything that will improve its function...no matter how slight.
I quite often stop with 400 grit or maybe 600...if I happen to have some around..this last time, I had to go from 320 to 1000 because that was what I had...I did need to begin with 240 grit because that HS was really bad on the bottom. I wound up with a mirror finish by using the 1000, but only because I went directly from 320 to 1000..it took a lot of rubbing to get rid of the 320 scratches.
A heatsink does not need a mirror finish..IMO..that is taking it to extremes.. I think a smooth finish (and flat) from just finishing up with 400 grit is fine. Several years ago when lapping became a fad of sorts, the guys were bragging about how smooth they could get their heatsinks..using up to 2000 (and higher) grits...actually, thats getting into the jewelers rouge metal polishing compounds.
Although I have lapped a few cheap heatsinks on just this desktop on some newspapers, I also use the plate glass foundation when I am doing a really good heatsink, and I am trying for a decent finish.
I myself got a piece of glass, sand paper, rubbing compound,and polish, then when to work on my copper heat sink.
Improvise - Adapt - Overcome
SafeSearch is off
Wow! 320 to 1000 in one step... I'll bet that took some rubbin'!!On really bad sinks I start with 200, to make them flat, the go up in about 200 grit steps (or whatever I've got thats close). Learned my techniques many years ago preparing specimins for microanalysis. I have jeweler's rouge but found that it increased my temps (I figure that it left some deposits).
Badog; If you have any questions just ask..expect a couple of degrees C improvement for a small investment in time and money.
Last edited by Happy Joe; 08-18-2004 at 05:37 PM.
I did mine last winter when the power was out.
Improvise - Adapt - Overcome
SafeSearch is off
Ahhh sooo, thats how you stayed warm all winter, huh!
Originally posted by Baddog
I did mine last winter when the power was out.
Yes, finishes can still be concave, convex, or just plain garbage all the way. But a company can easily pass off a well finished piece with a .001 dip than the toss finish it probably would have had a year or so ago. I agree that a HS should be checked for flatness, not just apparent finish, but how far you go, that is another question.
I can't say whether or not you can improve at a reasonable cost (I checked my swiftech GPU block, it's dead on flat, the SLK isn't on the otherhand). Would lapping to that same quality to get that .001 dip out actually net you more than a couple C if that? Like I said, I can't say as I don't have a nicely finished concave heatsink, nor do I have the will to maul it by trying to refinish it.
I'll let you guys decide.
New Security Features Planned for Firefox 4
Another Laptop Theft Exposes 21K Patients' Data
Oracle Hits to Road to Pitch Data Center Plans
Microsoft Preps Array of Windows Patches
Microsoft Nears IE9 Beta With Final Preview
Simplified Analytics Improve CRM, BI Tools
Android Passes RIM as Top Mobile OS in 2Q
VMware Updates Hyperic System Management
File Monitoring Key to Enterprise Security
LinkedIn Snaps Up SaaS Player mSpoke