Hardware Analysis Community Forums http://www.hardwareanalysis.com/content/topic/68424/ 2008-08-20T07:06:45-05:00 http://www.hardwareanalysis.com/content/topic/68424/ http://media.hardwareanalysis.com/halogo.gif text/html 2007-08-28T04:16:39-05:00 G. G. http://www.hardwareanalysis.com/content/topic/68424/?l=1#507327 I would do the hsf side myself.... I'd be too chicken to do it on the cpu side fearing damage to the whole package...... text/html 2007-08-28T04:01:27-05:00 McFly http://www.hardwareanalysis.com/content/topic/68424/?l=1#507325 I only lap the heatsink itself ... only takes 5-10 minutes and I usually see a 5°C drop, so it's worth it to me. <img src="http://media.hardwareanalysis.com/smilies/smile8.gif" width="14" height="14" border="0" alt=":_)" title=":_)"> text/html 2007-08-28T03:54:10-05:00 Sander Sassen http://www.hardwareanalysis.com/content/topic/68424/?l=1#507324 Lapping? What do you think thermal paste is for? Use a very (very!) thin layer to fill out any surface texture on both the heatsink base and the processor die to ensure maximum contact. DIY lapping always results in people making a nice curve onto the processor die, a CNC mill would work for the aluminum or copper heatsink base, wouldn't use it on a processor die though, way too fragile and prone to chip.<br /> <br /> Best regards, text/html 2007-08-28T03:42:59-05:00 FordGT90Concept http://www.hardwareanalysis.com/content/topic/68424/?l=1#507323 P V got what I was saying. Here's a crude pic:<br /> <a class="ext" href="/action/r/http://img513.imageshack.us/img513/5692/thermalpinscj9.png" target="_blank">http://img513.imageshack.us/img513/5692/thermalpinscj9.png</a><br /> <br /> I didn't finish it because Overlord is running minimized. It's too taxing on my poor 4000+. Anywho, that surface in the picture would be either built into the processor cap or the heatsink. There would be a fraction of a millimeter difference between the two to make way for thermal compound. The two sandwich together creating hundreds (if not thousands) of times more surface area cooling much better than any form of cooling can do today. Processor caps like that may start showing up as the core count climbs.<br /> <br /> Spikes would probably do better in the way of preventing air bubbles and the whole thing might have to be domed in order to for air from the inside out.<br /> <br /> <br /> Edit: Whoops, I forgot to mention that it's a two piece set. If the processor has pins sticking out, the heatsink has holes for them to stick in to. Because it would need a lot of thermal paste, I would place a ridge around the processor to catch the excess paste to prevent it from spilling into the socket. text/html 2007-08-28T02:10:35-05:00 phil http://www.hardwareanalysis.com/content/topic/68424/?l=1#507319 ford does have a point... you guys are just looking at it differently then what he's saying. <br /> <br /> he's saying that if the grooves in the cpu and HS fit together like<br /> <br /> /\_/\_/\_/\_/\_/\_/\ hsf<br /> /\_/\_/\_/\_/\_/\_/\ cpu<br /> <br /> .. if the grooves lined up, then there would be still 100% contact but much more surface area.... all be it, like 100000000000 to 1 odds to getting that with sandpaper<br /> <br /> where i think you guys would be thinking like this:<br /> /\_/\_/\_/\_/\_/\_/\/ hsf<br /> \/\_/\/\/\_/\_/\_/\/\/\ cpu<br /> .. where the grooves don't line up ... reducing contact area.. (this is what you get with sand paper)<br /> <br /> <br /> .. then theres the mirror and flat finish..<br /> <br /> hsf<br /> =============<br /> cpu<br /> text/html 2007-08-28T01:23:38-05:00 G. G. http://www.hardwareanalysis.com/content/topic/68424/?l=1#507316 Ford,<br /> <br /> the only place you want grooves are on the fin side of the heatsink.... more surface area for air to bleed the heat off the heat sink... as for the contact point between the cpu heat spreader and the contact point to the heatsink ... you want no grooves... you want as much contact points as possible. The bigger the contact patch the more area where the heat can transfer into the heatsink to be bled off on the other side. The reason you need thermal paste is to fill in those grooves between the heatsink and the cpu to make contact..... otherwise you have airgaps and those airgaps are a resistance to heat transfer to the heatsink. when you lap, you get rid of those grooves and to make sure the surface is perfectly flat to which both create a bigger contact patch between heatsink and cpu.... <br /> <br /> <br /> Flat and smooth (no grooves) between heatsink and cpu is the way to go... the flatter and smoother the better. That is why there is such a big improvement in the before and after lapping project of that article is all do to making both surfaces as flat and as smooth as possible. text/html 2007-08-28T00:14:18-05:00 Tam the Bam http://www.hardwareanalysis.com/content/topic/68424/?l=1#507310 Lol PV i wasn't on about tolerance, i was on about finish and in mm's! lol!<br /> <br /> 16/<br /> \/ finish<br /> text/html 2007-08-28T00:05:13-05:00 Tam the Bam http://www.hardwareanalysis.com/content/topic/68424/?l=1#507309 You're very clever aren't you ford. I give you a degree in Grooves.<br /> <br /> The rougher the finish the bigger the grain. The smoother the finish the smaller the grain.<br /> Not grooves! Less grain = more contact. Big grain = say good bye to ur CPU!<br /> <br /> In a nutshell, shiny mirror finish!!<br /> text/html 2007-08-27T23:49:00-05:00 phil http://www.hardwareanalysis.com/content/topic/68424/?l=1#507305 the problem with lapping at home... well a lot of people just end up sanding the same curve or an even worse curve into the damn chip... and the occasional person gets filings into the underside of the chip<br /> <br /> yes intel and amd should put more care into making them flat, but don't just grab some sand paper and go at 'er.<br /> <br /> ----------------------------------------<br /> tam, leaving your proc with 2thou/inch tolerence isn't worth the effort, just use a 0.01mm microcaliper, machinist level, some cutting oil and a round ceramic wheel stone.<br /> <br /> level the wheel stone flat (with level to 10000th/inch) confirm it by rotating it through the microcaliper. map the high/low on the chip with the microcaliper for reference, level the chip against ceramic on the highest point , then apply the oil in the path of contact on the stone as it rotates... 20mins later. you have a chip thats level to .0001/in, instead of .002/in text/html 2007-08-27T23:05:22-05:00 FordGT90Concept http://www.hardwareanalysis.com/content/topic/68424/?l=1#507300 Grooves help cool. The smaller the grooves, the more surface area, the cooler it runs. When you can't even feel the grooves, the grooves are microscopic: ideal. You don't want a perfectly smooth surface or it won't cool well at all. I'm still waiting for an interlocking processor cap and heatsink so that heat is dispersed deeper into the heatsink.<br /> <br /> <br /> I've only lapped one heatsink before because the underside was so course, rubbing it would tear the skin. text/html 2007-08-27T22:26:04-05:00 Sean Costello http://www.hardwareanalysis.com/content/topic/68424/?l=1#507296 ive known about lapping, just never done it. Mainly cause i dont have any sand paper <img src="http://media.hardwareanalysis.com/smilies/smile1.gif" width="14" height="14" border="0" alt=":)" title=":)"> text/html 2007-08-27T22:07:36-05:00 Tam the Bam http://www.hardwareanalysis.com/content/topic/68424/?l=1#507295 <br /> WHO would like to hire my CNC Mill to get a 0.002 finish on their CPU's??<br /> That's so shiny you can see ur ugly mug init!<br /> <br /> text/html 2007-08-27T22:05:49-05:00 Tam the Bam http://www.hardwareanalysis.com/content/topic/68424/?l=1#507294 <br /> They should be true flat anyways. AMD/Intel.... TAKE HEEED!!!<br /> text/html 2007-08-27T20:16:11-05:00 micro http://www.hardwareanalysis.com/content/topic/68424/?l=1#507285 Thanks gg for the post/link.<br /> It may be fun to do sometime. I wonder what % of the drop in temp is from removeing the nickle down to the copper. Though from what i see im sure most of the results come from the flat surface.<br /> <br /> The only variable i can see besides what was mentiones, is if he had a slightly thicker or thinner layer of thermal past on between tests.<br /> <br /> I had always assumed that the surfaces were flatter from the fatory, but i guess if they had to resurface them there, they would be chargeing more per cpu. text/html 2007-08-27T20:10:47-05:00 McFly http://www.hardwareanalysis.com/content/topic/68424/?l=1#507284 Neither is getting out of bed in the morning. ^_^