| Forum Home | ||||
| Press F1 | ||||
| Thread ID: 63674 | 2005-11-19 01:15:00 | Heatsink Lapping | Agent_24 (57) | Press F1 |
| Post ID | Timestamp | Content | User | ||
| 405453 | 2005-11-19 01:15:00 | I've decided to cool my computer the best way, starting with lapping the CPU heatsink - I've read articles and found that pre-made kits are available (or at least that's what I thought) Is there anywhere in NZ that sells these kits? I can't find a trace of them anywhere :confused:. Surely someone must sell them as they would seem to be a convenient way to get all the pieces you need. |
Agent_24 (57) | ||
| 405454 | 2005-11-19 01:23:00 | Who needs a kit? "Lapping the heatsink" just means "making it flat". Buy a sheet of 80 or 100 grit silicon carbide "wet and dry" paper, and find a piece of glass. (Modern "float glass" is very flat.) Rub the working surface of the heatsink on the paper (backed up by the glass). You'll see any hollows appear immediately. Just continue until the surface looks uniform. |
Graham L (2) | ||
| 405455 | 2005-11-19 01:35:00 | Why bother? On a microscopic level there will still be a massive unevenness. Thats what paste is for. |
pctek (84) | ||
| 405456 | 2005-11-19 01:52:00 | Agreed, but it will make a small (probably unmeasurable ;)) difference. Cast aluminium has a fair amount of shrinkage as it cools, so there's a relatively deep hollow in the centre of the "flat" contact surface. Grinding down the high surround means that the paste is just filling the small pits, which means is it working best. | Graham L (2) | ||
| 405457 | 2005-11-19 01:59:00 | Why bother? On a microscopic level there will still be a massive unevenness. Thats what paste is for. Because many heat sinks are far from flat, I've measured up to 0.05mm out of flat. The flatter the surface, and the finer the surface texture, then the greater the real area of contact, and the better the heat transfer. The paste is not meant to fill big hollows, only the surface texture, ideally there should be 'virtually' no paste actually separating the surfaces. There is a long thread somewhere on Press F1 describing temperature results of lapping, installing micro-thermal couples into the surface of the heatsinks etc. But before any lapping it only makes sense first of all to actually measure the flatness, as poor lapping techniques can make things worse rather than flatter. To measure flatness you need a reference flat, or a reference straight edge. Thick float glass as Graham says maybe flat enough for the job. A trace of engineers 'bearing blue' on the flat, and lightly rub the heat sink on the flat. You can then see the area of contact by the 'blue' transfered to the heatsink. Obviously a thick dollop of blue is useless, but it does wonders to make a surface look flat :) With a reference straight edge you look for 'daylight' between the edge and the heatsink. If the light is not coloured then the gap could possibly be measured with thin feeler gauge stock. In the old imperial days it came in 0.0005" increments, in metric it may be 0.01mm. A 'Stanley' knife blade with the sharp edge lightly removed on an Arkansas polishing stone makes a good poor mans straight edge. |
Terry Porritt (14) | ||
| 405458 | 2005-11-19 02:04:00 | Thats all nice and well, Though will still make no worthwhile difference, If the cpu's humming along at 30c then you go to all that effort for nothing. And the cpu is overheating then lapping wont cure it. A good enough project for a tool-maker with time on his hands........ |
Metla (12) | ||
| 405459 | 2005-11-19 02:13:00 | Well I'm running the latest version of Core-Center and the temperature got up to 57*C full load at full speed, this is with my Athlon 64 3700 Skt 754. I know my Athlon XP 2600 Barton ran at max 48-50 and idle at 34 usually the lowest (stock HSF), So I think I want to get my A64's temperature down a bit.. unless this is a normal temp for this CPU. I have CnQ running and It seems that sometimes when the muliplier changes (and sometimes when it's not) the FSB sometimes drops down as far as 99mhz and even up as far as 250mhz Is this true or is core-center just ballistic? I worried that something might happen to my harddrives because I've read that overclocking the FSB too much with SATA can cause data corruption. |
Agent_24 (57) | ||
| 405460 | 2005-11-19 02:25:00 | If you run the bus faster than your memory can work at absolutely reliably you will write faulty data to disk sometimes. If correct data on disk are read and are wrongy stored in memory, what is written out from that memory will be wrong. Manufacturers specify the speed their products will run at. They are usually coservative in their ratings, because they don't want stuff returned as faulty. If you run it faster, it's your responsibility to check that it is working correctly. |
Graham L (2) | ||
| 405461 | 2005-11-19 03:10:00 | If you run the bus faster than your memory can work at absolutely reliably you will write faulty data to disk sometimes . If correct data on disk are read and are wrongy stored in memory, what is written out from that memory will be wrong . Manufacturers specify the speed their products will run at . They are usually coservative in their ratings, because they don't want stuff returned as faulty . If you run it faster, it's your responsibility to check that it is working correctly . I don't usually overclock anything, and I don't have anything overclocked at the moment (due to that reason) I think core-center is lying, Since cpu-z tells me the FSB is stable at 200 . 0mhz even when core-center is telling me otherwise . But what about temperatures? |
Agent_24 (57) | ||
| 1 | |||||