go ahead man

Which model P4 is it? 3.2 sounds like the last of the V8 Interceptors... I mean P4s. It would be a shame to destroy a piece of history.

If you clean the CPU & Heat-sink then use a quality thermal paste I don't see how sanding it would improve heat transfer much if at all. I remember back in the day people were doing this but I don't believe there was any real benefit, a couple of degree difference would be statistical noise and make no difference to the chip itself except you made a mess and removed any markings from the top of the chip.

I wouldn't do it, try to find another fun thing to do with your computer.

TandySensation wrote:

Which model P4 is it? 3.2 sounds like the last of the V8 Interceptors... I mean P4s. It would be a shame to destroy a piece of history.

If you clean the CPU & Heat-sink then use a quality thermal paste I don't see how sanding it would improve heat transfer much if at all. I remember back in the day people were doing this but I don't believe there was any real benefit, a couple of degree difference would be statistical noise and make no difference to the chip itself except you made a mess and removed any markings from the top of the chip.

I wouldn't do it, try to find another fun thing to do with your computer.

It's a Prescott. The surface of most heat sinks and CPUs are not flat and sanding flat can lower temps by around 6-8c if you use a high enough grit. The best you will see is about 10-15c. But that's a extreme case.

TandySensation wrote:

Which model P4 is it? 3.2 sounds like the last of the V8 Interceptors... I mean P4s. It would be a shame to destroy a piece of history.

If you clean the CPU & Heat-sink then use a quality thermal paste I don't see how sanding it would improve heat transfer much if at all. I remember back in the day people were doing this but I don't believe there was any real benefit, a couple of degree difference would be statistical noise and make no difference to the chip itself except you made a mess and removed any markings from the top of the chip.

I wouldn't do it, try to find another fun thing to do with your computer.

oh its noticable

(If you just can't wait to see a lapped Preshott, click here 😀 )

they just arnt flat!
http://www.silentpcreview.com/files/images/lap/lapping.gif
here a old article on cpu lapping.
http://www.frostytech.com/articleview.cfm?art … cleid=82&page=1

It would be a shame to destroy a piece of history.

🤣
For the most part, P4 are just considered trash. There are tons of them about so no big loss.

smeezekitty wrote:

It would be a shame to destroy a piece of history.

🤣
For the most part, P4 are just considered trash. There are tons of them about so no big loss.

An dime a dozen in some places.

Back in the day when people first started considering the idea of lapping heatsinks and CPU tops, they meant well, and they had the right idea. I've always seen improvements in temperatures from "decking," as I'd rather call it, my heatsinks and CPU tops, much of the benefit is from abrading the surfaces as well.

Which brings us to why people began to pop out bad results, either not seeing any benefit or seeing worse temperatures: They were polishing the surface because in their mind, this was completely flat and has the most surface-area being cooled. This is true, but polished surfaces reflect heat rather than conduct it so well. So they were basically mirroring the heat back at the CPU to a degree.

I've done a lot of research into proper grits for finishing, even trials done at the industrial level for applications which are not computer related, and 800 was about as fine as you could go before you start negating the benefit--600 grit saw the most benefit so that's what I use.

It is funny how the first result when you google "CPU lapping" is this article. The article is great when it comes to detail and clarity of explanation, and in the end presents a meager 2.5C difference in temperatures. 😜 I bet the guy was quite disappointed with the results. Of course he tries to polish it (pun intended) with "it can be the difference between stability and crashes" (it can't, because 2.5C is within the scope of random day-to-day variations).

Overall - if you are a newbie and want to get introduced to the of CPU lapping and the complete uselessness of it, all at the same time - that's the article for you. 🤣

Perhaps it is because he took it to 1200 grit, which as you say, may be too much and actually negates some of the benefit. All in all, I'd be curious to see cases where it actually lowered temperatures at load by the aforementioned 6-8C, not to mention 10C+. I'd be wondering if it was just a case of an extraordinary poor heatsink / bad CPU sample, and how the final temperatures may compare to what is known to be obtainable from a good heatsink, without any of this mess. 😀

The whole point of lapping is to ensure that processor's IHS top surface and heatsink's under surface are both flat and making good contact across all the common area.

Mileage will vary. If IHS & heatsink are already flat, benefits will be minimal. On the other hand if they are far from flat benefits can be huge.

Don't forget the porous surface of the nickel plating too.
I myself have lapped at least a dozen cpu and heat sinks and would like to add this.
Some cpu's are concave and other on convex. Cpu’s with a concave ISH yield bigger drops then those with a convex as the die is under the center of the ISH. Older and cheaper cpu's also yield bigger drops then newer and higher end cpu’s. Another thing you need to do is lap both the cpu and its heat sink, you want them to both to have the same falt surface structure. Also 800~1200 grit is where I start to notice a drop in temps. 800~1200 grit on a cpu alone will drop temps about 2~3c, both the cpu and heat sink will do more around 4~5c. You start to see bigger drops around 1500~2000 grit and anything over 2000 grit is useless. But that all said, there is no guarantee you will see a drop in temps. Lapping takes a lot of time and skill to do it right and even I make mistakes at times.

here is a good link on a newer cpu/heatsink lapping with a 9c drop
http://forums.anandtech.com/showthread.php?t=2300800

EDIT:
There is also a lot of different methods you can use to lap a cpu and some yield better results then others. I myself soak the sandpaper and sand 100 strokes in one direction then rotate 90 degrees and sand another 100 strokes and jump up 200 grit at a time. I tend to stop at 1500 grit or 800 grit

dr_st wrote:

All in all, I'd be curious to see cases where it actually lowered temperatures at load by the aforementioned 6-8C, not to mention 10C+. I'd be wondering if it was just a case of an extraordinary poor heatsink / bad CPU sample, and how the final temperatures may compare to what is known to be obtainable from a good heatsink, without any of this mess. 😀

Drops over 10c are usually with an older cpu/heatink and cases where the cpu/heatsink are very uneven or damaged. You will see bigger drops with painted and stock/cheap heatinks as well. The heatsink I plan to lap is a older EK water block. This thread sparked enough interest that I think I'll go ahead and do it.

If you do go for make sure you do some good before and after testing, can you also take detailed pictures of the CPU and heatsink surface before and after? The image of the flat edge ruler along to the top of the heat spreader is interesting, maybe one like that too. Could you also clean the cpu and use good paste before you do the initial 'before lapping' testing. Sometimes I've wondered if the big boost from lapping was that the cheap/dried out/no thermal compound was the issue and after they lapped they reapplied some fresh stuff properly.

Will be interesting to see the results.

the other thing that is very useful but again not for the faint of heart is to delid your cpu, and get direct contact between die and heatsink

candle_86 wrote:

the other thing that is very useful but again not for the faint of heart is to delid your cpu, and get direct contact between die and heatsink

if it's not soldered 😵

TandySensation wrote:

If you do go for make sure you do some good before and after testing, can you also take detailed pictures of the CPU and heatsink surface before and after? The image of the flat edge ruler along to the top of the heat spreader is interesting, maybe one like that too. Could you also clean the cpu and use good paste before you do the initial 'before lapping' testing. Sometimes I've wondered if the big boost from lapping was that the cheap/dried out/no thermal compound was the issue and after they lapped they reapplied some fresh stuff properly.

Will be interesting to see the results.

I cleanned the cpu and heatink with 99% rubbing alcohol before installing it and it has had enough time to cure. I will be useing the same TIM (MX-4)
I will do the befor testing today and stat lapping it too.

I always thought that this is the purpose of thermal paste, eliminating the gaps between the two surfaces.
Never done any substancial reading on the subject, but with my obviously short knowledge on it I wouldn't expect to experience any big differentiation between a stock cpu/heatsink/fresh well applied thermal paste and after lapping.
That said, go on but please document it properly!

konc wrote:

I always thought that this is the purpose of thermal paste, eliminating the gaps between the two surfaces.
Never done any substancial reading on the subject, but with my obviously short knowledge on it I wouldn't expect to experience any big differentiation between a stock cpu/heatsink/fresh well applied thermal paste and after lapping.
That said, go on but please document it properly!

TIM pushes the air out from inbetween the cpu and heatsink. I'll try to document every thing properly, I'm thinking that there will be a 5c drop.

Better to know in advance if the IHS is soldered to the die or not before attempting delidding.

Northwood A or B with TIM, no problem:

Later P4 with soldered IHS, destroyed:

In s478 HT Northwood-C and Preshotts the IHS was soldered.

yep, but you can remove even a soderd ihs with time, patience and heat, just take your time.