Lapping My PIV

This is my trusty little P4 2.53

1

The Pentium 4 chip is covered with a solid heat spreader that protects the fragile core from being crushed by today’s heavy CPU coolers. As you see, the stepping model # and other information is printed on it.

I’ve been reading up on this heat spreader and found one person selling a P4 3.06 with a void warranty because he lapped it. I asked him why in a PM at another forum and he replied that most P4 heat spreaders are not even close to being flat and lapping them will give you a decrease in CPU Temperatures.

SO I decided to try it.

{mospagebreak}


Murdawg

Before You Lap

To lap any processor, you need the following:

  • Two or three different grits of sandpaper IE 600,1200,2000,3000 etc.

    1

    1
  • A perfectly flat surface (large sheets of glass work perfectly)
  • Rubbing alcohol

    1
  • Q-TIPS

    1
  • No fear of voiding warranty

    1
  • A pailful of courage

    1

    (Ed. note: OK, OK, I added the last two pictures to the piece. Get mad at me, not him for it. :)–Ed){mospagebreak}


    Murdawg

    First, The Disclaimer

    THIS ACTION CAN POSSIBLY DAMAGE OR DESTROY YOUR CPU. I TAKE NO BLAME NOR ASSUME ANY LIABILITY FOR SUCH. THIS IS A REVIEW ON HOW I LAPPED MY CPU AND A QUICK GUIDE ON HOW TO AND THE BENEFITS FROM. YOU DO IT, IT’S SOLELY YOUR RESPONSIBILITY.


    First, you put the CPU core down on the glass surface or whatever other flat surface you are using and use the lowest grit (roughest) sandpaper.

    1

    Moisten the sandpaper a little with some water and start to slowly and gently rub the heatspeader on the paper.

    After about a minute, flip over the cpu and with some rubbing alcohol and the Q-Tips and look at the results.

    You should see the some scratch marks where the sandpaper is rubbing away at the heat speader. You will notice the heat spreader is nowhere near flat and by the pattern of the scratch marks, you’ll see that they are not covering the entire heat speader but just in certain areas. These are the highest areas of the spreader which need to worn down.

    1

    1

    1

    CONTINUED page 4…

    Murdawg

    Clean up and get the sandpaper wet again and continue to lap the heatspreader and keep checking the buttom about every minute to check on your process.

    You will start to notice it change in some areas from the original aluminum to a copper color.

    1

    1

    1

    {mospagebreak}


    Murdawg

    Keep sanding as you it will start revealing more of the hidden copper and less of the original heatspreader.

    1

    1

    1

    1

    1

    1

    {mospagebreak}


    Murdawg

    Continue sanding until the entire CPU is copper-colored. Now you’ll know for a fact that your cpu heatspreader is very, very flat.

    1

    1

    Now when you reapply your cpu cooler you will need hardly any of the thermal paste or arctic silver you needed before.{mospagebreak}


    Murdawg

    I took the whole thing to the extreme and used some very fine polishing steel wool and gave the CPU a mirror finish.

    1

    1

    1

    This isn’t necessary and some believe it actually increases temps so I tested before and after.{mospagebreak}


    Murdawg

    Here’s are the coolers used:

    1

    1

    1

    And here are the results. All tests done with stock voltage and, after a reboot, running 20 minutes of FOLDINGATHOME 100%CPU usage

    BEFORE LAPPING

  • Stock Intel cooler with thermal pad 47C
  • Stock intel cooler with no pad and AS3 44C
  • AX-478 with Volcano 9 “COOL MOD” fan at low rpm 42C
  • AX-478 with Volcano 9 “COOL MOD” fan at high rpm 37C (75cfm)

    AFTER LAPPING

  • AX-478 with Volcano 9 “COOL MOD” Fans at low rpm 37C
  • AX-478 with Volcano 9 “COOL MOD” Fans at high rpm 33C (75cfm)

    AFTER MIRROR

  • AX-478 with Volcano 9 “COOL MOD” Fans at low rpm 38C
  • AX-478 with Volcano 9 “COOL MOD” Fans at high rpm 33C (75cfm)

    {mospagebreak}


    Murdawg

    You can see that going from the stock Intel cooler to the AX-478 helped me drop anywhere from 2-7C (depending on whether you prefer whisper silence or a 48db loud fan).

    The lapping dropped an extra 5C on low rpm and 4C on high.

    I tested after mirroring because like I said either some people believe that mirroring a product doesn’t help and I did see 1C higher temps at low RPM but that could easily be an error. However, mirroring didn’t provide any big bonus, either.

    All of these tests shouldn’t be taken absolutely literally because they are guaranteed to be + or – at least 1 or 2 degrees on any particular test. Pulling off a heatsink and reseating it doesn’t always give you the exact same temperatures, either.

    For anybody wondering about the base of the AX478, I found it to be near perfect and didn’t bother lapping or polishing it. I also used slightly less AS3 after the lapping and that could also have a small effect positively or negatively to the decrease in temperatures.

    All that being said, though, the lapping helped.

    I just got my entire DangerDen watercooling kit, so expect a hands on review and a boat load of benches with the Radeon 9700 and 2.53 water-cooled.

    If you have any question or comments, feel free to email me at the address below.

    Comment page 10…

    Murdawg

    While most of the experimentation done by DIYers is empirical in nature, the relationship between cause and effect should be evaluated critically. Additionally, extrapolations to a ‘general case’ from a sample size of one should be viewed as speculative – at best.

    The Intel HIS (integrated heat spreader), despite its name, is utilized to provide crush resistance for the CPU, as it is too thin to effect significant lateral heat dispersion. It is designed to deform under load, that load being the recommended clamping force.

    So what is known about the P4’s IHS installed flatness when it is observed ‘unloaded’?

    Nothing, thank you.

    Are ‘all’ (or even most ?) ‘unloaded’ P4 IHSs the same?

    My observations of descriptions in various forums (Overclockers Forum and ProCooling Forum) lead me to believe there is significant variation. But what NONE of us know is the IHS’s flatness when loaded.

    Is there ANY reason to believe that Intel intended it to be flat? No, quite the contrary: A convex surface will ensure (and promote) good contact and maximum pressure over the die area.

    At which juncture a critical reader might say “So what, look – lapping it dropped the temps.” OK, but why? Is enough known to establish a causal relationship? Probably, but I would suggest not enough to make much of a recommendation.

    The first difficulty is that the ‘flatness’ of the mating surface, the HSF, is not known – in any fashion. So if the lapping of the IHS resulted in a better ‘fit’, then the results would (also) be as described.

    Is the described lapping procedure going to produce a flat surface? I can state definitively: No. It WILL produce a ‘high’ center. (Do a search for ‘flatness’ on this and other site’s forums and many images posted by me can be seen.) This high center is not necessarily harmful, as at least some of Intel’s intended loading has been maintained.

    The author correctly addressed the uncertainty of temperature measurements. I would find the conclusions more credible if there were ‘before’ and ‘after’ data sets to compare, rather than single values. This is not being picky, this is about demonstrable validity and reasonable conclusions.

    In the near future, we will have a new generation of IHSs from both Intel and AMD to ‘play’ with, and with a 90 lbf load, disaster will befall those who ‘get it wrong’. Lapping a IHS without understanding the effects on the thermal joint system (CPU die -> TIM joint -> IHS -> TIM joint -> HSF or WB) is not recommended.

    Bill Adams

  • Be the first to comment

    Leave a Reply