Also shows how much the temp improves – Cody Stasyk
Ever since I started overclocking I, like many of you, have always been trying different things to lower my processor’s temperature. Sure, there’s the conventional things you do to shave a few degrees here and there. You buy the latest and greatest heatsink, you route your cables behind your motherboard tray and stuff them into empty drive bays to improve airflow, you try all sorts of ducting schemes, add a few extra case fans, maybe even mod your case so you can pop in a few 120 mm fans.
But what about after that? What else is there to do? Yes, I suppose I could go out and spend a couple hundred bucks on a water cooling setup, or get a second job so I could buy a Prometeia or Vapochill unit, but I’m on a bit of a tight budget right now, and I don’t have the time for a second job.
After searching around for additional ways to cool my P4, I came across an article right here, at Overclockers.com, where a person lapped his P4’s IHS (Integrated Heat Spreader) and saw his CPU’s temperature drop 4°C under load. After reading the article, I decided that it would be worth a try, so I lapped down into the copper of my chip’s IHS, reassembled my machine, ran Prime95 for an hour, and guess what? A 4°C drop in my CPU’s temperature at full load.
That held me for a few months, but then, as always, I began looking for more ways to keep my processor even cooler. I had seen a few people on the forums recommend removing the IHS from a P4 to make it run cooler, but as always, there were those who disagreed. Some stated that the IHS distributes the heat from a P4 evenly, over a larger surface area, resulting in better heat distribution, and thus, lower temperatures.
The way I saw it, the faster heat could be removed from the die, the cooler the processor would run. To me, having the die in direct contact with the heatsink (with thermal interface material between the two of course) would remove the heat faster than if the heat were to pass through a nickel/copper cap and an additional layer of thermal compound (I’ll explain later) before reaching the heatsink. I wasn’t sure though, and no matter how hard I looked, I couldn’t find any sort of article dealing with the removal of an IHS, so I decided to try it out myself.
Having just upgraded to a new P4, I was a little concerned about the consequences if I were to somehow screw up. So to avoid the risk of destroying my new chip, I decided to remove the IHS on my socket 423 Willamette P4 that I had been using up until a few weeks ago, instead. While it may be a different processor than the Northwood chips, I believe the results that I obtained are not specific to socket 423 chips, and I believe that had I used my Northwood chip instead, I would have seen similar results.
Before removing the IHS, I put together a barebones machine for testing the thermal results before and after my little experiment. Here are the specs of the machine:
- 1.7 GHz Pentium 4 (socket 423 Willamette) @ default speed,
but @ 1.80v (default is 1.75v for this chip)
- Asus P4T motherboard (i850 chipset)
- 256 MB of RDRAM (two 128 MB RIMMs, and two continuity RIMMs)
- GeForce 2 GTS graphics card
- Sparkle 300W power supply
- Stock Intel heatsink and fan
I laid the components of this machine out on a table for this test. The reasons for doing so were threefold:
- The ambient air temperature would remain controlled as long as the room temperature remained constant
- It allowed me to take better pictures of the machine
- I was too lazy to cram everything into a case 🙂
After setting everything up and planning everything through, I was ready to go.
Here’s a shot of the barebones machine:
I applied a thin, even layer of Arctic Silver 3 thermal compound per Arctic Silver Inc.’s instructions HERE.. I then mounted the stock Intel heatsink and fan that came with my processor.
To monitor the room temperature I used this thermometer:
It’s pretty old, and I know that this is an international website, so I apologize for the Fahrenheit readings, but for the purpose of this experiment, I am monitoring for changes in the room temperature, not its actual value. For the record though, before turning the machine on, the room temperature was about 64°F, or 18°C (yes, it is cool in my basement).
For each of the tests, I wanted to record my processor’s temperature under load, using Prime95. The problem I ran into, however, was that I only have a single hard drive at the moment, which I’m using in my newer computer. I tried hooking the hard drive up to my test machine, but it refused to let me into Windows XP.
Thus, I would have had to reformat and reinstall XP just to run Prime95 and record temperatures for this article. To tell you the truth, I would have done just that, if I had only had some way of conveniently backing up my hard drive (I’m not about to burn everything onto 50+ CD-Rs).
To measure my processor’s temperature throughout the experiment, I booted into BIOS and let my computer sit for twenty minutes in the “Hardware Monitor” section. After twenty minutes, I returned to the computer and monitored the CPU temperature for roughly half a minute to make sure it had stabilized and was not fluctuating. I then took a photo of the BIOS screen, displaying the CPU temperature.
After letting the computer sit for twenty minutes in the BIOS, this is what I observed:
With the IHS, my processor sat at 39°C in the BIOS. I was pretty confident in my results, as I observed similar temperatures when I used these components as my main computer in the past, but for the sake of precision, I decided to re-test the thermal results.
I removed the processor from the ZIF socket, cleaned the IHS with 99% isopropyl alcohol, reapplied Arctic Silver 3 thermal compound to the processor’s IHS in a thin, even coat, popped the chip back in, and re-mounted the stock Intel heatsink and fan. After twenty minutes of sitting in the BIOS, my computer reported these results:
After twenty minutes, my processor sat at 38.5°C. I attribute the half degree difference between the first test and this one to my application of the Arctic Silver 3. It’s pretty hard, if not impossible, to apply an identically-thick layer of thermal compound from one time to the next.
So the temperature of my processor with the IHS was 38.5°C – 39°C.
After testing the thermal results with the IHS on the processor, it came time to remove the Integrated Heat Spreader from my P4. Having found no guides or instructions of any sort on removing the metal cap, I tried to think of the best way of removing it without damaging the processor.
Taking a look at where the IHS and processor meet, I could see that the IHS was attached to the processor with some sort of clear, hard adhesive all around the IHS’s edge, except for the middle section on each side, where a small gap between the processor and the IHS existed. Unfortunately, I did not take a photo that can show this in great detail, but I’ll try to demonstrate:
Here is the processor:
Here’s the exact same photo, but with red markings indicating where the clear, hard adhesive holds the IHS onto the rest of the processor, along the edges of the IHS. Notice the gaps in the adhesive, where a slight gap exists between the processor and the IHS:
Here’s a closer shot of an edge of the IHS:
Here’s the exact same photo, but with the contrast and saturation punched way up and the color balance altered to more clearly show where the hard, clear adhesive holds the IHS onto the rest of the processor:
In the above photo, you can see where the adhesive (bright pink) exists along the edge of the IHS (blue).
I decided to try to remove the IHS by inserting a thin blade into the gap between the IHS and the processor, and sort of “saw” my way through the clear adhesive.
I started sawing, but it quickly became apparent that I was getting nowhere. The adhesive was a lot harder than I thought it would be. After trying numerous methods of “sawing” and “slicing” and becoming progressively frustrated each time, my gentle “sawing” method eventually changed to more of a “grab-the-chip-and-jam-the-blade-towards-the-corner-with-as-much-pressure-as-possible” method.
Twenty frustrating minutes and two broken blades later, I had managed to hack my way through most of the adhesive along the four sides of the IHS.
The corners, however, were another story. No matter how hard I tried, I couldn’t sink the blade into any one of the corners’ adhesive. The problem, as I saw it, was that I couldn’t apply enough force with the small blade I was using, so I grabbed a longer blade.
I placed the long blade at one of the corners of the IHS and pushed it in, towards the corner, with my thumbs. After struggling with it for a while, the blade slowly started to slice into the corner’s adhesive, and I had soon after finished cutting through the first corner:
After doing the same to the other three corners, the IHS was only holding on by a few strands of adhesive, so I took my blade and wedged it below one of the edges of the metal cap:
After some minimal pushing, the blade sliced through the remaining adhesive on the one side of the IHS, and with a little prying, the entire thing popped off, exposing the die:
You can see where the adhesive held the IHS onto the rest of the chip, with the black marks around the edges of the raised brown part of the processor, and you can see where the gaps in the adhesive existed.
Notice all the dirt and powder around the die. This is from lapping the IHS months ago, as some of the nickel and copper powder must have seeped beneath the IHS, through the gaps in the adhesive. Also note the white thermal compound on the die, which was very oily.
Here’s the processor after I cleaned it up:
Here’s a shot of the Integrated Heat Spreader:
I found the thermal compound that Intel uses between the P4 die and the IHS to be fairly oily. Looking at the image above, it would appear Intel uses a fairly large glob of this thermal compound during the manufacturing process, as a fairly thick chunk of it framed the die.
With the IHS removed, I popped my P4 back into my barebones testing rig:
I applied a thin, even layer of Arctic Silver 3 to the die and then mounted the stock Intel HSF. While mounting the heatsink, I was sort of half-expecting to hear something crack, as the stock Intel heatsink retention mechanism applies a lot of pressure to the heatsink.
Luckily, I heard no such sound. Since I removed the IHS, the heatsink sat slightly lower than when the IHS was atop the processor, so a little less pressure was needed to mount the heatsink, which was definitely a good thing.
Before I booted up, I checked the temperature of the room – it was about 64°F, or 18°C, just before I booted. Compared to temps just before testing the thermal results with the IHS on the chip, it appears that this time the room temperature was just a fraction of a degree lower. For all intents and purposes, I am going to consider this difference negligible, as it is hardly even noticeable and would not have had an impact on the results of my thermal testing.
After booting the machine up and letting it sit for twenty minutes, this is what I found:
So my first test resulted in my processor running at 36°C; two and a half to three degrees lower than with the IHS.
Now let’s take a look at the results from my second test:
The results of the second test were identical to those of the first, so it’s fairly safe to say that, for my Pentium 4, removing the IHS lowered its temperature by two and a half to three degrees Celsius.
Note the motherboard temperature. With the IHS on the processor, the motherboard temperature was 25°C, but here, without the IHS, the motherboard reports its temperature to be 27°C. Why the difference? My best guess is that since the processor runs cooler without the IHS, the heat it pumps out is transferred faster to the heatsink than when the IHS is present.
Thus, the heatsink would become slightly hotter without the IHS on the processor, and thus the air exiting the fins of the heatsink would be slightly warmer. This warmer air would, in term, raise the temperature of the motherboard, since it blows overtop of it after leaving the fins of the heatsink.
It is important to remember that these temperature readings were taken from within my BIOS, not from within Windows while running a stress program. If I had been able to take the thermal results while the processor was under a heavy load, the impact that removing the IHS would have had on cooling would have been larger than what was observed in this experiment, since processors consume more power when under load than when they are sitting idle in a BIOS.
Factor in overclocking (ie: increasing the speed and voltage of the processor), and the difference would be even more significant.
As I stated earlier, although I tried this experiment on a socket 423 Pentium 4, I don’t believe that the results I obtained are processor-specific. This is, of course, just a hypothesis of mine, so I could be right and I could be wrong.
So would I recommend removing the IHS from a Pentium 4? Well, yes and no.
No, because the actual process of removing the IHS can be immensely frustrating and difficult. Also, if you are not “careful”, you could render your processor useless, not only while removing the IHS, but also while trying to mount a heatsink. I know there has been a lot of debate over the main purpose of the Integrated Heat Spreader, but no matter what it is, the IHS does provide excellent crush protection for the processor.
On the other hand, if you are careful, you have patience and you can afford to replace your processor if something goes horribly wrong, then yes, I would probably recommend it.
There are definitely rewards to be reaped here. Removing the IHS could lower your overclocked P4’s load temperature by four or five degrees Celsius, and that could possibly enable you to push your chip a little further.
If anyone has any questions or comments on anything in this article, please send me an email.