Pitfall of Prescott's Stock Heatsink Fan: stock Thermal Grease versus Arctic Silver 5

[death metal]

Pitfall of Prescott's Stock Heatsink Fan: stock Thermal Grease versus Arctic Silver 5

***Introduction***
Since Prescott has been released, I have been able to build quite a few rigs based on this really highly overclockable bad boy. The flavors of Prescott that I'm lucky to get in touch with spans the 2.4A youngest of the family, to the hippy 3.0E big brother, all based on Socket 478. There have been a lot of reviews of Prescott, and most of the articles never failed to mention how "hot" this baby is. And to combat the heat, comes a new thermal design solution from Intel: al-cu combination.

(I won't delve and discuss anything in-depth about Prescott, and its architecture, the net offers a lot of information on that already. Rather, I'll be sharing what I found to be something to watch out for using the stock thermal design solution)

***System Specifications***
Pentium 4 2.8E@3.6Ghz, 1.525vcore
Asus P4C800
2x512MB OCZ EL GOLD PC4000
40Gb ATA133 HD
Thermaltake XaserIII LANFire
Windows XP SP1, Clean Install

For monitoring, I use Coolermaster Aerogate II, Thermaltake Hardcano 12, and Motherboard Monitor.
For heating up the CPU, I use PCMark2002/Prime95(1)/Prime95(2), PCMark2004/Prime95(1)/Prime95(2), Aquamark03, 3DMark01SE, 3DMark03, SiSoft Sandra Burn-In/20runs


***Trial by Fire***
Test1: Stock Heatsink, Stock Thermal Grease
Test Routine: Running the above mentioned benchmarks for two weeks. The temperatures listed is already an average of the number of times the data has been taken within the duration of the test.

Ambient Temp: 33c
30min Idle: 57c
60min Load: 66c

The test performed above will be our basis.

***A freaky turn of events***
Test2: Stock Heatsink, Arctic Silver 5
Test Routine: Running the above mentioned benchmarks for two weeks. The temperatures listed is already an average of the number of times the data has been taken within the duration of the test. Note, at times, I have to push down the OC to 3.5Ghz just to ensure I meet the 60min load time because the system is showing instability at high temperature.

Ambient Temp: 31c
30min Idle: 62c
60min Load: 72c

Surprisingly, Test 2 shows that the AS5 seems to be performing very poorly. I suspected I may have missed something in the steps. But I actually followed the instructions on the product, and have been using the system and shutting down/restarting afer use during the two week period. Now, to be fair, I decided to scrap the data I gathered during the two weeks, and decided to observe the set up again.

***The doubt of the unbeliever...***
Test3: Stock Heatsink, Arctic Silver 5 (heatsink cleaned, AS5 reapplied as per instructions, see quote below)
Test Routine: Running the above mentioned benchmarks for two weeks. The temperatures listed is already an average of the number of times the data has been taken within the duration of the test. Just like with Test 2, at times, I have to push down the OC to 3.5Ghz to achieve stability.

Ambient Temp: 32c
30min Idle: 61c
60min Load: 72c

AS5:
Important Reminder:
Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity. (This period will be longer in a system without a fan on the heatsink or with a low speed fan on the heatsink.) On systems measuring actual internal core temperatures via the CPU's internal diode, the measured temperature will often drop 2C to 5C over this "break-in" period. This break-in will occur during the normal use of the computer as long as the computer is turned off from time to time and the interface is allowed to cool to room temperature. Once the break-in is complete, the computer can be left on if desired

As the data shows, not much changes. I highly doubt that the stock heatsink is the culprit here. Test1 shows it is a capable heatsink. I wouldn't fault AS5 as well, my past experience with this product meets my expectations all of the time. So, could it be that the new thermal grease that Intel is using on their new heatsink be a better performer than AS5? That's the only logical explanation I can think of as far as my little coconut can discern. But what I really fear is that, I may have screwed up my data collection in step1 for some reason. I have no way to retest the set up since I have already scrape off the thermal grease =) yay!!!

I research about this on the net, but I can not come across any info regarding this. Oftentimes, the enthusiast will skip the stock heatsink and use an aftermarket fan. And if anyone is using the stock heatsink, I haven't found anyone who use AS5 with it. Well, I have no other means to verify my findings so I just kept silent and dismiss my test as a freak of nature =). Well, until a week ago, I got a request for a new rig. So I sent in my proposal for a 2.8E based rig and we got a "go" =). Afer acquiring the CPU, I put it up on my test rig above and performed my tests.

***Into the pit of fire, I dived back again***
Test4: Stock Heatsink, Stock Thermal Grease
Test Routine: Running the above mentioned benchmarks for 3 days. The temperatures listed is already an average of the number of times the data has been taken within the duration of the test. CPU sSpec is the same.

Ambient Temp: 34c
30min Idle: 57c
60min Load: 65c

Comparing it to Test1, this solidify my belief that my initial tests was accurate. This is both a relief to me, but also very intriguing. So I dismiss the idea of me screwing my initial baseline test, but there is still one lingering question: does the new thermal grease really performs better than AS5? (To be honest, I don't think the thermal grease used with Prescott is any different with Northwood, I never noticed any difference actually).

(continued...)

[death metal]

***A mystery.....solved?***
So my quest is now focused on finding the fault of AS5 (or is it the heatsink's fault?). I popped out the heatsink and studied how it is making contact with the CPU heatspreader. Indeed, nothing really spectacular, except the fact that the copper portion is a bit extruded compared to the alluminum fins.......wait......wait.....I think I got it........

Reading AS5 instructions again: http://www.arcticsilver.com/arctic_silver_instructions.htm

10. On an Intel P4 or Athlon64 type CPU with a large metal heat spreader, put a small amount of Arctic Silver onto the center of the heat spreader as shown in the photo.
Only a small amount of Arctic Silver is needed

P4- About the size of an uncooked grain of short-grain white rice or 1/2 of a BB.

Picture Linked from AS5 without Permission =): http://www.arcticsilver.com/images/p4_as_dab1.jpg


......Since the vast majority of the heat from the core travels directly through the heat spreader, it is more important to have a good interface directly above the actual CPU core than it is to have the heat spreader covered with compound from corner to corner.

Picture Linked from AS5 without Permission =): http://www.arcticsilver.com/images/p4_as_dab2a.jpg

Now, with older Northwood's heatsink/fan solution and with other aftermarket heatsink fan, the steps is perfect. If anyone has observed how AS5's characteristics is on this "flat" heatsinks, AS5 tends to spread out, covering a large area of the heatspreader. And in the event that AS5 doesn't spread out "that much", it still maintains contact with the heatsink. *DING* *DING* *DING*.

I know, I know, please don't kill me. I know that heatsinks are supposed to be "flat" to provide better contact, and hence, better heat transfer. But hey, don't spank me for believing for a second that the heatsink is not at fault. Look at my Test1 and Test4 results. See how I missed it? Now, let's prove it with a simple test.

***Harking back, I conquer thee***
Test5: Stock Heatsink, Arctic Silver 5 Plus Arctic Alumina
Test Routine: Running the above mentioned benchmarks for 3 days. The temperatures listed is already an average of the number of times the data has been taken within the duration of the test. Full stability during test.

Ambient Temp: 31c
30min Idle: 56c
60min Load: 64cc

Muahahahahahahahahahahhaa.......MUAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA...............

So, what's with Arctic Alumina? Well, if anyone is at lost (I can't blame you, I am not a good writer), like what I said earlier, I need better contact. And I'm sure you expert guys know that and wanting to kill me for this freaky mistake. But like I already said again, the stock heatsink/grease performed better and I was under the impression that for some reason, the stock thermal grease is doing some magical thing. But the fact is, the thermal grease provided with stock sink is good since it provides a means of contact not only to the copper core but to some degree, to some of the alluminum fins as well. So what's with the Arctic Alumina?

Please take a look at this image (shamelessly linked without permision from SilentPCReviewhttp://www.silentpcreview.com/):
With ThermalGrease: http://www.silentpcreview.com/files/images/intel-hsf/intel-hsf4.jpg
Without ThermalGrease: http://www.silentpcreview.com/files/images/intel-hsf/intel-hsf3.jpg

You will notice the thermalpad covers the copper core, and actually, this pad spreads almost throughout the CPU's heatspreader providing contact to the other alluminum portion of the heatsink and not just the copper core. Now, here's what I did:

1) I followed AS5 instructions, and placed the heatsink on the cpu
2) I then removed the heatsink again, spread some Arctic Alumina along the copper core sides (aluminum side)
3) Replaced the heatsink to the CPU

The crappy picture: http://www.sleektech.nl/~dm/temp/as5alu.jpg

With the addition of Arctic Aluminina into the scene, I was able to hit the "magic" numbers that the default thermal grease provided. Note that adding the Arctic Alumina needs to be a little bit "more than usual" since the slightly protruding copper core will prevent the bonding of Arctic Alumina with the CPU's heatspreader if its applied in the "usual" fasion (see Arctic Alumina's instructions).

***I lay down to rest***
As a concluding remark, I would like to stress out that the stock heatsink/grease works great with little to decent OC. As long as you are not going to tweak the voltage (stock OCing), I doubt that you'll need better aftermarket heatsink fan. And if you decide to replace the thermal grease like I did (and most I read on the net), don't forget to ensure that the outer alluminum sides, lining the copper core, is also getting contact with the heatspreader, but there's a catch:

a) If you use AS5, use Arctic Alumina to give contact to the sides or
b) Skip AS5, and use Arctic Alumina exclusively

That's all folks, till next of my mis-adventures....

Credits:
Thanks to myself for providing all the needed test equipment =)...

* Kudos to Intel for the great chip...
* A whooping hurraah for Asus motherboard, your CPU retention survives my misadventures...
* Big thanks to the fast RAM by OCZ that maintains stability all throughout the testing...

[Snafu]

Tres cool report :cool: :thumb: .

In the end either use a dab of AS5 in the centre and the Alumina around the edges or just strictly use the Alumina.

I have a Zalman 7000 AlCu heat sink coming and wonder if I should use the grease it comes with or go for the Artic Silver/Alumina combo. Possibly also lapping the heat sink (not sure if the later is necessary with a new heat sink).

Thanks again for the great report :thumbs: .

[death metal]

tnx mate =), this may seem to be the reason why some early reviews of Prescott with Stock HSF (but with AS5 to replace the grease) shows extreme rise in temperature.