Madshrimps reviews CPU grease (TIM)

Felger Carbon · Post by **Felger Carbon** » Sun Dec 16, 2007 3:12 am

What impressed the heck out of me in this review is the vast difference between two methods of spreading the grease: spreading it yourself or letting the HSF-mounting do it. Skip straight to the last page of the review and see the graphs. You will be amazed!

The difference between brands was, by comparison; insignificant.

sjoukew · Post by **sjoukew** » Sun Dec 16, 2007 3:23 am

Indeed

thanks for the link

psiu · Post by **psiu** » Sun Dec 16, 2007 6:50 am

Wow--that's a big enough difference that I would believe it's probably legit.

Hmmph.

Wibla · Post by **Wibla** » Sun Dec 16, 2007 7:29 am

Looks like im switching both TIM and application method

thejamppa · Post by **thejamppa** » Sun Dec 16, 2007 8:11 am

Hmmm, I've always used AS5's method of spreading it. It seems I need to change to dollop way ^^

dhanson865 · Post by **dhanson865** » Sun Dec 16, 2007 8:29 am

Surprising but IHS reverses the trend. Apparently with a IHS you should NOT spread your goo.

Lawrence Lee · Post by **Lawrence Lee** » Sun Dec 16, 2007 10:12 am

thejamppa wrote:Hmmm, I've always used AS5's method of spreading it. It seems I need to change to dollop way ^^

AS5 has always recommended using the "dollop way" for CPU's with heatspreaders.

thejamppa · Post by **thejamppa** » Sun Dec 16, 2007 10:45 am

Lawrence Lee wrote:
thejamppa wrote:Hmmm, I've always used AS5's method of spreading it. It seems I need to change to dollop way ^^
AS5 has always recommended using the "dollop way" for CPU's with heatspreaders.

Darn, you were right. How come I always have had thought it was recomending the spread way.... Maybe I am getting demendend...

Lawrence Lee · Post by **Lawrence Lee** » Sun Dec 16, 2007 11:41 am

It's cool... I didn't know at first either. When I got my first tube of AS5 I spread it all over my P4 heatspreader... it was so sticky that when I uninstalled my Zalman 7000, the whole CPU came out of the socket with it. So I looked it up and whatdya know - they decided to change the application method. Luckily my CPU survived.

merlin · Post by **merlin** » Sun Dec 16, 2007 5:08 pm

Felger Carbon wrote:What impressed the heck out of me in this review is the vast difference between two methods of spreading the grease: spreading it yourself or letting the HSF-mounting do it. Skip straight to the last page of the review and see the graphs. You will be amazed!

The difference between brands was, by comparison; insignificant.

This all makes me glad I bother to remove the default thermal interface from oem heatsinks and reinstall with ceramique using dollup. It's a great confirmation of the advantages of tim. I'm wondering one thing, what is the composition of the Tuniq TX2 thermal interface? If it's ceramic based, I'd definitely pick it over ceramique.

spc · Post by **spc** » Mon Dec 17, 2007 12:43 am

Wish MS had tested this method too : carefully spread TIM all over the IHS, lightly draw a razor blade across the surface of the IHS to scrap off the excess TIM; repeat this procedure on the base of the heat sink. What you want is the thinnest and most evenly spread layer of TIM. You don't really want any TIM on the IHS, only in the crevices/cracks of the IHS & heat sink.

gb115b · Post by **gb115b** » Mon Dec 17, 2007 2:18 am

yeah, from the photos it looks like he's used waaay too much tim in his spread method....

i don't use a razor, just a nive flat piece of plastic and i get great temps...

spc · Post by **spc** » Mon Dec 17, 2007 4:34 am

Another thing I noticed was that in the dollop method, the TIM had not really covered the entire IHS. Probably if MS had used a slightly larger dollop, the TIM would be spread over a larger portion of ( if not all over ) the IHS.

Post by **NeilBlanchard** » Mon Dec 17, 2007 5:12 am

Hello,

spc wrote:Another thing I noticed was that in the dollop method, the TIM had not really covered the entire IHS.

OTOH, the CPU(s) are located only in the center of the chip. The results he got already reflect the relative effectiveness of the two methods. I'm certainly changing my method...

pipperoni · Post by **pipperoni** » Mon Dec 17, 2007 6:30 am

spc wrote:Wish MS had tested this method too : carefully spread TIM all over the IHS, lightly draw a razor blade across the surface of the IHS to scrap off the excess TIM; repeat this procedure on the base of the heat sink. What you want is the thinnest and most evenly spread layer of TIM. You don't really want any TIM on the IHS, only in the crevices/cracks of the IHS & heat sink.

While I'm not trying to knock your method, the beauty of the dollop method is that as the TIM heats up, it starts to flow on its own due to the pressure of the mounting clips. Excess TIM simply gets squeezed out the sides and big voids as is typical with IHS get filled.

Going for a very thin layer of TIM on both contact surfaces is a good strategy (I use it myself) as it involves a less messy cleanup and with very flat surfaces will get very good contact.

I bet over the course of a year, the effect of TIM pump out would be more significant with the dollop method and that the results if retested a year from now would be the same between dollop and spread.

JazzJackRabbit · Post by **JazzJackRabbit** » Mon Dec 17, 2007 1:42 pm

I'm vary about the MS results. Technically spread should be better than dollop because of the increased contact area between heatspreader and heatsink. Many people mistakenly argue that since the CPU die takes only a small portion of the total heatspreader area in the center, the paste should be applied only in the center. However, that is a mistake. It is hard to explain without a picture, but what people forget is that heat from CPU die travels in all directions in the heatspreader including to the outer sides of the heatspreader. It's not like only the center of the heatspreader will be hot and transferring heat to the heatsink while outer sides will be cold. It is true that the center of the heatspreader will be the hottest spot, but the outer sides of the heatspreader will be hot as well. So it makes sense to cover entire heatspreader with thermal interface.

The only two situations when spread method may be inferior to the dollop method is when 1) people put way too much thermal paste on the CPU, or 2) the CPU heatspreader is concave, so extra layer of paste on the outer sides of the heatspreader prevents proper contact in the center.

What I guess has happened is MS applied way too much paste on the CPU as indicated in their Tuniq TX2 picture. That is simply way too much and it was practically dripping from the heatsink by the looks of it when they took it off.

I always use spread method. I apply it on one side only, either CPU heatspreader or heatsink itself, whichever is smaller. That is in case with stock intel cooler I apply paste on the cooler itself because it's smaller than the CPU heatspreader. In case with Ninja I apply paste to the CPU heatspreader because Ninja base is larger than CPU itself. I usually drop a small drop in the center and spread it with scalpel as thin as possible removing any excess paste. When my scalpel starts exposing the surface I know I removed as much excess paste as possible. I've always done it and never had any problem. My temps have always been reasonable. Needless to say I won't be changing my method based on what appears to be a flawed MadShrimps review.

cmthomson · Post by **cmthomson** » Mon Dec 17, 2007 7:21 pm

Well, I'm on the completely opposite side from JazzJackRabbit.

[I'm talking specifically about AS5 on Intel IHS here...] The blob/streak method with this TIM uses the combination of the CPU heat and the mounting pressure to evenly spread the TIM from the center to the edge of the IHS, which automatically eliminates air bubbles.

This is superior to trying to spread an even layer of TIM and somehow have no air bubbles at all (obviously impossible).

On the other hand, on south bridge chips (typically in nowhere-near-flat ceramic packages), I prefer the spread method, which even though it won't eliminate air bubbles will maximize TIM/case/HS contact area.

spc · Post by **spc** » Mon Dec 17, 2007 7:32 pm

NeilBlanchard wrote:Hello,

spc wrote:Another thing I noticed was that in the dollop method, the TIM had not really covered the entire IHS.
OTOH, the CPU(s) are located only in the center of the chip. The results he got already reflect the relative effectiveness of the two methods. I'm certainly changing my method...

The point I was trying to make is : if the TIM had covered a greater portion of the IHS, the better the results would be. By definition, the IHS is a heat spreader, so anything that assists heat dissipation from the IHS should result in lower temps.

SPCR or FrostyTech could test the various methods mentioned in this thread using one TIM.

Felger Carbon · Post by **Felger Carbon** » Mon Dec 17, 2007 9:22 pm

spc wrote:Wish MS had tested this method too : carefully spread TIM all over the IHS, lightly draw a razor blade across the surface of the IHS to scrap off the excess TIM; repeat this procedure on the base of the heat sink. What you want is the thinnest and most evenly spread layer of TIM. You don't really want any TIM on the IHS, only in the crevices/cracks of the IHS & heat sink.

In a perfect world, where all IHSs and HSF bases are perfectly flat, the above makes a lot of sense.

Where perfect flatness does not prevail, extra TIM is needed to fill spaces which would otherwise (using infinitely thin TIM layers) be filled with air... highly undesirable. A thick (by comparison) layer of TIM is vastly to be preferred over air!

Controlled experiments using TIM would have to use flat surfaces. If sandpaper and elbow grease is required, so be it.

cmthomson · Post by **cmthomson** » Tue Dec 18, 2007 12:56 pm

I want to congratulate jmke for this meticulous set of measurements.

However, it occurs to me that using the stock Intel HSF will cause an understatement of the differences among TIMs and application methods.

Here's why: the stock HSF has a variable-speed fan that the BIOS controls based on the measured CPU temperature. So an inferior TIM or method would raise the temperature and thus the fan speed, resulting in a lower measured temperature than would be the case with a fixed-speed fan.

A more accurate approach would be to use a fixed-speed fan on an aftermarket HSF.

That said, using the stock HSF does have some merit: for a fairly large segment of the population, replacing the TIM but not the HSF has some appeal, and the article clearly points out the desirability of doing so. The result is both lower temperatures and lower fan speed relative to the completely stock build. This probably applies most to the non-DIY PC owner wanting to make a simple improvement to (say) a Dell machine.

alglove · Post by **alglove** » Tue Dec 18, 2007 4:39 pm

Can anybody explain to me why the Intel Stock Thermal Pad temperature goes down from 59.9 degC to 55.6 degC when the application method is switched from "spread" to "dollop"?

It is a thermal pad. Presumably, it just sits on the bottom of the heatsink. You don't spread it, nor do you dollop it. So why the temperature difference?

cmthomson · Post by **cmthomson** » Tue Dec 18, 2007 5:49 pm

alglove wrote:Can anybody explain to me why the Intel Stock Thermal Pad temperature goes down from 59.9 degC to 55.6 degC when the application method is switched from "spread" to "dollop"?

It is a thermal pad. Presumably, it just sits on the bottom of the heatsink. You don't spread it, nor do you dollop it. So why the temperature difference?

The TIM that comes pre-applied (in three stripes) on the stock HSF can be used only once. I think the 55.6 number may be a typo.

Note it is not a pad; it is three blobs of TIM that melt and spread after installation, just like most aftermarket greases.

willyolio · Post by **willyolio** » Tue Dec 18, 2007 7:20 pm

i've never really understood the reasoning behind the "spread" method. the whole point of thermal grease is to decrease the tiny air gaps between two surfaces (which are already quite flat).

now, unless you're capable of spreading grease across a surface and make it even flatter and smoother than the mirror-polished base of a heatsink, you'll be putting in more air bubbles/gaps than you're taking out.

i'd do a hybrid way. spread first, but scrape all of it off with a razor. ALL of it. the only stuff that might remain are the bits inside the tiny cracks. after that i dollop.

alglove · Post by **alglove** » Wed Dec 19, 2007 1:10 pm

cmthomson wrote:The TIM that comes pre-applied (in three stripes) on the stock HSF can be used only once. I think the 55.6 number may be a typo.

Note it is not a pad; it is three blobs of TIM that melt and spread after installation, just like most aftermarket greases.

Ahhh, so it is not a pad. That is where I was confused. That could mean that they ran one test with the three stripes, and another test with the TIN spread out? You say, however, the TIM can only be used once. Hmmmm.

Maybe the 55.6 number is a typo, but that is a 4 degree difference. That difference is on the same order as the other "blob" readings. In other words, I am wondering if the true difference in temperatures from using the "blob" method under load was really 2-3 degC, not the apparent 6-7 degC.

jimmyzaas · Post by **jimmyzaas** » Tue Dec 25, 2007 9:50 pm

crap and here I got suckered into buying Arctic Cooling MX 2.. when my AS5 is apparently better.. damn it all!

at least it was cheap..

Madshrimps reviews CPU grease (TIM)

Madshrimps reviews CPU grease (TIM)

Re: Madshrimps reviews CPU grease (TIM)