Arctic Cooling MX-2 question

[ImageMD]

I've been having difficulty trying to find what the ideal amount and technique is of applying this particular paste (I looked through the stickies, searched the forums, googled..).

Last time I built a system I just made a super-thin veneer of paste that I spread with a plastic card...but, after recent searches, it seems that today people are either putting a small dollop / "cooked grain of rice" right on the center of the IHS or a "line" extended from one end of the IHS to the other....and then just plopping the HSF on top and twisting it a bit.

These are the company's own instructions:
http://www.arctic-cooling.com/downloads/installation/install_mx2_retail.pdf

But, I'm just not sure that's the ideal way to do it. Any input would be appreciated.

 

[rainman]

The company's instructions seem to work just fine for most of us here. Last time I checked, they did design the stuff...they ought to know how it works best.

 

[Brahmzy]

I disagree with the rice sized dollup method. I have tried various methods well over 100 different times building PCs and have found that an extremely thin uniform spread across the entire IHS is a MUCH better method. Too many times using the single drop method have left the corners/edges naked. That doesn't make for proper heat conductivity. Even done various temp tests to back it up.

 

[ImageMD]

The company's instructions seem to work just fine for most of us here. Last time I checked, they did design the stuff...they ought to know how it works best.

Well, part of my uncertainty is that I think they made their instructions for singlecore CPUs, before C2D and C2Qs even came out. It's identical to their MX-1 instructions anyways.

 

[filthysanchez]

I just make a thin line across the IHS.

 

[rainman]

Well, part of my uncertainty is that I think they made their instructions for singlecore CPUs, before C2D and C2Qs even came out. It's identical to their MX-1 instructions anyways.

Well, if you spent more than 13 seconds on this website, you would find: This thread stickied at the top of this forum. If you look in the second post, it shows how much to use. It comes from this document.

http://www.arcticsilver.com/pdf/appinstruct/as5/ins_as5_intel_quad_wcap.pdf


Now, that's AS5 and not MX-2. But it is the same exact principle. You're just taking up the air space between the IHS and your heat sink.

I disagree with the rice sized dollup method. I have tried various methods well over 100 different times building PCs and have found that an extremely thin uniform spread across the entire IHS is a MUCH better method. Too many times using the single drop method have left the corners/edges naked. That doesn't make for proper heat conductivity. Even done various temp tests to back it up.

By the way, unless your testing showed temperature variances greater than 10* based on application method, your results are statistically insignificant. The corners and edges don't matter compared to the area centered over the actual cores.

 

[ImageMD]

Well, if you spent more than 13 seconds on this website, you would find: This thread stickied at the top of this forum. If you look in the second post, it shows how much to use. It comes from this document.

http://www.arcticsilver.com/pdf/appinstruct/as5/ins_as5_intel_quad_wcap.pdf


Now, that's AS5 and not MX-2. But it is the same exact principle. You're just taking up the air space between the IHS and your heat sink.
.

I already saw that, mentioned that method in my first post, and thought I was being clear when I specified I'm asking particularly about MX-2. There is no consensus in this thread so far. Maybe I'm overthinking things and in the grand scheme none of this really matters. But it doesn't help that you contradict yourself by now advising I should follow AS5's method over AC's.

 

[Brahmzy]

By the way, unless your testing showed temperature variances greater than 10* based on application method, your results are statistically insignificant. The corners and edges don't matter compared to the area centered over the actual cores.

Sorry, I've done very repeatable testing MANY times on this. My resusts DO MATTER in fact. Granted they are only about 2 degrees different almost EVERY time, but that does matter to me. Just like buying a Ultra 120 and using two fans. Why spend time to lap and IHS an $80 HSF fan sandwich only to get the "middle" on the contact area covered. That's just friggin' DUMB. And the results prove it. The IHS is, in itself, a heat conductor - the ENTIRE thing. Sheesh.

 

[Sphere]

Sorry, I've done very repeatable testing MANY times on this. My resusts DO MATTER in fact. Granted they are only about 2 degrees different almost EVERY time, but that does matter to me. Just like buying a Ultra 120 and using two fans. Why spend time to lap and IHS an $80 HSF fan sandwich only to get the "middle" on the contact area covered. That's just friggin' DUMB. And the results prove it. The IHS is, in itself, a heat conductor - the ENTIRE thing. Sheesh.

I've spent a fair amount of time with different methods, and I cover the entire IHS, that's where I have had the best results as well.

 

[rainman]

It's all the same. You all can fanboy whatever method you want. But the fact of the matter is simply: it doesn't fucking matter that much.

2* is not a valid difference. That is easily within the probabilistic error of the measurements you're making.


Cover the cores. That's all you really need to do.




As for contradicting myself...you weren't happy with the answer of "use the manufacturer's methods," so I gave you another manufacturer's reccomendation.

But, please, everyone continue to freak about statistically insignificant numbers.

 

[Brahmzy]

It's all the same. You all can fanboy whatever method you want. But the fact of the matter is simply: it doesn't fucking matter that much.

2* is not a valid difference. That is easily within the probabilistic error of the measurements you're making.


Cover the cores. That's all you really need to do.




As for contradicting myself...you weren't happy with the answer of "use the manufacturer's methods," so I gave you another manufacturer's reccomendation.

But, please, everyone continue to freak about statistically insignificant numbers.

Continue to run at least 2 degress hotter than me then. Thanks. I can't even understand this is up for question. Any OC pro will tell you to cover the entire thing. This is not because it just sounds or looks good - it's because it's been PROVEN. If you call 2 degrees cooler EVERY TIME insignificant, then I guess you're not a real enthusiast and don't care about the best possible method. Let me guess, you also believe RAID0 is a no-no and provides no additional gains?

 

[rainman]

It's up for discussion because we're not talking about the most reliable means of measuring the temperatures, now are we? If we were ALL measuring temperatures like Intel wanted, then it would be a different story. Speaking of...if that IHS was so incredibly important to be completely flat and covered in paste, why then would the STOCK Intel assembly not have enough paste on it for that coverage? Why also would they give you guidelines to cut a channel in your IHS to measure temps in accordance with their documentation...if that IHS mattered any more than just the areas over the cores.

However, there are just too many variables in this whole shebang to consider 2* significant other than for bragging rights. I'm still running 35-40* cooler than Intel's max specified temperatures.

Let me put it another way. Cores 1/2 on my processor are currently running at 39. Core 3 is 36 and Core 4 is 38. Do you really think the way I applied my paste matters when the cores are on the SAME DIE? Under load, the cores are 62/62/54/54. Oh wait, no they're 58/58/55/55. Oh, now they're 63/63/56/56. Hm...I suspect this is just the natural variance in the 1) DTS on each core; 2) TIM between the IHS and cores...THAT I CAN DO NOTHING ABOUT!




RAID has nothing to do with this discussion other than a non-sequitor and a lame-ass personal attack. There are pros and cons to every RAID configuration. RAID0 provides performance increases but no data security. Use it if you want, but don't cry if you lose all your data...or have a reliable backup plan.

 

[Brahmzy]

It's up for discussion because we're not talking about the most reliable means of measuring the temperatures, now are we? If we were ALL measuring temperatures like Intel wanted, then it would be a different story. Speaking of...if that IHS was so incredibly important to be completely flat and covered in paste, why then would the STOCK Intel assembly not have enough paste on it for that coverage? Why also would they give you guidelines to cut a channel in your IHS to measure temps in accordance with their documentation...if that IHS mattered any more than just the areas over the cores.

However, there are just too many variables in this whole shebang to consider 2* significant other than for bragging rights. I'm still running 35-40* cooler than Intel's max specified temperatures.

Let me put it another way. Cores 1/2 on my processor are currently running at 39. Core 3 is 36 and Core 4 is 38. Do you really think the way I applied my paste matters when the cores are on the SAME DIE? Under load, the cores are 62/62/54/54. Oh wait, no they're 58/58/55/55. Oh, now they're 63/63/56/56. Hm...I suspect this is just the natural variance in the 1) DTS on each core; 2) TIM between the IHS and cores...THAT I CAN DO NOTHING ABOUT!




RAID has nothing to do with this discussion other than a non-sequitor and a lame-ass personal attack. There are pros and cons to every RAID configuration. RAID0 provides performance increases but no data security. Use it if you want, but don't cry if you lose all your data...or have a reliable backup plan.

Is Intel trying to squeeze every possible ounce of on-the-edge performance / cooling out of the extremely low-clocked stock processors? No. Why did you even bring a stock cooler up into question? If you are using Intel's stock cooler design as a HSF design benchmark then we need to end this discussion right now. Good grief. There's so much headroom in they're stock proc and coolers it's ridiculous - by design. It's all about manufacturing costs. Somethin' tells me Intel will never ship a Thermalright Ultra-120 with a tube of MX-2 with their retail boxed processors.

At least your RAID0 comments were sound. I'll agree to disagree so this doesn't turn into another Crysis type thread (we all know the one I'm thinking of.)

 

[PliotronX]

I disagree with the rice sized dollup method. I have tried various methods well over 100 different times building PCs and have found that an extremely thin uniform spread across the entire IHS is a MUCH better method. Too many times using the single drop method have left the corners/edges naked. That doesn't make for proper heat conductivity. Even done various temp tests to back it up.

This is interesting, I think it is statistically remarkable because most TIM tests have declared winners over a single degree and I would think the esoteric watercooling folk at procooling would agree that is well within the margin of error, but the repeatability lends some credence to your experience. I am curious as to the status of your IHS samples and heatsink surfaces though, are any of them lapped? In spreading across the entire IHS, how much compound do you end up using? Do you prefer razors, credit cards, or finger in a baggy to spread?

 

[rainman]

Is Intel trying to squeeze every possible ounce of on-the-edge performance / cooling out of the extremely low-clocked stock processors? No. Why did you even bring a stock cooler up into question? If you are using Intel's stock cooler design as a HSF design benchmark then we need to end this discussion right now. Good grief. There's so much headroom in they're stock proc and coolers it's ridiculous - by design. It's all about manufacturing costs. Somethin' tells me Intel will never ship a Thermalright Ultra-120 with a tube of MX-2 with their retail boxed processors.

I'll agree to disagree.

However, I brought up the stock cooler because of the detail that Intel puts into the specification. There's a lot involved in cooler design, including how to "correctly" measure temperatures. I think all of us would benefit from a trial done within Intel's thermal measurement specifications. They provide specs on how to measure ambient, "core," and other temps...if we followed their guidelines to test everything from stock Intel designs to TRUEs to various water cooling rigs...and different methods of paste application...

Perhaps this argument would be easier to solve...as this would eliminate a few of the variables.

Personally, I tried both methods of paste application. There was no difference in temps in my setup.




However, I'm disappointed that you didn't take the opportunity to quote Intel on me:

All thermal interface materials should be sized and positioned on the heatsink base in a way that ensures the entire processor IHS area is covered. It is important to compensate for heatsink-to-processor attach positional alignment when selecting the proper thermal interface material size.

But then I could counter with:

The size of the contact area between the processor and the heatsink base will impact the thermal resistance. There is, however, a point of diminishing returns. Unrestrained incremental increases in thermal interface material area do not translate to a measurable improvement in thermal performance.

My argument is simply that we're at the point of diminishing returns. It is up to each person to decide how much work is too much work for the next degree cooler.