| 80-way Thermal Interface Material Performance Test |  |
| Reviews - Featured Reviews: Cooling | |||||
| Written by Olin Coles | |||||
| Sunday, 14 June 2009 | |||||
|
Page 14 of 14
Final ThoughtsThis article closely mirrors the classic tortoise and hare analogy: two completely different methods of reaching the same destination, and the fastest (best performing) product isn't always the winner. As I've reiterated over an over again in the article, no thermal paste will offer better results than a properly prepared processor and cooler surface with mounting hardware that creates very high contact pressure. To the average hardware enthusiasts, this doesn't mean very much, since they don't usually want to spend the hours/days/weeks of time lapping, polishing, buffing, and testing the various aspects of their tweak to achieve the best performance. In a classic Hare fashion, they simply want to buy the product and enjoy the same results as earned by the Tortoise. Overclockers know: this isn't a foot race, and slow, steady, deliberate preparations and testing are the only way to obtain dramatic performance improvements. What began as a huge mind-boggling collection of TIM products has been refined into a test-proven selection of premium Thermal Interface Material. In retrospect though, the testing methods were not perfect, and certain aspects were unnecessary. The reasons are simple: if you want to truly 'test' thermal paste, you need a controlled mechanism for deriving results and eliminate influencing factors such as pressure and finish. There are plenty of these methods available, such as Hot-Wire, Axial Flow, and Hot-Plate methods of determining thermal conductivity, but delivering results expressed as W/ m·deg K would be lost on most readers (who might also criticize a test foreign to them). I have often thought that the many months of work have amounted to only a shadow of what really should have been; but unfortunately most readers would also reject my own testing theory, which includes a ceramic ring filled with a thick but consistent amount of material. It occurs to me that if we want to test how much thermal energy can be transferred through any particular material, we should be using a thick amount with basic compression trapped by a non-conductive barrier, and not base our results on nearly immeasurable thermal paste film resulting from good compression. Scientifically sound thermal testing and real-world performance are two very different things, but it's difficult to convince readers otherwise. Many years ago, back when Intel Pentium 4 processors were heating homes across America, the average computer system processor would require extreme measures to keep it stable at its base speed. Even now as we experience AMD Phenom II and Intel Core i7 processors that can run warm even without activities such as overclocking or adding additional voltage, they usually operate considerably cool with the stock manufacturer-supplied cooler. This is proof-evidence that CPU cooling has reached a peak, and the need for aftermarket cooling on a stock computer system is, well, cooling. Overclockers are a different breed, however, and the end result of more-efficient processors translates into more stable overclocks. So what comes next? With this article Benchmark Reviews will likely discontinue large-scale TIM comparisons, since there's really no purpose in comparing unavailable or discontinued items. At the same time, we will search out a better method of comparing one material against another, since replicating user experience doesn't generate enough disparity between materials. Armed with the information we've provided in this 80-way thermal material comparison, you can apply the tips we delivered in our Best Thermal Paste Application Methods article, and then begin your own project with our Introduction to Overclocking Guide. 80 Thermal Pastes ComparedBenchmark Reviews does not endorse any particular manufacturer or product, regardless of our test results. Our official position on thermal interface material is very clear: every product tested will transfer thermal energy beyond the needs of average computer users. If you do not overclock your computer system and push hardware to the breaking point, any thermal paste featured in this article will suffice without penalty. However, if you're not just looking for something to fill space between the processor and cooler, and every degree of performance counts, then perhaps you'll be interested in our suggestions to help you attain the overclock you're burning to reach. Listed in order of enthusiast performance results, Benchmark Reviews offers the following products for consideration:
If you have any questions or comments regarding this article, Benchmark Reviews really wants your feedback. We invite you to leave your remarks in our Discussion Forum. It's the only way we can meet your expectations. Thank youBenchmark Reviews recognizes that this project would not have been possible without third-party support, and we sincerely thank the following companies for their assistance:
Related Articles: |
|||||
Comments
This is one of the best review ever consider the amount of writing and scrutiny the writer had been through! I cannot give you enough compliment!
Makes me want to know more detail about how to prepare/polish the heatsink surface as it seems really important. Like exactly how you go about it and exactly what materials are used? I imagine you could make it worse rather than better by not doing it well?
Brilliant article, thanks again!
Haven't even seen the result charts yet, but just having the pictures of different ways to apply the paste is hugely helpful.
Thanks again!
Great article. Thanks for all the work.
If you're looking for a follow-up article - investigate why the machine OEMs use such a thick layer of ITM in their boxes when it's blazingly obvious that a thin layer works much better.
I would expect that even the ones that are without a curing recommendation would do better. That would be a great comparison.
Still, great review. It does however tell us that near the top, we are probably chasing diminishing returns here. I suspect that the 0.5 degrees could be within a margin of error.
found out on my test that any paste that require spread method. will give max/best result right away but over time the paste bleeds out of place, due to air bubbles. therefore decreasing performance.
i learned so much, and not just about TIM
Now I'm dying to see how the new ones like MX-3 go in the next test.
Now I just use a bit more TIM. I didn't mind lapping my old E4500 as there wasn't much at stake but I didn't lap my (current) O/Ced QX9560.
Now I just use a bit more TIM. I didn't mind lapping my old E4500 as there wasn't much at stake but I didn't lap my (current) O/Ced QX9650.
I am not talking about "Shin-Etsu MicroSi G751", I am talking about "Shin-Etsu MicroSi G765"
I happened to run out of Arctic Silver one night while assembling a build for a customer and I had a small tube of Radio Shacks white Tim. I used it on a a few builds and was surprised at the performance. The only thing that was worth mentioning stating is that it does dry out after 3-4 years and needs to be replaced IMHO.
Anyway, great review, later........................................
It would be grate if each product's electric conductivity was listed in the charts at the final results.
There is a slew of information in this article and it's definitely going into my bookmarks. Great job and many thanks for the extensive effort that was put into this project!
Helped me alot.
Or even a dumb (non-contaminating, non dangerous, cheap and available) poor mans product like toothpaste, or ??
Or a test showing the best product with poor application technique?
Seems like with all the work, a few more comparisons made to show the importance (or lack) of this.
Was repeatability also ensured for accuracy of claimed results? (eg, 30 tests later, do the same test with XYZ and ensure same results obtained)
By the way I had a ghetto thought about this If Money is no problem why not weld the heatsink base and CPU IHS...
Will love if this is done..
Sincerely,
/Antonio
You did a hell of a job
Keep the marvelous work
Funny how a 2 or 3 hour a day PC user could eyeing up the next upgrade path before as5 has cured. Any way I'm off to lap my sink
Try reading the article.
Must say though, that adding the viscosities of each thermal goop was a brilliant move as it's hard to find any real info on them elsewhere.
I would love to see some of the newer TIMs tested and added in a section at the end of this article as an addendum. I have been very interested in the metal TIMs I have seen around. Some people claim they are getting a 10c difference using a metal TIM over a TIM like AS5. I've never tried a metal TIM myself, and although I hear they are a little bit of a pain to clean off I would still like to try one.
As it stands this is still the best article around on TIMs, but it could use a little bit of an update.
Saying they nearly all perform the same may have been true 2-3 years ago, but not so much now. I would sure like to put some of those claims to the test, but since I can't get TIMs like Liquid Ultra locally I can't test them myself.
Frankly, until they are, no one can make any claims regarding them and be taken seriously, and that includes the naysayers. I have seen these claims on ocn. I can't counter them because I haven't tested these new TIMs.
Have you?
Then there are the people this article is aimed at, the enthusiasts who care that the difference between the poorer TIMs are better TIMs can be 5c. When you're overclocking, and aiming for a particular clock rate, then 5c can be pretty important.
Your question is akin to comparing the fuel in a street car to that in a race car where a few octane can make a difference.
AND when I read all of the comments that were written thanking you and praising you for your all of your hard well thought out work, Nice elements table too, Then there were the "when are you going to do a new one" Persons. I have to just shake my head sometimes in wonder at the different densitieses that are exhibiteded in such exchanges.
I have worked in metallurgicaliFielde ild much of my life and pretty well versed in such matters, had to laugh at the suggestion of soldering the processor to the cooler, I wonder do people actually take the chance on refinishing the surface of the processor? I can see flat lapping the cooler but I would never touch the processor, except to install.
Thank you for all of you efforts and I enjoyed all of it and agree with all of your findings.
Bay Blues
P.S I also put the antec 6 on my 5670 card voided the warrantee, but I more than likely extended its life its getting 32c vs 40c plus before.
When you put up 80 products agains each other the possibility of favoritism practically vanishes and the real picture begins to emerge.
I do agree with several here who have expressed wishes this test included results after long term use and curing of the TIM application(s). At the same time I must understand the time and cost required to do such an in depth test with so many products. All in all considered this is a great trade off.
In Short, I didn't think it mattered. It does! Buy it!
One thing I think could have really driven your important point home would have been to add a short table at the end, showing the "Before and After lapping" temp results using an A+ and a D- (perhaps thin and thick examples of each) on the illustrated Ultra-120 eXtreme heat sink, thus demonstrating the benefit obtained from prep, and the much lower importance of the paste used.
Also, as someone who laps metal pieces flat and polishes them to a mirror finish in my daily work, checking with an optical flat/monochromatic light, I'd like to emphasize to the masses that a FLAT surface is not the same as a SHINY one. Obviously, flat AND shiny is what we're after, but it's important not to "unflat" a surface in the pursuit of "ultrashiny", which is easy to do if not performed properly.
Finally, in answer to a question asked upstream to the effect of, "why would anyone not overclocking worry about this stuff?"-- those of us who are in quiet environments, particularly audio, like to keep our stock-volted machines as quiet as possible, which is partially accomplished by improving heat dissipation thus enabling lower, quieter, fan speeds.
Thanks again for the insane number of hours this (and previous articles) must have taken, including the excellent presentation.