[Review] Indigo Xtreme vs. AS5, MX-2, IC Diamond, Shin-Etsu X23-7783D

[Church]

Vapor: you can also for one or two pastes (eg. some very thick or very fluid one) experiment with different aplying methods for effectiveness (no-no, not for all of pastes, as it will grow needed time/work for test to complete to non-acceptable ammount) - eg. using credit card or finger covered with some plastic bag to spread paste, or good-old single rice grain size blob in center, or one bigger in center and four smaller in IHS corners, or stripe of paste in center, or using few drops of acetone to make thick paste like TX3 more easily aplyable, or using some heatgun alike for same purpose, and whatever else popular method exists ). Hmm, and for this mentioned TX3 imho it would be interesting to test vendor's claims that it's effectiveness doesn't differ much from thickness of aplied layer. Pitty though that another claim i'd like to be tested/find out about takes too long (one about liquid ultra being less diffusing into surfaces then liquid pro), - i doubt required half a year-year will be good fit for reasonable test finish timeframe .

[Major]

I'll be lucky if I'm not doing TIM tests for the next 90 days, but I decided I want to leave as few unturned stones as possible. So here's the current list of what's going to be tested....

AM3:
MX-2, AS5, Indigo Xtreme, and potentially IceFusion (I have questions about its stability/consistency based on prelim tests). This test will be released first.

Then the long haul will set in on LGA1366, where I will be testing, in batches, the following TIMs:
AS5, AS Matrix, OCZ Freeze, Shin-Etsu X23-7783D, Shin-Etsu G751, Zalman STG1, Zalman STG2, MX-2, MX-3, Gelid GC-Extreme, Tuniq TX-3, ThermalRight ChillFactor 3, Feser H-Bridge, Scythe Thermal Elixer, Coolaboratory MetalPad, Coolaboratory Liquid Ultra, IC Diamond (yes, again...I'm giving it one more shot), Gelid GC-2 (potentially), CM IceFusion (potentially), Indigo Xtreme (hopefully)


Looking at between 20-24 tests there

I kinda can't wait to see all the data roll in

What an exhausting exercise!! hats off to you Vapor!

I'd like to see some weird stuff tested, like chewing gum, wax paper, hair jell, and finally some butter (non salted)

[Church]

I'd like to see some weird stuff tested, like chewing gum, wax paper, hair jell, and finally some butter (non salted)

I recall it being done already somewhere , IIRC even tooth paste was tested alongside butter

[Conumdrum]

Great dane poop was tested. Old link, can't find it.

[Nickel020]

Except that what they did there can't be called "testing" because the measurement error was so big.

Hats of to you Vapor, I know that I wouldn't have the dedication to pull through with this! And I second churchy concerning the application methods. I know we talked about this, but I'll say it again: I'm a big fan of the line method for testing because you can easily judge the amount of TIM used, while with the blob method you can never be sure how much TIM you really applied this time, the variation is much greater with blobs. Would be great if you could look at line vs. blob (and then settle for line ). I'd prefer that to a review of some of the more obscure pastes, and I think it will be quite useful for the readers as well, as the choice of the application method can be as important as the choice of the TIM itself (or close to it anyway).

[Church]

I recall on some other forum posted method about one blob in middle and 4 small in corners. Unfortunately i don't remember which paste was used, but that was the method that user got best evenly && thin spread paste and best temps. I don't recall what was bad with line method, but with single blob in center when he took off testmounted cooler to take pics of resulting aplied paste layer, paste on cpu's IHS && cooler's bottom wasn't spread to very corners of IHS too. It was spread there too if he aplied more paste, but that didn't make temps better (probably of too thick resulting layer in center). Line method showed similar temps to single blob, and only 5 blob method worked for him best, that's why i suggested that as worthwile to test aswell.

[Vapor]

I really don't see obscure stuff happening....in some of my prelim testing, a couple made-for-thermal-performance TIMs were already having a hard time coping with the heatload and I don't really feel like playing Pin the Tail On the Donkey trying to find a household item with acceptable performance that may or may not exist (for reference, I already did test a generic thermal pad and it completely failed...and a pure silicone oil was a near-failure, and it's a sibling component of most TIMs out there!).

As for application method, I'm also interested in what is best, but the hurdles to overcome for testing are significant. I reckon at least 15 mounts per method. Add onto that the individual differences between TIMs (a line may be best for one TIM, but almost definitely not for all), and you're looking at an undertaking that's best left to the individual manufacturers of each TIM, IMO. As such, I'm going to be following manufacturers instructions for each TIM. I personally don't think making a line is easier than making a blob in the center, I think they're all equal until you start adding steps, like spreading (I'm generally anti-spreading....though the painting from Zalman STG1 is pretty darn fun).

I'm already looking at 20-24 TIM tests at 2.5-3.5 days per TIM at 100% testing efficiency (I think I want to do 7 mounts per manually applied TIM and 5 for pre-measured like IX/LMP), at 12 hours per mount. Any hiccup (power outage, brown out, testbed issues, travel, whatever) will just make things even slower. Bare minimum I'm looking at 20*2.5 = 50 days of straight testing and at the highside I'm looking at 24*4 = 96 days of testing. Step back and think about just how long of a test this is going to be....

[rge]

IC7 site has a good explanation (with transparent mounts) on why many manufactures, that have extensively tested varying tim application methods in lab and outside lab, recommend using blob in middle....which definitely holds true for thicker tims like IC7, MX3, etc, that rely on high bulk thermal conductance. Using large enough pea size, and I measure mine each time, I get very good reproducibility and coverage over entire IHS and get consistently good mounts, often within 0.5C using 9 crystalfonz sensors. Using any other method, I get more erratic results, especially hand spreading (which introduces air vacuoles), and I never beat a good pea size mount with any other method (once the ideal pea size is worked out for particular tim).

If trying different methods though a) have to be reproducible b) still need to figure out how much to use, given high conductivity thick pastes perform differently than low interface resistance greases. Thicker pastes like mx3 and IC7 have low bulk thermal resistance (high bulk conductivity) but relatively higher (higher than grease anyways) interface resistance (IHS to tim and block to tim interfaces), ie they dont fill voids as well as grease, hence excel in independent testing with bondlines needed to fill gaps of 150 microns and higher (typical cpu/IHS), (bulk conductivity becomes more important than lower interface resistance in 150 microns or higher, interface resistance relatively more important in 50 microns lower bondlines). Silicone based grease tims like corning TC 5026 claims to beat all other pastes on market, even though has a much lower bulk conductivity by half, because it does have lowest (or did at one point) interface resistance. And test done (by corning granted), illustrates the point since they stacked the deck/cheated...they used two perfectly mated surfaces (not an IHS and waterblock) where they could use a bondline thickness of 20 microns and beat best performing shin-etsu. My guess is most IHS/cpu interfaces have variance that would preclude using 20 micron bondlines, more like 100+, pretty sure mine does looking at tim footprint.

Guess my point is, even if you found one IHS/block that worked ok with some other method, it would only apply to that one particular mate....and blob vs line vs many blobs vs spreading has already been exhaustively tested on multitude of cpu/ihs....so guess I would rather see same consistent method used ie blob or manufacturers...but then Im not doing the testing so not my decision

[masska]

I've never been able to figure out if you're Enerdyne or not I'm considering the offer, but I am very hesitant to accept because I am way overdue on the AM2/AM2+/AM3 kit testing

Anyway....running temp sensor validation my i7 930 right now and, lo and behold, at high temperatures (about 10C higher than I ever got with my 920), there's a different response curve Looks like I'm spending the next few days remapping the 930

I'm not an Enerdyne employee, however I' am the Australian/Asian distributor.
Let me know if we can help.

cheers

[Church]

As for application method, I'm also interested in what is best, but the hurdles to overcome for testing are significant. I reckon at least 15 mounts per method. Add onto that the individual differences between TIMs (a line may be best for one TIM, but almost definitely not for all), and you're looking at an undertaking that's best left to the individual manufacturers of each TIM, IMO

But how about if as i've said before - not for all TIMs, but just for two from different ends of scale of thickness/viscousity? (or choosing TIMs with less cure time to even furthier cut down test time?) (and less mounts per method. 3 mounts per each of two TIMs for each method are better then nothing + also maybe not all mentioned methods makes sense to test, it's just what i've summed up from reading various sources)
P.S.
Hmm, after seeing that IC webpage about aplication testing i started to love idea of aplicating with using see-through glass for quick glance of quality of aplied layer. Should be good for quick eliminating obviously wrong results with too much air bubles or too thick/thin layers, or wrong ammounts of used TIM without even testing temps ..

[Nickel020]

I personally don't think making a line is easier than making a blob in the center.[/URL]

I think I didn't make myself clear here. I'm not saying it's easier, I'm saying it's easier to get consistent results with the line method, and you're AS5 results point toward this as well, as they have a lower variation than those tests done with the blob method. It's simply not humanly possible to apply a consistent blob size without using some special syringe that applies the same amount every time*. A blob is a very complicated geometric body, as opposed to a line, which is simply defined by length and diameter. So with the blob you might be looking at +/- 20% (guesstimate) variation of TIM used between mounts, while with the line method you might be closer to +/-5%. This partly explains the variations in your first round test results imho.


*It might actually be a good idea to just use a small syringe with a scale on it similar to that that Liquid Ultra comes in. That way you can read of the scale the amount of TIM you just used for a blob

Edit: I'm also not saying that blob or line are better, I'll again second churchy in that testing exemplary TIMs with the line and blob method (MX2 for a thin viscosity, something else for a thick), but I think more than 3 mounts will be necessary. I realize how much of a constrain time is, but this should be considered and maybe done instead of testing some of the more obscure TIMs. I for one don't know of a single article with trustworthy data that compares line with blob method, so I think this would be very interesting.

[johnnymossville]

I really do wish there was a more consistent way to apply TIM. I think my instincts incorrectly tell me to use too much! That's why I'm liking the indigo xtreme method. I can only do it their way.

[Church]

..So with the blob you might be looking at +/- 20% (guesstimate) variation of TIM used between mounts, while with the line method you might be closer to +/-5%.

+/- 20% of TIM ammount used, not of resulting temps ..

I for one don't know of a single article with trustworthy data that compares line with blob method, so I think this would be very interesting.

Yup, that's why i want to push all that extra testing job to skinnie/vapor/martin aswell, because they are of that lesser bunch tests of whom i trust

[Nickel020]

Of course of TIM used, but no one here can honestly claim to know how much such a variation of TIM used actually affects temps. So we should either try to control the amount of TIM used more accurately or figure out how the amount of TIM used affects temps. The former is definitely easier, hence my argument for using the line method or a "controlled"/reproducable blob.

[Vapor]

Well the high leakage sample from AMD is a bust

Previous user lapped it and even worse, it's cooler running at max acceptable/stable settings than my 'frigid witch' 965C2 (169W vs. 162W). Going to go with my 1055T (187W at 3850/1.58V....going to run it at 3850/1.61V for this though, looking at just under 200W). Figure it's okay to bump to above 1.6v since I'm only testing 3-4 TIMs on this CPU, thanks to Nickel020's idea

Finished remapping/calibrating the thermal sensors of my i7 930 a few days ago and finishing up the remap/calibration of the 1055T today. Definite differences in behavior between them, and the 930 was giving much 'cleaner' data.

Tonight might be the first official mount of this TIM roundup

[Church]

Wish you luck in your work and eagerly anticipating results

[cleiomar]

I plan on using Indigo-Xtreme in my build once I get the rest of my loop. Would I be able to just turn the the pc off once the reflow happens instead of re-connecting the pump?

[Captain_Harlock]

No, you MUST reconnect the pump. Its part of getting a propper reflow. when the temps start to change the second time is when you turn the rig off and wait a few minutes for everything to cool down. Then you're done. Dont skip the pump step and just shut off the rig. The block needs to cool down quickly to help set the IX as its final part of the reflow and thats what reconnecting the pump does. Just follow the instructions to the letter, and dont try to adjust them. If the instructions dont seem clear, then by all means ask us about it (as the guys at IX may take a day or two to respond to any emails), and we will be glad to offer help. Just dont assume you can do without a step as you will end up wasting a $10 application of IX and will ahve to start all over again. Not worth it to save a few minutes.

[DarthBeavis]

After reading the reviews I ordered some new TIM (have used Ceramique for the last few years) but I am concerned because I want a TIM that will help me run cooler not hotter. The description at PPCs:
*** HOT *** IC Diamond 24 Carat Thermal Compound - 4.8 gram
*** HOT *** Indigo Xtreme for Intel Core i7 Processors (LGA1366 socket)

Did I order the wrong kind? is there a ***COLD *** kind I should look for? Or ***Holy crap this shiat is cold ***

[cstkl1]

After reading the reviews I ordered some new TIM (have used Ceramique for the last few years) but I am concerned because I want a TIM that will help me run cooler not hotter. The description at PPCs:
*** HOT *** IC Diamond 24 Carat Thermal Compound - 4.8 gram
*** HOT *** Indigo Xtreme for Intel Core i7 Processors (LGA1366 socket)

Did I order the wrong kind? is there a ***COLD *** kind I should look for? Or ***Holy crap this shiat is cold ***

lol bro.. ure funny

but this tim me loving it
its really made for i7
the hotter ure clocks
the bigger the difference
especially when ure running high uncore/vtt voltages
so far its arond 6c difference for me from mx3(still my fav for everything)
and gelid gc2

[DarthBeavis]

lol bro.. ure funny

but this tim me loving it
its really made for i7
the hotter ure clocks
the bigger the difference
especially when ure running high uncore/vtt voltages
so far its arond 6c difference for me from mx3(still my fav for everything)
and gelid gc2

hoping to do my 980x build next week or so (3-way SLI 480s and Asus Rampage III mobo with 12 gigs Corair Dominoator GT ram):
http://www.xtremesystems.org/forums/...d.php?t=245448

[the finisher]

hoping to do my 980x build next week or so (3-way SLI 480s and Asus Rampage III mobo with 12 gigs Corair Dominoator GT ram):
http://www.xtremesystems.org/forums/...d.php?t=245448

Can't wait to see this.


As always, eagerly awaiting Vapor's testing results.

[elmysterio]

I want to change thermal paste and not know what to bring. What do you recommend to put on NB, SB motherboard and CPU? AS5, Prolimatech PK-1, MX-3 and IC Diamond 24 Carat?

[Nickel020]

NB and SB doesn't really matter. For the CPU I'd use MX-3 (that's out of the ones you listed; I personally use Liquid Metal, since I don't care about surface blemishes on the CPU/cooler).

[Vapor]

From that group, I'd probably use MX-3 as well.

AMD tests are about half way done! All 7 IceFusion mounts are done, all 5 Indigo Xtreme mounts are done, 3.8 of 7 mounts of AS5 are done, with just 7 mounts of MX-2 remaining.

Difference in break-in curves of AS5 is probably the most interesting result so far. Not sure if it's the different bow of the CPU-360 block, the flatter IHS of the 1055T, the increased mounting pressure, or the center dot application method (vs. vertical line for i7), but the initial AS5 break-in curve is significantly flatter than it was previously.

Other interesting note to share so far is how differently AMD and Intel processors behave at thermal threshold and how it makes installing IX on an AMD CPU a little less confidence inspiring. Also of note, I botched my first IX mount (), it's amazing how much different the reflow process is using an all-metal CPU block than a delrin + copper base block. I restarted the pump too soon and only ended up getting IX over about half the IHS--temps were still good enough that I didn't realize anything was awry until after I took off the block though

I'm also considering adding IC Diamond to the AMD suite--I think I want a 5th TIM in the roundup, I think I have enough ICD for 14 mounts (7 AMD, 7 Intel), and I think it'll be interesting to see if it performs better than last time considering the large increase in mounting pressure.

Then it's onto the Intel testing