Bitcoin Forum

For reference, the acrylic block is 4" x 4"

https://i.imgur.com/aZSnf.jpg

https://i.imgur.com/tBDF0.jpg

https://i.imgur.com/kygrm.jpg

https://i.imgur.com/ZYGMs.jpg

https://i.imgur.com/UkEVe.jpg

This is so true. Nice way to demonstrate it.

Tell that to MSI.  I had to clean off the TIM on all my MSI 7xxx cards because they put way too much.  Plus I like MX-2 and had a bunch left over.   ;D

This or a similar thread was posted last year.  Not that I disagree with what you see here, but it doesn't tell the whole story... the effects of thermal heating and cooling will change the shape, thickness and nature of the thermal paste spread you see under that acrylic.

So this test is kind of meaningless in a real world scenario.

I have the same paste btw. Got it with my Cooler Master Hyper 212+ cooler.

Ditto. It's not AS, but very good stuff for just everyday use. Love the cooler, too, btw.

Correct, but having less of a blanket of TIM insulating the chip = good.

I don't see the problem. Unless your thermal paste is as thick as modelling clay, I don't see how you can over-apply. Any excess will simply ooze out of the side, and thermal paste shouldn't be thick enough to prevent this, especially with as much pressure as most heatsinks apply via tight screws and springs. For my GPU cores, I've always done an X with a dot in each "quadrant" with my shin-etsu x-23-7921-5 thermal paste, and never had a problem. My temperatures are always lower than what they were originally. IMO, it's better to have full contact and wasted thermal paste than have partial contact and wasted surface area.

Just don't spread it, then it bubbles everywhere.
Unsure what happens when heat is applied post bubbling - so that could just be a moot point.

Heat that acrylic block up to 80C !!! ;P

That too. I refuse to spread because it can make air pockets, which is extremely poor conductor of heat energy. Also not all heatsink are equal. It may be concave or convex, or have minor warping that you can't detect. Spreading thermal paste thin will leave huge air pockets in your heatsink defects. Let your thermal paste do it's job and spread on it's own.

Most heat paste does NOT conduct electricity. If it did I would have lost many gpus or cpus / boards already.

UGH .. I *HATE* TIM pads. also can't stand them on northbridge heatsinks. I've had it happen when I'm installing a new motherboard and might knock the northbridge heatsink that it pops off. I have taken to simply redoing the northbridge heatsink with arctic silver 5 (not the thermal adhesive though) since it's just the right kind of sticky. Just the regular thermal compound seems to hold much better than those awful pads.

Temps on CPUs and graphics cards have always been able to go lower by redoing the heatsinks properly. I'm not sure why manufacturer insist on crappy TIM pads. I'd pay a few bucks more if the heatsink was factory installed with AS5 instead.

I really despise the big sheet of TIM that comes with (especially Koolance) water blocks.  Not only do I not have a secondary way of ensuring contact because there are no mounting attachments for RAM, VRMs, and such, I have to cut the stupid crap myself.

I do like the very thin adhesive on the Swiftech RAM heatsinks (which I never use for RAM anyway) but I wouldn't use TIM on anything performance critical.

FYI, EK does the same thing.  Not a big deal for me and I've done quite a few waterblocks.  The thing that drove me nuts though, was when EK didn't use spacers etched in the waterblock.  You had to get a small amount of TIM and use that to stick to the block and hope that it doesn't move when you attach it to your card.  Thankfully they learned with their latest series (HD 7xxx & GTX 6xx) and applied the spacers within the block.

They'ed rather cover their ass and put too much instead of not enough...

I agree with what you're saying theoretically, but the BFL FPGA singles I have all had way too much paste on them. I even had one half it's hashrate because thermal paste dripped onto some transistor knocking out one of the FPGA chips. I wiped all the paste off and applied a very small amount of paste to the chip and it's been running fine for about two months now. Doing the same to my other units has given an overall benefit in throttling reduction (I started doing it because of the prior thread). This stuff is not meant to be slathered on, and doing so hinders the conduction.

I used to have a pic of the thermal paste on the transistor. Not sure where it went.

True, but as ssateneth pointed out above, the final thickness of the paste has nothing (or not much) to do with the amount applied.  I think air pockets or insufficient amount of paste are more important problems.

What color was the paste?  I have been trying to educate the assemblers on proper application of paste and we switched to Arctic Silver MX-2 from the crap we had been using before.  I'm wondering if it predated the change.  If not, I may need to go talk to them again about application technique.

This is what I use but its pricey

http://www.coollaboratory.com/en/products/liquid-pro/

http://www.vortez.net/articles_pages/coollaboratory_liquid_metal_pro_thermal_compound_review,6.html

I haven't seen that before, how well does it work compared to MX-2?  Can I get some in the US?  I would love to test it out... I like Shin-Etsu but it's way too hard to apply for the assembly folks - is this stuff harder to apply than a basic paste?

I like it, I happened to find it on frozen pc about a year ago, no have I done any testing to see what temps do. Before purchasing I did reviews on it and it done better then the artic silver 5 in most reviews and now it seems there is a paste application also. im sure if you need a large quatity they would sale direct there are not many distributers here it seems.

http://www.vortez.net/articles_pages/coollaboratory_liquid_metal_pro_thermal_compound_review,6.html

a review I found

Never heard of the stuff, but it's def interesting! But it sounds like it's more of a pain than it's worth. Removing the "paste" requires sanding down both the chip, and the heatsink.

MX2 is pretty difficult to apply properly. Its almost like chewing gum. I would suggest you switch to MX4 which is (marginally) more effective, but more importantly, very easy too apply,.

Yes you have to sand the surfaces, but that would help to keep a proprietary aspect to the equipment, if the unit is rma'd you can tell if someone tried to make mods on thier own. Which would of course void the warranty making bfl not responsible for said damage. Just make sure your heat sink has a copper base, it will corrode aluminum. You can have copper base going into aluminum fins and it's ok, but on high heat the reviews speak for them self.

I don't use liquid metal products anymore. I used some with my CPU and waterblock and when it came time to dismount it to sell off and upgrade, the waterblock and cpu became nearly fused. The stuff completely dried up and it was EXTREMELY difficult to pry it apart. The waterblock (fully copper) is ruined now, and the CPU had to be sanded and lapped to be in resaleable condition for use in air cooling. I'll take a picture when I get home from work.

I am sure that is why they make the sealed pads now: http://www.frozencpu.com/products/5509/thr-35/Coollaboratory_Liquid_MetalPad_Thermal_Interface_CPU_Pad_501249.html?tl=c127s451b74 (http://www.frozencpu.com/products/5509/thr-35/Coollaboratory_Liquid_MetalPad_Thermal_Interface_CPU_Pad_501249.html?tl=c127s451b74)

According to the MSDS it's an alloy of indium, copper, and bismuth that melts at 59 C. Not surprised that it fused with the water block made of copper! 

I'm getting better results on gpu's and singles with thicker pastes. So long as it's not pouring at the sides, your not using to much.

In my case, it was white.

https://i.imgur.com/mfRAn.jpg

When reading my message, keep in mind I may be remembering my college courses wrong...

That said...  The purpose of thermal compound is to fill in the microscopic gaps between two metal surfaces, maximizing thermal conductivity.  Any more than enough to fill in those gaps and it effectively becomes an insulator.

What I recall from my school, is how to apply thermal compound to the CPU, so that way the compound more effectively fills the gaps.  If I recall properly, just putting some compund in the center and installing the heatsink isn't the most effective method (it doesn't fill in those gaps as well).  But to take a plastic bag over your hand, and smear the compound on the surface of both the CPU and heatsink. Any air that gets trapped when installing the heatsink should be forced out after so much time when the CPU gets hot.

I been applying thermal compound this way for years and tends to save me a lot of compound and usually shows 5-10C cooler themps than factory application.

Have I been doing it wrong all these years or do I need to go back to school??

I got my information from the Artic Silver site and the answer is: it depends on the chip:
http://www.arcticsilver.com/intel_application_method.html#

but a lot of chips now do not use a "middle dot" any more but instead use a line or multiple lines.

However, if you don't know a lot of information about the chip, your method (labeled "surface spread") works fine, and is the method I used. One thing is that the application on the heatsink side is really thin, almost too thin to consider to be applied. It just greys the area. For the singles, I actually put dots of the arctic silver on the corners of the chips and dry press on the heat sink and then remove it to get some markings for the tinting. Then I color it in the lines.

As far as I understand for those documents and others, you are right about the air gaps being removed by heat after use. That's why there's a burn in time.

+1, its what I use now (was MX2) and what I put on my single. made a 2-3 C difference. the original paste was gray and very thick (almost clumped) BTW.

the springs on a singles heatsink are pretty weak, so a thinner paste should be better. or at least easier to apply correctly.

I've been trying to do away with the springs and go with a nut and bolt system with loctite. I absolutely detest that spring system on the current HSF.

If we are talking about bfl asic cooling design, best to just use copper rather than worry about paste too much? (I know it's pricey) Shin etsu ftw.

Also, vote spring hsf design. Nuts and bolts are not a long term 24/7 solution (although they sound like one).

It's kind of immaterial now, though... the ASICs use a totally different HSF, so the problems should not reoccur anyway.

We've actually ordered some of that paste/sheets above to test out.

Chewing gum?  I don't why yours is that thick.  I use MX-2 all the time (prefer it over AS5 which is thicker).  I wouldn't mind trying MX-4 if it weren't 2-3x more expensive.

Yes, MX-2 has always been... about the consistency of taffy. First thing I had to do was get that junk off my BFLs and apply AS5.

I mailed Butterfly some size 4 nylon nuts and bolts. They said over-tightening could be a issue. I don't really think that can happen with nylon though. The threads don't hold if you over tighten. Definitely nicer then those springs. They seem to be holding fine, I didn't even use loctite.

I might have been confusing MX2 with MX3. Have a look here, check the video:
http://www.hardwareheaven.com/reviews/986/pg2/arctic-mx-4-thermal-paste-review-thermal-compound.html

I have a big bag (1500 ct) of nylon nuts and bolts that we are going to try out, Cuz.