Coolers & TIM

Howdy y'all ...

I'm curious what Thermal Interface Material (TIM) others are using in conjunction with their Corsair coolers, specifically the Asetek coolers as I have an H100i v2. Is anyone using a liquid metal TIM (e.g. Coollaboratory Liquid Ultra)? What about any other TIMs? I'm currently using Noctua NT-H1 and it seems pretty decent but I'm also looking at replacing with Thermal Grizzly Kryonaut.

Funny you ask this... I had been using Kryonaut for the last 16 months or so and thinking I was getting better results than my prior favorites of NT-H1 and GELID-X. When I got my cooler back last week I was down to my last squeeze of Kryonaut and put it on. End of the tube is never great and I new I didn't do a good job. The temps were fine, but I wiped it off and took out a now 14 month old tube GELID-X. It was brand new and unopened. The CPU temp sensor was about 2-3C better on the same standard I always use. No surprise, I knew I did a poor job on the last one.

 

So I ordered more Kryonaut and two days later wiped off the GELID and put it back on. Really smooth and easy coming out of the tube and it was a perfect patch. Fire it up. Same temps as the botched Kryo job and 2-3C warmer than GELID-X. That made no sense, so after letting it ride for a few days I tested again with the same results. I then wiped it again and put Gelid back on. 2-3C cooler again. I went back and looked at all my test data for the last year and it was perfectly consistent with the results I was getting now at the CPU temp sensor on the lid. Core temps are always hard to pin down and the max values are more or less the same corrected for air temp, but the onboard CPU temp sensor on my 5930K was consistent at this same overclock and voltage I have maintained. I made an error in my initial assessment of the Kryonaut and in my mind it was better than what I was doing before It was not. At three times the price of GELID or NT-H1, I am done with it, although I have a brand new $29 tube to find something to do with.

 

I also used the Grizzly Conductonaut (Liquid Metal) for a brief period at the start of the Summer. It is still on my GPU chip/cold plate. Not much difference on the GPU where the coolant temp is always the controlling factor for final GPU diode temp. On the 5930K, it was 3-4C better than Kryonaut at the CPU temp sensor. However, that is only 1-2C better than GELID-X and it does come with a cost. I had the Conductonaut on for 30 days. In that time, it striped the finish off the cold plate and left a residue on my CPU lid that doesn't want to come off with normal cleaning methods. There is a warning about not using it with aluminium heat sinks, however despite what anyone says, not too many cold plates (or CPU heat spreaders) are made entirely 100% copper or any other single metal. It is going to be an alloy and there very well maybe nickel, aluminum, or something else that reacts with it. My 4 month old cooler looked like it had been in use for 5 years. I may have damaged any resale value for the 5930K and I am scared to take off the Titan and see what's there. Additionally, you really need to be super careful about using this stuff. If you have never used liquid metal anything before, imagine what happens when your break a thermometer with liquid mercury. Every drop breaks into a 1000 more drops, which break into a 1000 more, etc. You need to take what you are applying it to out of the case and prepare a genuine safe zone. After I was done and pulling back, the tiniest little drop came out the end of the syringe. It landed on my work table, but nicked the tip of a screwdriver. It split into hundreds of balls which raced off the table surface like tiny little insects. I doubt it will ever come out of the carpet. That was a droplet a fraction the size of a pinhead. That's a lot of risk for 1-2C. If you delid your processor, then I am more inclined to go with liquid metals. From an application standpoint, liquid metal is actually quite easy and precise. It's the safety concerns you need to watch because if you loose a drop inside your case, you are in serious trouble.

Yeah, I'm familiar with the dangers of LM TIM ... and I have delidded and used it for that (after applying liquid electrical tape to protect the other parts of the processor). It is definitely funky, funky stuff. The difference from delidding was ... everything that "they say it is". A good 20C at stock clocks. But I am reluctant to use it anywhere else.

Your experience with Kryonaut is interesting ... and certainly different from the current reviews. Questions: what did you use for cleaning the old TIM? And what application method did you use?

And yes, I'd chase after a 2-3C difference. My temps are currently pretty good ... max of 83C under extended 100% load (2 hour RealBench Stability Test) with the average around 76-77C and only brief spikes over 80 ... that's with a stable 5.1 Ghz overclock. I'd like to get those brief spikes under 80C. At stock clocks, it barely touches 70C under extended load.

I generally use a 90%+ isopropyl alcohol to clean off the old TIM. I have some other 'specialized stuff', but its milder and not any better. I use AIDA for temp evaluation and I can look over the line graph for 30 min chunks of time to really track variances (or the lack thereof). If I am strictly interested in heat capacity, I may also use Linpack with AIDA monitoring to record the same information as usual.

 

I was surprised with the results as well. I keep the room temps, fan speeds, testing conditions, atc. and the results from the past year were very linear. 22.1C room temp and 52-53C CPU temp. 24.3C room temp and 54-55C CPU temp. I had about 30 test sample to look back and compare, plus in all I ended up wiping and reinstalling the GELID and Kryo 3 times each on the current tubes. It was consistent. My take away from all of this is the thermal conductivity for the TIM is not the dominant performance factor. GELID-X is rated at 8.5 wm/k. Kryonaut at 12.5 wm/k and the Conductonaut liquid metal version at 73.5 wm/k. A 1-2C difference between 8.5 and 73.5 somewhat devalues the statistic. Also of note, I think GELID-X has changed their formula a bit. When I first started using it, it was a royal pain to spread and super gritty. I would heat the tube by placing it in a bag in hot water. Totally unnecessary now and it looks very much like NT-H1 and spread very easily. They do sell a Gelid Pro (or similar) that is much more viscous, but oddly with a lower thermal conductivity rating. Perhaps they renamed the old product.

 

I don't know how much any of this will help with spikes. Most often spikes are the result of internal conditions. A better TIM might reduce the average and consequently slightly lower the peak values, but not necessarily. My peak data from all of these are very similar and largely dominated by room temperature and voltage. There is no clear pattern with the various TIMs. Normally, I am more interested in peak values from a safety standpoint, but for evaluating conductivity of the TIM, the CPU temp sensor is better, at least on HW-E. Different CPUs might have better or less optimal sensor positions to measure surface temperature.

 

If you are interested, you might as well try it. I would think HW-E with higher cores and larger surface area would be a good test example, but ultimately each CPU and cooler combination is different. Since you are delided, it might be worth the time and trouble for the liquid metal, especially if you don't have resale plans for the CPU or cooler. I can't compare the Conductonaut to Cool Lab. My one experience with Cool Lab was several years ago and did not go very well. I had to scrap it and redo with normal TIM. The Conductonaut was quite easy to apply and I do think that has value.

So ... an update. When upgrading to a new case, I decided to go ahead and repaste from the Noctua to the Thermal Grizzly.

Initial results were ... disappointing. Over 90C while stress testing. Regularly hitting 40-45C on idle. :eek: Not what I was hoping for. Reset the BIOS to stock clocks as I didn't have time to clean and repaste at the time. At stock clock, temps were high but acceptable (~80C with RealBench)

I did check the mount; the block seemed solid so I stopped there. It turns out, that was a mistake. When I went to clean off the Thermal Grizzly and put the Noctua back on, I discovered that one of the standoffs wasn't fully screwed in to the backplate. So ... fully 1/4 of my CPU didn't look like it had contact with the plate. Certainly, this explains the temperatures! :bigeyes:

Screwed the standoff in tightly, quick repaste, remounted the CPU ... quick run of ROG RealBench. At stock clock, max temp of around 65C with RealBench. Reset to the 5.1 Ghz OC and rerun ... max temp of 81, which is a degree or two less than with the Noctua NT-H1.

I found applying the Thermal Grizzly according to directions to be frustrating, messy and time consuming. Dot/lentil method would definitely be preferred for me. It also seemed a bit thicker than the Noctua though warming it up did help with that quite a bit.

Yeah, I have gone through periods of trying to paint the lid. Depending the applicator it can come out like Vermeer one day and Jackson Pollock the next. I think most people are better with the dot or if you have a high core CPU a pre-smushed square in the middle.

 

Incidentally, I went another two rounds each with another tube of Kryo and the same Gelid X last week. Same results. One of my Kryo applications was flawless and it about measured even with the X-treme, but at triple the price why bother? I have done a full work up on NT-H1 vs the other two on my current hardware. 1-2C sounds about right - at most.

I have the H100i V2 on an 8700k cpu with a small overclock to 4.8 GHz and my temps have been 80C on Prime 95 with the thermal paste that came with the cooler. I decided to try Thermal Grizzly Kryonaut and my temperatures went up to 90C. I reapplied the Grizzly twice and got the same temps. First time I used a line and the second time I spread out the paste. I'm going to check to see if one of my standoffs got loose.

It's also possible that you applied too much of the paste. If you use the "recommended" application method, it's easy to put too much on. It should be very, very thin. I found that it helps to warm it up ... put it in a ziploc baggie and then in a glass of super-hot water from the tap for about 10-15 minutes. That'll warm up the paste and make it easier to spread.

Or you can use the ol' dot about the size of a lentil in the middle. Actually, for that 8700K, you might want to do a strip as the dies are more rectangular than square (as on the Skylake/Kaby Lake procs).