Liquid Metal and 33+ tape

I wanted to gather some information from experienced individuals who’ve used 33+ tape to surround their CPU die and the longevity of it.

It’s rated up to 105c and I have machines out in the wild since Haswell with a Zero percent failure rate so far. I’ve even taken some apart as some buyers didn’t want LM and everything looked like the day it was applied minus the die etching/staining that occurs.

Every once in awhile (including today) I get the viewer who is absolutely positive that this is a terrible method and wants to argue without looking for evidence to support their case.

So this post is for those that want their case supported regardless of what side you’re on.

 

Super33+ tape was amongst the first methods to institute insurance in your machine.

Now there are two-three more methods on the table that people like to endorse which all seem perfectly fine with some being better than others.

The top option is building yourself a foam dam around the cpu die, its a pain (like all things pursuant to performance) but its well worth the trouble for everyone that has used this method.

Next up is using transparent nail polish, a specific type though with a specific chemical that isnt reactive to Liquid Metal. Sorry I cant give more detail on that front. Anyways you put about 3 layers of the stuff around the CPU die iirc.

Next up is simple super glue.

Im sure others like @Falkentyne could give better detail though, iirc, he was much more educated on this subject than I.

 

Super 33+ tape and nail polish are usually used to insulate the exposed or SMD resistors or traces around CPU and GPU cores.
Super 33+ tape can also be used to help cover around the "gaps" in a LGA socket also, if applying LM on top of an IHS, but kneaded eraser is a much better option there. Foam dams can work also but it's much harder to get a foam dam to work around an un-delidded LGA socket, because you have the gaps and the retention mechanism in the way.

Foam dams are used (especially on GPUs and on BGA CPU's, and LGA with some extra work--Kneaded eraser is a better option there though) to both stop LM runoff onto the PCB or motherboard, and also to help insulate against oxygen oxidation of liquid metal (LM exposed to air greatly accelerates gallium absorption, even nickel plated copper, which has much lower absorption rate than bare copper).

Foam dams are essential on BGA and GPU's (where kneaded eraser won't really play well due to thickness and very low pressure heatsinks. And you don't want thick polyurethane foam either. Air conditioner foam works very well. Good dimensions for a sheet of foam are like : 24" x 13-1/2" x 1/4"
or something like this:
https://www.amazon.com/gp/product/B002GKC2US/ref=ppx_yo_dt_b_asin_title_o03_s00?ie=UTF8&psc=1
(1/4" thickness is the key, and around 20-30 ppi).

 

I just applied Thermal Grizzly Conductonaut aka "Liquid Metal" to my ThinkPad W540 a few days ago. I used a conformal coating which is somewhat toxic to breathe so it requires adequate ventilation when it's applied but it is the recommended solution when doing such things. I know there's tons of various kinds of tapes available, and some folks do use clear nail polish but conformal coating is specifically designed for this purpose so I'd never even consider recommending anything else.

First I cleaned off all the relevant parts (top of the i7-4930MX CPU core, top of the Nvidia Quadro K2100M GPU core, bottoms of the heatsinks as completely as possible with some isopropyl alcohol and a micro-fiber cloth - it's amazing how much previous thermal compound you'll find with a micro-fiber cloth compared to just a napkin or tissue or whatever, also highly recommended and you can wash/re-use the micro-fiber cloth anyway). I then took pieces of Scotch tape - seriously, and applied it to the tops of the CPU and GPU cores, then carefully cut off the excess so only the very tops, the flat shiny surfaces, were protected. Then I applied conformal coating, and this is what I purchased (this is not an affiliate link or anything, not attempting to make a buck from folks, it's just the direct product link):

https://www.amazon.com/MG-Chemicals-Silicone-Modified-Conformal/dp/B008O9YIV6/

The "Maximum constant service temperature of 200 °C [392 °F]" means this stuff isn't going to be bothered by a piddly ~100C heat buildup (if that much) from any modern CPUs so this is really the stuff to use for this coating-for-protection activity.

I applied one coating using the brush applicator that's part of the packaging of the conformal coating, slow easy strokes across the CPU and GPU dies. Gave it 15 minutes to dry and again adequate ventilation when doing this is an absolute must - I basically set my work space up next to a window, then I put a small desktop fan (high output, some Holmes "super fan" type thing) so the airflow blew across the motherboard and directly out the window during this whole process and I barely even smelled the conformal coating at all because of how I set things up.

I applied a second coat the same way, gave it 15 minutes, then figured why not so I applied a third coating and another 15 minutes. The thing about the ThinkPad W540 heatsink assembly is that is is a hybrid made of copper for the actual heatpipe part of the assembly but aluminum for the rest of it, and of course Liquid Metal is quite destructive to aluminum if it makes contact with it - it renders aluminum extremely brittle and there's no going back.

So, I then got out some more Scotch tape, applied two layers of it to the copper areas on the heatsink/pipe assembly, cut it to match the size of the CPU and GPU cores, and then I applied 3 coats of conformal coating to the entire surrounding area where the CPU and GPU cores would mate with the copper. I also blew out the fan assembly with compressed air - I had redone the cooling system only 2 months ago so there wasn't anything "stuck" to the fan blades, nothing caked on the heatsink fins, etc, it as all clean with a few quick blasts of air.

When I was satisfied the CPU and GPU dies and the copper areas on the heatsink/pipes were as protected as I'm going to really make them, I then set to work applying the Liquid Metal and of course I was deathly afraid of applying too much pressure to the syringe and having it spurt out like it did in Linus's video on his YouTube channel. Luckily I got it right on the first attempt for both the CPU and GPU and then also to the copper areas on the heatsink/pipes. Spread it out using the included swab applicators, no issues, and once I was finally happy with everything I did one final check of it all and then I reassembled the entire laptop and did some testing (of course I did some testing prior to this as well).

My original thermal compound was Arctic Silver 5, been using it for the past decade or so and never had reasons to complain about it, really. Running Prime95 for 10 minutes I got temps in the 95C range which of course is quite hot, and that's from an idle temp of about 41C. I knew Liquid Metal should provide me some decent drop in temps, and it did.

Upon booting up the W540 with Liquid Metal applied, I never even heard the fan spin up, from the time I pushed the power button to the time I was at the Windows 10 desktop (about 8 seconds 'cause of my Intel 5450s enterprise class SSD, obviously). I then set about firing up Prime95 for 10 minutes, using CoreTemp to keep an eye on the temps.

After 4 minutes I decided to stop the testing because the temps never went over 71C.

Am I a happy camper? Damned right I am.

I did more testing with wPrime to 1 million places, again, about 71C max temp. Idle temps? About 34C now, so again, a very happy camper. Played some Borderlands 2 which even in spite of being a ~7 year old game and not massively CPU or GPU dependent did tend to warm up this machine quite a bit with the AS5 on it, usually pushing about 80C with all defaults (no overclocking anything, stock clocks across the board, stock voltages, etc and I do use Intel XTU to modify things sometimes to keep the heat generation down a bit).

With Liquid Metal now in place, setting Borderlands 2 for my native 1920x1080 resolution and then literally maxing out every graphics option the game offers - and I do have the 4K texture packs installed (even in spite of me not having a 4K display the textures are still utilized and they do make it look much better overall) - I was still never seeing more than about 71C for temps across the 4 cores (+/- 3C either way since the 4 cores of this i7-4930MX aren't perfectly matched on the heatsink assembly, maybe lapping it someday could get it a few points lower).

Regardless, the Liquid Metal application plus the use of some of Thermal Grizzly's thermal pad material on the other pads of the heatsink/pipes assembly (on the GPU RAM chips and two or three other spots) now has this W540 running cool, and I hardly ever notice the fan when it's running compared to previously when it was quite noticeable. The Minus Pad 8 thermal pad material I used is here (again, a direct product link, no affiliate crap):

https://www.amazon.com/Thermal-Grizzly-Minus-High-Performance/dp/B00ZJSZ6WA/

Can't say anyone or everyone else might get such results, but I will say this: if you want some serious drops in temps, and you're willing to take your time and do things right and really do it carefully with the right materials, you can end up with some seriously great differences in the operating temps for even an older much hotter running setup like this i7-4930MX and the Quadro K2100M.

Pretty sure our computers would thank us for it if they could.

As always, YMMV, just be careful with the application of Liquid Metal - can't stress that enough - and if you do use the conformal coating, again, adequate ventilation during the application and drying periods, that stuff is toxic and highly flammable as well. Not trying to scare anyone off, but it's amazing to me how many YouTube videos use it in their thermal compound application and instructional videos but they never - not one of them that I've ever seen - warn viewers that it's a toxic substance and adequate ventilation must be taken into consideration.

Anyway, have fun, always...

 

You have to be careful with the tape and CPUs with IHS. If heat melts it, you may end up with IHS basically glued to heatsink. It happened to me when upgrading from 7700K to 8700K, it wouldn't move even a bit, I had to bend heatsink a little to reach socket lever and release it.

 

You have to use tape thats rated for high temps.

 

I prefer Kapton tape. It's designed for electronics and can even withstand the high temps of soldering. It sticks really well and removes easily. It's awesome stuff. It's slick, transparent and not stretchy like electrical tape.

I just came up with an idea, today in fact, that I believe will be superior to using a foam dam. I am testing it now on the MECH-15. I think it will be equally effective, easier and less likely to cause any contact interference like some experience trying to use foam that is too thick and/or dense.

Here is a picture to look at. I will explain below.



So, what I did was made myself some Kapton tape "straws" by rolling it lengthwise into a really narrow tube with the sticky side out. You'll want to use some narrow Kapton tape to keep the straw diameter small. You don't want it to be a large diameter straw, or thick with multiple layers wrapped around itself, as that could result in contact interference between the CPU/GPU and heat sink. My roll of Kaptop tape is about 1/2 inch wide, so it makes a very "skinny" straw. It collapses effortlessly and seals the air space. Space the straw far enough away from the die that it won't accidentally get in between the die and heat sink when it gets sandwiched into place. I just switched from Kryosnaut to Conductonaut today with these Kapton straw dams. Instant 10°C temperature improvement on the CPU and about 15°C on the GPU.


@Prema @Donald@HIDevolution @thattechgirl_viv @Falkentyne @Papusan @Ted@HIDevolution @Kevin@GenTechPC @jaybee83 @Robbo99999 @iunlock @Rage Set @XMG @B0B

 

And, the results...
GoW4 Benchmark - 125 FPS on Ultra/Max settings

GoW4 - Temps @ 30 minutes of gameplay (Conductonaut on CPU and GPU)

 

Liquid Metal FTW!

This stuff works as well as Conductonaut, spreads easier than Conductonaut, (doesn't try to ball up as much,) and it is less expensive. I have ordered another tube of it so I can have some extra on hand. Thermalright Liquid Metal Thermal Paste, 79 W/mK High Performance, Silver King Heatsink Paste

For the skeptics, a picture says more than 1,000 words.

Kryosnaut versus Thermalright Liquid Metal Paste ( Bare Die @ 5-Minute Stress Test)



Pro Tip: Brother @Falkentyne recommended these for liquid metal application, and they are AWESOME. They work fantastic! Much better than brushes or cotton swabs. They also work nice for cleaning up old paste and reaching into tight places on the PC to remove dust. Buy some... you'll be glad you did.
100 PCS Disposable Lip Brushes Lip Gloss Applicators

 

So same performance but easier to spread and cheaper? Would you still recommend the standard run off protection against liquid metal or is it on the level of an arctic silver? If it offers the viscosity of a standard TIM that's a major plus.

 

On a laptop or a vertically mounted desktop motherboard/GPU, you should still use run-off protection of some sort, as the viscosity is the same as other liquid metal. It is not a paste consistency. It is easier to spread simply because it seems to adhere to the surface(s) much better. You don't have to scrub or massage it into the surface a lot to make it stick. It applies more like paint. I'm not sure why that is, but I am certainly not going to complain about that, LOL.

If you remember, the old Liquid Ultra use to go on that way, then Coollaboratory changed something. Now Liquid Ultra and, to a lesser extent, Conductonaut have something different in their formula that almost seems to repel the surface you are trying to apply it to. This goes on and sticks nicely, like the old Liquid Ultra used to.

Run-off protection...

 

Coollaboratory made a new liquid metal last year with more paste-like texture for easier application.
Coollaboratory Liquid Extreme

I'm sure they screwed up their own recipe/formula to be more equal Conductonaut (Conductonaut beat them in a couple of tests). Then a few year later make a new formula equal old Liquid ultra and brand it with the "Extreme" surname to be able to charge more (+5%) for the barely higher 3 W/mK or the easier application. None should change their own winning horse (for Coollaboratory, screw up the old working Liquid Ultra recipe)

I don't trust Coollaboratory anymore. And none should trust the W/mK numbers floating around from the different brands.

 

I agree. I stopped buying Coollaboratory products several years ago. They violated the cardinal rule. "If it ain't broke, don't fix it." It seems to be a popular concept to ruin excellence instead of leaving it alone. Superior products don't stay superior when you start screwing with them. Stupidity is abundant. We are completely surrounded by retards in the tech world.

 

@Mr. Fox after a few months did the straws hold? Planning on repasting myself tomorrow and I wanted to try your method.

Can't easily find of 33+ or Kapton here, would TESA tape also work: https://www.tesa.com/en/consumer/tesa-insulating-tape-electrical-pvc-tape.html

 

That kind of tape might be too thick. You'd have to try it and find out.

I haven't taken the MECH-15 apart to inspect the Kapton straws. I've taken it on numerous trips and start another one tomorrow. I have no reason to suspect it is not working as intended.

That said, there is nothing wrong with using foam, as long as it is thin and compressible enough. The foam insert from a cheap 120mm fan filter like the one shown below should work well.

 

The tape is indeed quite thick, but it works, at least for the moment.

 

Great. Looks like it wasn't "too thick" so you should be good to go. If it was going to be an issue it would have been obvious from the start. I don't think you have any need to worry based on the nice temps you are seeing.

 

Is that heat resistant? They melt and burn easily in my experience...

I reckon you really don't want that to happen under the heatsink D:

 

Tesa isolating tape is heat and even flame resistant. But let's be clear if you apply LM en your temps >90 degrees You're Doing It Wrong ;-)

 

OK, I took it apart today and the straws had held up fine. The heat sink was stuck in place pretty well--more than I anticipated--and had to be carefully pried off (to avoid bending it). I found none of the liquid metal had spilled out anywhere and the Kapton tape straws had done exactly what I intended for them to do.

I put it back together using foam dams cut from the air filter element (shown in previous post) and that works nicely. I ran a Time Spy Stress Test (full 20 passes) and the temps leveled out at 80°C max on CPU and GPU. That was with no GPU undervolt. The CPU and GPU were within about 2°C of one another, with the CPU and GPU constantly trading spaces at being the hotter/colder part.

I took the motherboard out to investigate why the top panel between the display hinges gets so stinking hot all of the time and the only thing I can see to explain it is the CPU, GPU and GPU memory, and the VRMs for both (all of which generate heat) are just too near the chassis to not make it hot, and no air movement to evacuate the heat trapped in that very narrow air space. The other reason it gets hot is that it is metal. Plastic would get equally hot, but wouldn't feel so hot to the touch as metal. Plastic and metal have their pros and cons, but metal still wins in my book.

That hot chassis thing kind of sucks, but the alternative is that the heat stays on the silicon and that is definitely not a good option. Better for the chassis to take the heat than leave it on the chips. To facilitate using the chassis as a heat sink, I added thermal pads to serve as the conduit. That DEFINITELY works. The chassis gets A WHOLE LOT hotter now and the CPU and GPU both run cooler now, LOL. 0.5mm pads (just some old used ones I had in my junk box) fit perfectly. I placed the pads directly on foil cover on the opposite side of the CPU die, GPU die and GPU memory.



@B0B

 

I used the thin foam from a 120mm fan filter for liquid metal dams as I previously mentioned when I added the thermal pad between the mobo aluminum foil shield and chassis back on March 1, 2020. Yesterday I took apart the MECH-15 to see what size resistors I needed to shunt mod the 2060 and found the foam dams were in stellar condition, doing exactly what I intended them to. I reassembled it using the same foam dams. They were in great shape and totally reusable.

@Falkentyne @Papusan @B0B

 

Is that a 7+1 phase vrm for the 2060? Damn.

 

These resistors certainly are tiny little critters. They look exactly like those on the 1080 Ti and 2080 Ti, but they're about 1/3 the size of them. I ordered 8 ohm through-hole resistors from Digi-Key so I can tack them to each end instead of stacking them. Should be easier to deal with considering how small they are.

 

Can you take a clear pic of a mosfets and also the entire gpu power delivery? I'd like to take a closer look at what Eluktronics's implementation is.

 

I will try to get a better image when I take it apart again. I only have my LG V30 phone camera. I'll take it out in the Phoenix sun next time. That may help with the clarity.

I was all ready to go with the shunt mod today until I discovered the extra resistors I had from the 1080 Ti and 2080 Ti shunt mods were massive in comparison. I think one of those resistors would cover up both of these, LOL.

 

Yeah for sure. As far as I know only mxm gpus use desktop size shunt resistors. On laptops they just use flat ones. Didn't even know there were miniature versions of the desktop shunt resistors myself.

Looking at the picture again there seems to be another pair of R010 resistors near the battery pins. That might be a separate rail that feeds extra power to the gpu. I'll need a clear high resolution pic to know for sure.

 

Yes there are. The PCB is labeled RCH1 and RCH2 for those two R010 resistors next to the battery connector. The two near the GPU capacitors are RCH5 and RCH6.

 

Yup, they look identical in design, just much smaller.
Here is a look at one of them on the 2080 Ti.

I did a shunt mod on the 2080 Ti using through-hole resistors, but they were 10 ohm and didn't work. Had to drop down to 8 ohm. But, they were easier to tack in place. I just tinned the wire first and it was super easy to attach (and remove). Only had to touch it with the soldering iron like 1 second on each end.


I will use these two Fire Strike stress tests from today for comparison after the shunt mod. With Eluktroboost it already holds boost clocks better than most competing laptops, but I'd like to see it at least hold 2100MHz sustained under load. Temps are pretty decent as well. I hope that allows it to hold higher core voltage as it did with 1080 Ti and 2080 Ti. I really hate Turing and Pascal firmware algorithms. First one was with "Balanced" power plan. Second was using "High performance" power plan.

 

ah, I wish i thought of that. keen to see how easy/tidy it turns out. i might redo mine if i ever need to open mine up again, its rather ugly at the moment .

max temps on your gpu show as 79c. probably not much more wriggle room on voltage? still benefits to be had from the card not dropping clocks due to short spikes n power limit throttling though.

 

This should be kind of interesting to play with. Don't have the resistors yet, but this was delivered today. This is an old resistor I took off of a shunt-modded GPU before selling it a couple of years ago. The image is a little blurry because it is still in the plastic bag, and the bag isn't perfectly clear.

A couple of shots of a part on my desktop mobo. The close-up is with the camera lens about 1/2 inch away from the part.

 

Yeah one of the many downsides to the popular crappy approach to laptop design is a unified heat sink. Having the CPU and GPU taking a dump on one another is a really stupid way of trying to manage things. That said, it was a stress test of 20 loops of Fire Strike so it is not necessarily exemplary of a normal use scenario. With such a mickey mouse thermal solution it is a wonder that it wasn't 99°C. I also find it interesting that the cooling on this product is better than average in its class, yet people still choose to buy inferior competing products.