Been running a EVGA 1080Ti with LM for nearly 2 years with a custom loop. Flashed the bios with a KP version to not have the voltage restrictions. It achieve a 10C benefit at my max core overclock of 2105 MHz. I’d recommend any enthusiast to try it, you just need to research the knowledge and use some care and common sense when applying. The benefits while minor are there. BZ has done some great past videos on how to handle LM properly. Ironic that the viewers who sent in these cards hadn’t watched them.
Well BZ, no, let me explain: It's all about the amount of heat transferred per area. I had a 1080ti gaming X which pulled around 330W at 117% power. The cooler was phenomenal I thought, then I started noticing how quick the temps were rising. I fiddled with the soft temp limit and saw that the cooling wasn't actually adequate and the card was soft throttling. I repasted with MX-4. The temps were better but only a couple degrees. I had LM on hand (Thermal grizzly stuff) which I used for my 7700K delid(s). I covered the smd caps with nail polish, applied just enough and I buffed it in, that is really important. The card afterwards would stop at 69 degrees celcius under the same fan settings, I lost like two degrees in a year and reapplied. I used it like that for about two years and I just sold it, of course disclosing everything and letting the buyer actually try it beforehand. The buyer was skeptical at first but later he was like "looks like this LM stuff really works". I sold him an unopened 1g pack of conductuonaut along with the 1080ti. The reason LM sticks to tin is Ga-Sn amalgamation btw, not some sort of Sn-Sn debauchery. Ga amalgamates with tin very much. It amalgamates with Cu and Ni too btw, it's just that those two are really slow conquests for Ga.
Does any of this matter without the control temp for your assessment? Cause this just goes over anyone's head when given numbers from only one half of the equation. Liquid metal generally just isn't necessary especially with newer cards in regards to overclock unless you are purely going for a degree or two change.
@@killzone_GG mate I don't have precise figures for you, lm amounted to +30ish MHz at -20 degrees celcius. Then I cleaned the lm and applied mx4, that was nearly as bad as it was before lm. Heat conductivity of Si and (Ni plated) Cu are fixed values so if they are saturated then the bottleneck is the TIM. GP102 is a large but really hot chip, so it is safe to assume that the system is closer to thermal transfer saturation than something like a TU-104 or whatever, thus lm would make more sense for a 1080ti here. 3090's silicon dwarfs 1080ti's in sheer surface area so the heat density is lower while the amount of heat generated is substantially higher, thus you have a colder card that draws more power. There is supposed to be a 1080ti gaming x lm video on youtube with a somewhat proper scientific methodology, you may check that one out.
@@killzone_GG also, I am giving you delta T so unless you want to incorporate the heat generated by electrons strayed by the increased temperature there is no point in asking for a control temp. 2 or 3 degrees doesnt matter here anyway.
@@ahmetrefikeryilmaz4432 Okay that's a fairly interesting point you make. When it comes down the the smaller specifics of technology I don't know a ton about the different thermal applications liquid metal has besides videos reviewing the jist of it. Thanks for expanding my current knowledge of the subject,
A GPU I picked up online had liquid metal applied but somehow a small blob of liquid metal got itself situated under the core bridging two BGA balls together causing a black screen during driver initialisation and there was no sign of previous liquid metal clean-up or remnants of it 🤷 . I Fixed it with a sweet wrapper that I swiped across the BGA and the liquid metal came out quite easily.
Yeah, gum wrappers also work to "suck up liquid metal TIM, as does aluminum foil, just make sure to abrase the foil with some sandpaper first to remove the oxide layer to allow it to absorb faster. Someone I used to work with also claimed that heat under vacuum would also work, basically strip card down to PCB, put in vacuum chamber and use a coil element heater to raise the temp, claims the liquid TIM would boil off but I don't see how that would work, maybe with mercury but not other modern TIM materials, boiling point even under vacuum would be too high.
Pure isopropyl and air blasting with 90PSI from a compressor got rid of all the spillage under the VRAM chips on a 1070M that was causing a code43 I use liquid electrical tape all over the place to protect with LM, and also to protect against leakage as now I'm into the quietness of modding laptops with detachable watercooling which causes already overkill low temps (9900K @ 150W at 80C and 2x 1080's at 60C with horrible looking DIY soldered stuff but NO NOISE) so I have gone back to good goop
Best thing is probably to use a copper wick, like for other desoldering. If necessary, trim it so that it gets in the right place. Galinstan wicks pretty well onto copper. Air blasting risks blowing the metal into something else.
Mother of god he used waaaaaaayyyyy too much LM. Like you should be basically painting the LM on so thin that it barely covers the silicon and do the same to the heatsink. Hell I did my entire 3090 die with the amount he has globbed off the sides of the die. (15C delta over the water, it never fluctuates clock speed because it's always at full boost and has no TDP restriction)
There is nothing to worry about if you know how to apply liquid metal. A thin layer is more than enough, it's more like "painting" the gpu with an hyssop (and the base of the heatsink too), isolating is not required at all!
I think the elephant in the room is that this particular GPU has 10x the necessary amount of LM on it, and THAT is the best way to create problems.. Basically video starts to make sense at 15:00. Kind of much more important than thinking about protecting anything IMO. I've been putting LM everywhere since it become available on mass market and if you just paint it over with a cotton stick there won't be any issues described here. This looks like someone just squeezed a whole drop and then just put a cooler over and moved it around quite a bit.. then disassembled it and applied MOAR😂
I experienced the capacitance issue you mentioned around 10:38 when I first put liquid metal on my 1080Ti. I made sure to protect the core, memory chips and exposed copper of the PCB with a generous layer of conformal coating, but yet I experienced instability and heavy artifacting in games. When I removed the heatsink I saw that a tiny blob of liquid metal had squeezed out from underneath the core and onto the (protected) substrate. All my problems went away the moment I wiped off the stray liquid metal. I guess the little bit of extra capacitance was enough to either upset the memory timings/signal delays or cause crosstalk between the individual tracks.
For covering SMDs and the space between the memory chips and the PCB, I've used silicone conformal coating. It might take a few coats to build it up enough to close the gap, but it's easy to apply and made for this kind of application. Be sure to have plenty of ventilation though, there are some pretty heavy solvents in it. I'm not sure if it inhibits the gallium-aluminum interaction, but it might be worth trying. If that works, it would be super easy to coat the exposed aluminum.
@@kekoraaaa it probably works fine, but conformal coating is rated for long term high temperature operation. A 55 mL bottle has lasted me a long time and wasn't terribly expensive.
I just used clear nail polish over the components around the die. And. I didn't pour liquid metal on the die like a savage so I didn't have to worry about it coming out anyways
Why not just put the LM on the tip if the BRUSH and PAINT IT over? Why would anybody point the syringe towards the GPU and shoot? That is ... stupid. This guy was TRYING to kill the GPU.
THEORETICALLY, with ALOT of flux and a hot air station just flooding it with flux and heating the chip to the solder melting point MIGHT break the bridge between 2 pads since (being liquid) the surface tension of the flux will push the solder back into place and separate the bridge. but the only SURE way to fix is reball as you said.
I've put LM on 6600XT Red Devil in my case. Got like 15*C lower hotspot, 500 RPM less with same power consumption and GPU is way quieter and that's what i was going for :D Right now with 2x P12 Slim under GPU whole thing is 0 RPM for most of the time, fans start only during gaming or work in some programs and it's still, only for a while due to lower temperatures. Yeah, i don't get better performance, yeah i don't try to push clock as much as possible, because every game runs really nice on 1440p/144Hz (Not FO4 or Ryse obviosly :V) and every work which need to handled is done fast for me. If you want quieter PC then i think it's worth doing so if you get like 300-400 RPM less and lower temps. Why? Becase you can lower them even more and get same temperature with better acoustics which for many is pretty important. Yeah, it's not a level of custom loop with 2x360 45 mm with pull fans with great static pressure but for cheap i think it's great. LM + some nail polish is like 20$ for the whole thing or even less if you get your mom, girlfriend of sistem nail polish :D Yeah, you void your warranty (i live in EU) but if neccesary you can open GPU without destroying seals
LM is relatively safe when applied properly. Together with what was on the cooler this GPU got drowned in 10-20 times the ammount of LM that was needed. Twice the ammount of regular TIM you would've needed to re-paste this card.
On modern GPUs you have the shitass autoOC, that's EXTREMELY sensitive to temperature. And you can still do voltage curve corrections, which is also helped by dies temperature reduction. And noise reduction.
With modern autoOC do you need dynamic clocks or actual one click OC like in radeon settings? Because dynamic clock speek is a great advancement since the pascal days.
the auto OC adds like 10MHz per 10C temp improvement. Unless you're going for high scores it doesn't make a difference and if you are going for high scores just use dry ice or LN2.
@@fVNzO Dynamic clock speed screws over voltage control and can unreasonably make your clock jump around for no reason at all, including frequency jumping above a stable point "because it can and fk you". Which happens all the time, if you don't fix the voltage curve to a flat line after a frequency point. And Radeon Settings don't have one click OC, you still need to keep minimum and maximum clock 100 MHz away from each other for best results (close to no frequency dips and solid clock stability).
@@PolskiJaszczomb yes, dynamic clock speed with manual tuning is excessively annoying. I meant under stock but maybe I'm misinterpreting you. Radeon settings not giving two haystacks and a pair of nailclippers what the fuck you actually tell it to clock is indeed very annoying.
Personally I covered my caps and core edges with (nonconductive) thermal paste. As for any gains, whole 3 C drop under water cooling (and 6 C on hotspot), compared to 6 C on CPU and quite a bit more on hotspot. It's less than worth it, unless you're running thermally close to the edge. And for memory, really extra not worth it due to lower thermal density per chip and the risk of BGA being shorted. (Unless you put thermal paste or lacquer or epoxy on the edges.) Using it just on core has very low likelihood of a spill as long as when you take it apart, you do so carefully. Using it on memory... well... it will go everywhere. Additional recommendation is to also put it on the heatsink as a very thin layer as well, the surface tension will glue both surfaces together and vastly reduce the chances it gets squeezed out. Using thermal paste, in addition to being thermally conductive, forms a semi-liquid gasket.
Conductonaut is some good shit. And the missus' clear nail varnish. - 30c on 5700xt hotspot but card was used and likely had other issues as it bounced off 110c junction when i got it.
2:42 There are voltage curve controls in MSI Afterburner. Liquid metal allows me to get more out of the voltage curve that the stock thermal compound wouldn't. 🤷♂️
Frankly, that poor Saphire card looked like the guy didnt know what he was doing when he applied the TIM. Ive been running LM on all my GPUs for almost 5 years and havent had a single problem. I isolate the most endangered parts around the chip and only use a minimum of LM. That guy has SO much spill it couldnt have been applied correctly.
There is not even small protection on SMD :P I applied it first time in my life on a GPU and done like 5-6 layers on everything around die just to be sure ^^
Liquid metal is something that requires user skill and depending on any coatings present it takes time as well, Ive applied it 4 times to my PC at this point and the minimum time ive spent to properly get it to spread on the IHS and Cooler was 40 minutes Its infinitely easier to apply to a CPU vs a GPU because of the annoyance of most GPU coolers and how you mount them I use Liquid metal on my CPU because i need the maximum possible heat transfer i can get because of the ridiculously small cooler i am forced to use because small cases that are thin enough to fit where i need my PC they all have limited CPU cooler space And i have a 2 slot 3090 and i can say for sure that liquid metal on my card would not be good because the GPU stands vertically and chances of it pooling and dripping out are too high for a 2k card thats £2000 not $ or anything else (i was not going to wait any longer than i did) But even then my 2 slot 3090 is cooler than most of the rest of the 3090’s out there, it came stock with thermal-right pads and paste, tho i changed out the rear pads that where silicon and replaced the front pads anyway because i didn’t realise till the card was open so had the pads anyway
40 minutes to get it to spread? That means you are being way too delicate with it. Use a bit of force to press on the cotton tips and it will take 2-4minutes. Start from the center and use circular motions. You can do back and forth motions at the edges, but slow down at the corners.
@@Tealc2323 press to hard the cotton swap slips and your utterly FK’ed Press to lightly it will not stick But for me its more because dried out liquid metal doesn’t play nice with still liquid liquid-metal (at least not for me) But regardless it mostly comes down to full coverage and making damn well sure you dont have to much As well as testing and applying more if needed or adjusting the mounting pressure of each corner I do not allow for cores to have a temperature difference at all, them must be the same temperature or the liquid metal is not doing its job correctly Also once it dries out if the cooler shifts your shit out of luck and have to replace it 80c becomes 110c for me if it dries and unwelds itself (i say liquid metal welds things together when dry)
Vertical should be fine, you sound careful with the applications. I ran l vertical 9 months with rear IO at top of case and since February vertical standing up in a bench. As well as a few months in a stock case layout, its had plenty of opportunity to move round. You'd probably need to massively overdo the application for it to be an issue. Admittedly not on a 3090.
This is why you'd have to coat the entire board in a protective coating. Personally I'll stick to paste as having to double mate the surface and then add a fresh layer again a few months later sounds a bit much for my taste.
I think I get your point: the liquid metal under soldered components is basically a dead card (unless reballing etc...), but it is probably less likely to happen if someone does something cleanly (which is, I agree, not always easy but still), whereas the chip shortage problem is (to me) more likely to happen, but is "easier" to fix (not for the majority of people though). So I do not think putting protection on the exposed SMDs is that useless, given how easy it is to put too much liquid metal and get some runoff. And also, replacing SMDs is not an easy task to perform when not having spares available and the skills to diagnose, and perform the replacement.
Liquid metal was fun, but after killing a $1000 card with it, I'm out. It wasn't spilling that was the problem, but the waterblock becoming so strongly attached to the die, that removing the waterblock ripped a chunk of silicon from the middle of the die. This is what always happens on pure copper surfaces, the liquid metal hardens in about a year, maybe sooner. Nickel plated surfaces might be fine though.
I delidded my 7700K and put liquid metal under the IHS, I used nail polish to cover the gold pins on the PCB and everything works great but now im worried that the liquid metal will leak over time and gets out from under the IHS. I only glued the four corners of the IHS.
I have Conductonaut on my 5700xt with a Morpheus cooler and my wife's 5700xt sapphire pulse. I use conformal coating, lots of layers. And I'm very careful with my application. Just did the sapphire card a couple of months ago, mine has been well over a year. The sapphire pulse dropped 6-7 degrees, while gaining 66mhz and 50mV on the core with the same voltage/clock curve in furmark. Worth it for the laymen? No. Fun for an enthusiast? Definitely. My main reason for it is to achieve lower noise levels. I hate screaming GPU fans.
Isn't there a clear coat over the SMDs surrounding GPU? When I apply LM to GPU I use Tolulene free nail polish just as an extra, and while apply I kinda scrub it onto surface and kinda sticks on rubbed surfaces(cooler and silicon). I also assume due to capillary action and being rubbed on, it doesn't ball up unless it excessive and overflows. Good suggestion to protect mem and chip as extra!
About Gasket. I had discussion with one user on YT under similar video (about LM and how danger it is for newbies in PC). And look's like we have found solution. There are companies which make rubber\silicone o-rings. You need to measure how tall GPU\CPU die from PCB to top edge. This will be absolute minimum size of o-ring. You need just a bit thicker (0.1 or 0.2). Too thick - excessive pressure on PCB (we don't need that!). Diameter should be higher then CPU\GPU die size (no stretch!). When o-ring is installed and LM is applied just place heatsink on it. Pressure will squize o-ring and cover all gaps around CPU\GPU die. Flip it and screw bolts. Technicly it should work. Just need to find specific size and also it should hold AT LEAST 120 degree by C!
There liquid gasket products for automotive engines that replaced rubber gaskets. Can be picked up cheaply at auto parts stores. Much easier than making gaskets.
I actually reduced Junction Temps of my 7900 xtx on water from 71C to 59C. Benchmarks got only slightly better but I'm getting close to 4090 stock benchmarks
The Intel ihs has superior nickel plating to cpu and gpu coolers, using liquid metal against an average cooler will cause the gallium to harden and damage the cooler, or at least require sanding/lapping
I did 4 layers of clear nail polish on my 3070 laptop GPU,applied LM on only 70% in center of chip and heat sink.Half a year good so far temps are same as I initially did it just have to keeps fans clean,over 10 degree cooler than paste
First: the mouth breather that spilled that much LM should not touch any more hardware. Way too much LM and looks like Hellen Keller applied it. 6900xt/6950xt say differently. The difference in LM vs any paste is massive for hotspot temps. They do have voltage control. They do scale with voltage up 1.35vgpu.
I had LM under a memory chip on a 1070M, caused code43 Blobs are easy to deal with, just suck up everything you can with the conductonaut tube Pure isopropyl makes LM able to be picked up by paper towel thats what I use to pick up most of it Pure isopropyl and blasting from an air compressor (90psi) got rid of it from under the VRAM chips, fixed the card Liquid electrical tape (you could call it plastidip) is awesome
Compressed air. I have a commercial air compressor, you’ve gotta be careful though, it will blow the liquid metal out from between the cpu and the heat sink.
you need to know what to do with liquid metal in gpu and you are wrong is totaly worth the lm in gpu im using lm at everyone gpu i have if you use it correct hof 2080ti is around 10c lower than the best thermal paste out there
I feel i should mention that most Commercial applications use leaded solder for three main reasons better electrical conductivity, lower melt temperature reducing chance of component damage and Cost, Leaded solder is 1/3rd the price of the cheepest lead-free solder
This is why I use Arctic Silver Céramique 2 on everything. Not the best thermal paste in performance but it's both non-conductive and non-capacitive. Meaning you can spill it everywhere and it won't cause any problems. I'm not going to risk very expensive electronics to win a few degrees.
The safest way to do this is to put the LM first into a (clean) container, and then use the cotton bud to move it from there onto the die. Don't use the syringe directly on the die.
I got a 1080 ti that is dead (got it to fix it) and it have visual damage on the chip by liquid metal, makes no sense to do LM on GPU as it won't do much on temps anyways.
I also don't think it worth too. Unlike CPUs, on GPUs the cooler is already in direct contact with the chip, there is no IHS in between to affect thermal transfer. Apart from the fact that GPU's are much more locked to overclock than CPU's with minimal gains and also much easier to detect anything to deny the warranty later. In the same way that selling sealed GPU's is by far a lot easier than opened ones. For things like nitrogen, we use usual paste anyway.
I am in a 50/50 situation where my vertically mounted GPU dripped towards the bottom. Just cleaned it up, fingers crossed. When I turned my PC on, it flashed on and off 3 times and then nothing. Onboard works and GPU does not detect. Only dripped on the little things on the inside of the metal GPU chip surround. No burnt smell or obvious damage so I am somewhat hopeful. I only used it because I had no paste and now I am mad at myself for not waiting for it to arrive.
Update 3 - Everything seemed ok so shut down and restarted with the hdmi attached to the 6800 and it is all back to normal. Never will I use liquid metal again.... Just checked prices and where I live a 6800 is still over $500 Aus and it is water cooled meaning the block would be a waste as well. I am so exceptionally happy I did not destroy the GPU.
The paste on the VRAM is a yikes, too. Not really worthwhile, coverage is bad, and with the wrong z-stackup it's not conducting to the heatsink well and/or it's bending the pcb to make contact.
I remember you mentioning this on one of your streams, and after explaining it, I understand why it now appears to be a bad idea. I was seriously considering it for my gtx 690 and recently sourced Radeon 7990, admittedly. For the former, because the reference cooler seems like a good candidate for liquid metal and thermal efficiency improvements. For the latter, because it's a hot running bastard at load, even after replacing thermal pads across the card and applying Arctic mx-5 on the dies. Unfortunately, sourcing a water block for either is a challenge; and the only Arctic branded cooler I could find was for the 690, which made it unwieldy and huge. No such luck finding a replacement cooler for the 7990. :(
Im using LM on my RTX 2080 and I have no problem , Because the die is on top the LM Can't spill IF you reassemble the card with the cooler always on the down and you put the GPU on top. This guy OBVIOULY had no Idea what he was doing. The application was wrong and the reassembly also Wrong.
careful with silicone, as its acidic you'll corrode/oxidise a lot of the metals(and cause shorts), so you'll need food safe or non acid curing silicone just fyi
Isn't liquid metal gallium? Like on thermal grizzly conductonaut stuff. I've been putting it on shunts for shunt modding and have yet to have one get the solder dissolved. Also liquid metal on GPU can still drop 10C or more if you have a good enough cooler, which is great for reducing fan speeds.
Not mostly gallium, but gallium is the lowest-melting component, so it's what causes the dissolution, yes. As for noise, well, the channel's not "Actually Quieter Computing".
It’s a mixture of gallium, indium and tin. It’s a so called eutectic mixture which means that it „is a homogeneous mixture of substances that melts or solidifies at a single temperature that is lower than the melting point of any of the constituent“ Indium has a melting point of 156 degrees, gallium 30 degrees and tin 230 degrees. The eutectic mixture brings the melting point below 30 degrees Celsius (to around -19 degrees) otherwise the liquid metal paste would not be liquid at room temperature. I suppose they also add stuff to lower the surface tension of the mixture because the surface tension of Galinstan (which thus mixture is called) is about ten times higher than water, which is why it forms those tiny beads or balls that roll or flow around everywhere and why Lm is such a bitch to clean up because it doesn’t easily stick to anything
@@Numfuddle I'm familiar with eutectic mixtures, but I don't think it's relevant (as long as the gallium is liquid). Pure gallium will destroy aluminum and embrittle steel, for instance.
In one video, you offhandedly said that ddr5 2x8gb kits are trash. Could you go into more detail if all I care about is running an XMP profile and gaming, how much of a difference would there be between a 2x8 vs 2x16gb kit of 5200 cl40? Thinking of building a new pc as AM5 drops and I don't really need 32gb of ram.
How the F do you apply that much LM. Seriously. If you can't handle it don't use it. But this is not in reverse a non recommendation like BZ says in my point.
Sadly a hardware component died for it. Don't mind spending hundreds of dollardos or what you have on something you are excited about, but don't take time to read up how to prepare for application of LM. But than again some will only learn a lesson the hard way.
Liquid metal is fine, if you like to live on the edge / are careful with it's application. Really, you don't need that much. As for making a gasket, hot-glue could work with a card like this, sure it would not be perfect but better than nothing.
ive used LM on my 1660ti and it worked great due to the small die area as it was relatively flat. LM is amazing but as you get in with big dies, 2070+ 3070+ you see that LM is actually a bad idea. its too thin, like water. there will be INTENSE hotspots that the gpu cant measure. i saw artifacts on my 2080s and my 3080ti with LM and i promise you, i know how to apply it properly. if you want LM to work PROPERLY on GPU dies you need to LAP the die AND the Cooler. (goodluck) the die is not flat, and neither is the cooler. since then, for GPU's i use Kingpins Extreme paste, it works flawlessly. for CPU's, i Delid, Lap the IHS, and use LM on both sides. IHS side and Die side. works flawlessly. dont use it on large die GPU's.
I put LM on my GTX 780 ages ago when I was young and dumb (still dumb, but less so now). Had it on my Naked Die 3770k too so I thought, hey, why not. Ended up removing it after not-so-great results and its only much later now that I realise how fucking lucky I was to not kill that card. Didnt protect any of the components, only put it on the GPU die (not the cooler too, as you're supposed to) and probably a ton of other dumb things i'm forgetting. I really wanna put LM on my vega just to see how it is, though I think I already know. Took me many attempts to get a good mount with good hotspot temps, yet I still wanna try it.
Given Vega has problems with package height (the HBM and the Core on the same substrate), probably not a good idea. Liquid Metal isn't as thick as thermal paste and works best in between really flat things. *cough* Radeon VII thermal pad
@@InternetEntity Still gonna try it one day regardless, maybe soon. I know vega likes thicker thermal paste like kryonaut, which is what I have now. My card has a molded package by the way, HBM and core are the same height....in theory. Hotspot still scares the shit out of me.
Derbauer put a motherboard in the dishwasher to clean of vaseline, would it help on this card? Does LM dissolve with alcohol, would it make sense to dump it in a tub of Iso alcohol?
put liquid metal on my Vega64 when i put on a waterblock. varnished everything around it. been running fine for 3+years never goes over 45degrees and getting over 150mhz higher. ( runs up to 1800mhz though hits 1750 average.)
What is this about not being able to change the voltage on GPUs? is this something with RTX 3000 and RX 6000 specifically? I remember tuning voltage on 10 series and a Radeon 7
I found out that and went overboard on the voltage of the 6950xt with lowering the voltage I can have a stable clockspeed of 2550mfgz on 1120 mv with a hotspot of 87 Celsius
Call me lucky, but I managed to remove LM from under my laptop GPU memory chip and it's still working normally. Was not easy. I then went all out with conformal coating and reapplied, all good. Only GPU I'd apply LM is a laptop's anyways, different scenario.
As someone who had a go at Liquid Metal: Its kinda not worth it. Good thermal paste and good contact gets you within a few degrees of Liquid Metal in my experience. I also had issues with large temp deltas between Hotspot and Die Temp aka it sometimes wasnt easy to have good contact with the super thin Liquid Metal film instead of a blob of thermal paste.
"short" video 😂 Interesting 😁 I got about 5 or maybe 10c reduction on my flashed and watercooled 980....it really helped it clock higher. I guess that's an exception :)
First system (laptop) I used LM on I was able to overclock a 680M up from 620Mhz stock to 700ish on AS5 to beyond 1GHz, without any further mods In laptops where radiator volume and airflow is limited, LM is AMAZING
I wonder if an ultrasonic cleaner could get the liquid metal out from beneath the memory/core? Because we already saw with the smd components that it didn’t really deteriorate them, so maybe there’s a chance all you need to do is clean under the chips without removing them.
Yes an ultrasonic cleaner will work...People think these things are more fragile than they actually are...I have literally put Motherboards and GPUs in the dishwasher before...As long as you don't use soap and you let it dry completely afterwards it shouldn't hurt it at all...I usually give it a 99% alcohol bath also just to be sure
I run vega 56 (with blower cooler) sitting on 64s bios. So it has unlocked extra clock speeds on core and memory. And with Vegas the overall temps are kinda irrelevant becasue Hotspot temp exists. For preference overall which is shown is Radeon software 70C meanwhile hotspot is 110C. Which already is degrading chip slowly and also thermal throttles. So putting LM saved me basicly 10C in winters im living warm with my heatspreader but in summers im better touching grass outside. Temps go crazy even tho i have undervolted core and memory a bit.
The only thing that helped my Vega 64 was using kryonaut and really butter the biscuit (use a lot). Igors lab writes that the screws opposite the PCI should be tightened down first, I do not know if it makes a difference, but I followed the guide. I also used washers to tighten the mounting pressure, but it did not change the temps. I removed the fans and shroud so only the heatsink was left and installed 2 noctua 140mm fans on the card instead. Managed to keep the hotspot in the 90es after all that work. I was not able to change the clocks in the AMD driver, the Asus bios somehow blocked it, downclocking 10MHz or something stupid like that, made the card unstable. Installing the Asus software to adjust fans and clocks made a conflict with the AMD driver and the GPU fans turned off until the card thermal protection kicked in and ran fans at very high speed, the fans turned off at 65 degrees again and everything started over. Wanted to try and bios flash to non Asus bios, the program denied it and I was not brave enough to push on with more tinkering after that.
I messed up witk LM..got some spreading of it ..tried to secure with spraypaint ..now when starting upp , it works normally so you can see startupp ,and sometimes the windows loadup icon ..then the screen wents black ...so i have supposed its toasted 🤷
If you benchmarked and have a reference you only need to see if their are out of order changes. LM doesn't degrade like other thermal paste, unless it dissolves into the metal you applied it to, mostly as eMpTy said copper comes to mind. So to speak for your delided chips if you applied the LM between chip and IHS you won't need a change anyway. For your laptop application it depends on the heat sink material and your degradation. If there is no big gap in temps over what you got when first applied you not have to change anything. If the temps were to go up you might go for a reaplication. Never change a running system. If nothing changes it stays the same.
@@Airwave2k2 Thats what I thought. Everything is good except the macbook and it could be that I don't use it enough to know if the temps went up. Thanks!
Nice, been thinking about this since I started the video lol. Nickel plated GPU cooler and temps are the same after 14 months. Had to flash a custom bios just to change fan activation and cut off as the GPU was so "cold" the fans weren't cooling the bloody GDDR6.
There's repair channels that basically don't work on any boards that are covered in liquid metal because there's a very high chance you end up in a never-ending rabbit hole of trying to figure out which component is dead or whatever.
I had a pretty weird liquid metal mishap a while back, I killed a 980 even though there was no liquid metal on the SMDs or anywhere else on the card.. core coverage was good but it still died leaving a scorch mark on the cooler surface. I know the liquid metal killed it I just have no idea how 🤔 If I was in the US I'd send it to you just to satisfy my curiosity.
Was it an EVGA ACX card? My B-stock 980Ti's VRMs went up in smoke and I found out a while later that cooler did not do a good job cooling VRMs. Also, use nail polish and brush it onto the SMDs to protect them, colored if you can because you can easily see if it covers fully
@@kekoraaaa It's an MSI Gaming 4G, there's no visible carnage on the VRM or anywhere else on the board though. I had it in SLI (wtf was I thinking ^_^) with an identical card on stock paste and that card was fine. I think actually had somebody check it and they said the core is dead which sounds about right given the edge of the chip left a mark on the cooler like it ate through the nickel plating... must've been one hell of a short. RIP 980, expensive lesson learned 😁
damn, that's quite a big mess, anyway for the price of liquid metal I really wonder why can't they just always ship a protective liquid like just a few small amount of grams specially made to protect stuff around, I mean the liquid metal is not exactly that expensive from what I saw people doing similar mixes and we pay a lot for 1gr so it should have come with a protective specially made product too. People usually use nail polish, however in the past many years ago I used nail polish many times for other random stuff and for me after some time it always got fully cracked and most didn't even had heat into the equation so I actually wonder if nail polish is actually any good as a permanent protective solution against liquid metal cause I'm planning to delid a CPU, but never touch it again for the next decade unless temps really start to get strange...
![BOWKYLION - วิงวอน (ex-change) [Official MV]](http://i.ytimg.com/vi/cLdnZWDaO-w/mqdefault.jpg)
Been running a EVGA 1080Ti with LM for nearly 2 years with a custom loop. Flashed the bios with a KP version to not have the voltage restrictions. It achieve a 10C benefit at my max core overclock of 2105 MHz. I’d recommend any enthusiast to try it, you just need to research the knowledge and use some care and common sense when applying. The benefits while minor are there. BZ has done some great past videos on how to handle LM properly. Ironic that the viewers who sent in these cards hadn’t watched them.
Well BZ, no, let me explain:
It's all about the amount of heat transferred per area.
I had a 1080ti gaming X which pulled around 330W at 117% power. The cooler was phenomenal I thought, then I started noticing how quick the temps were rising. I fiddled with the soft temp limit and saw that the cooling wasn't actually adequate and the card was soft throttling. I repasted with MX-4. The temps were better but only a couple degrees. I had LM on hand (Thermal grizzly stuff) which I used for my 7700K delid(s). I covered the smd caps with nail polish, applied just enough and I buffed it in, that is really important. The card afterwards would stop at 69 degrees celcius under the same fan settings, I lost like two degrees in a year and reapplied. I used it like that for about two years and I just sold it, of course disclosing everything and letting the buyer actually try it beforehand. The buyer was skeptical at first but later he was like "looks like this LM stuff really works". I sold him an unopened 1g pack of conductuonaut along with the 1080ti.
The reason LM sticks to tin is Ga-Sn amalgamation btw, not some sort of Sn-Sn debauchery. Ga amalgamates with tin very much. It amalgamates with Cu and Ni too btw, it's just that those two are really slow conquests for Ga.
Does any of this matter without the control temp for your assessment? Cause this just goes over anyone's head when given numbers from only one half of the equation. Liquid metal generally just isn't necessary especially with newer cards in regards to overclock unless you are purely going for a degree or two change.
@@killzone_GG mate I don't have precise figures for you, lm amounted to +30ish MHz at -20 degrees celcius. Then I cleaned the lm and applied mx4, that was nearly as bad as it was before lm.
Heat conductivity of Si and (Ni plated) Cu are fixed values so if they are saturated then the bottleneck is the TIM.
GP102 is a large but really hot chip, so it is safe to assume that the system is closer to thermal transfer saturation than something like a TU-104 or whatever, thus lm would make more sense for a 1080ti here.
3090's silicon dwarfs 1080ti's in sheer surface area so the heat density is lower while the amount of heat generated is substantially higher, thus you have a colder card that draws more power.
There is supposed to be a 1080ti gaming x lm video on youtube with a somewhat proper scientific methodology, you may check that one out.
@@killzone_GG also, I am giving you delta T so unless you want to incorporate the heat generated by electrons strayed by the increased temperature there is no point in asking for a control temp. 2 or 3 degrees doesnt matter here anyway.
@@ahmetrefikeryilmaz4432 Okay that's a fairly interesting point you make. When it comes down the the smaller specifics of technology I don't know a ton about the different thermal applications liquid metal has besides videos reviewing the jist of it. Thanks for expanding my current knowledge of the subject,
Saygılar hocam :)
A GPU I picked up online had liquid metal applied but somehow a small blob of liquid metal got itself situated under the core bridging two BGA balls together causing a black screen during driver initialisation and there was no sign of previous liquid metal clean-up or remnants of it 🤷 . I Fixed it with a sweet wrapper that I swiped across the BGA and the liquid metal came out quite easily.
Yeah, gum wrappers also work to "suck up liquid metal TIM, as does aluminum foil, just make sure to abrase the foil with some sandpaper first to remove the oxide layer to allow it to absorb faster.
Someone I used to work with also claimed that heat under vacuum would also work, basically strip card down to PCB, put in vacuum chamber and use a coil element heater to raise the temp, claims the liquid TIM would boil off but I don't see how that would work, maybe with mercury but not other modern TIM materials, boiling point even under vacuum would be too high.
Pure isopropyl and air blasting with 90PSI from a compressor got rid of all the spillage under the VRAM chips on a 1070M that was causing a code43
I use liquid electrical tape all over the place to protect with LM, and also to protect against leakage as now I'm into the quietness of modding laptops with detachable watercooling which causes already overkill low temps (9900K @ 150W at 80C and 2x 1080's at 60C with horrible looking DIY soldered stuff but NO NOISE) so I have gone back to good goop
Best thing is probably to use a copper wick, like for other desoldering. If necessary, trim it so that it gets in the right place. Galinstan wicks pretty well onto copper. Air blasting risks blowing the metal into something else.
I wonder if a sonic cleaner could do anything against such problems.
Mother of god he used waaaaaaayyyyy too much LM. Like you should be basically painting the LM on so thin that it barely covers the silicon and do the same to the heatsink. Hell I did my entire 3090 die with the amount he has globbed off the sides of the die. (15C delta over the water, it never fluctuates clock speed because it's always at full boost and has no TDP restriction)
Yeah, you could cover 10 GPUs with what was applied there, it's literally insane. That was likely done by a very generous person.
There is nothing to worry about if you know how to apply liquid metal. A thin layer is more than enough, it's more like "painting" the gpu with an hyssop (and the base of the heatsink too), isolating is not required at all!
I think the elephant in the room is that this particular GPU has 10x the necessary amount of LM on it, and THAT is the best way to create problems.. Basically video starts to make sense at 15:00. Kind of much more important than thinking about protecting anything IMO. I've been putting LM everywhere since it become available on mass market and if you just paint it over with a cotton stick there won't be any issues described here. This looks like someone just squeezed a whole drop and then just put a cooler over and moved it around quite a bit.. then disassembled it and applied MOAR😂
I experienced the capacitance issue you mentioned around 10:38 when I first put liquid metal on my 1080Ti. I made sure to protect the core, memory chips and exposed copper of the PCB with a generous layer of conformal coating, but yet I experienced instability and heavy artifacting in games. When I removed the heatsink I saw that a tiny blob of liquid metal had squeezed out from underneath the core and onto the (protected) substrate. All my problems went away the moment I wiped off the stray liquid metal.
I guess the little bit of extra capacitance was enough to either upset the memory timings/signal delays or cause crosstalk between the individual tracks.
For covering SMDs and the space between the memory chips and the PCB, I've used silicone conformal coating. It might take a few coats to build it up enough to close the gap, but it's easy to apply and made for this kind of application. Be sure to have plenty of ventilation though, there are some pretty heavy solvents in it. I'm not sure if it inhibits the gallium-aluminum interaction, but it might be worth trying. If that works, it would be super easy to coat the exposed aluminum.
Have you tried using nail polish instead? It's worked well for my CPU SMDs and only takes 1 coat if you do it right
@@kekoraaaa Conformal coating is specifically made for this purpose, so I'm not sure that's a great idea.
@@kekoraaaa it probably works fine, but conformal coating is rated for long term high temperature operation. A 55 mL bottle has lasted me a long time and wasn't terribly expensive.
Thanks! ✌
I just used clear nail polish over the components around the die. And. I didn't pour liquid metal on the die like a savage so I didn't have to worry about it coming out anyways
Bit harsh on savages TBF
Why not just put the LM on the tip if the BRUSH and PAINT IT over? Why would anybody point the syringe towards the GPU and shoot? That is ... stupid. This guy was TRYING to kill the GPU.
THEORETICALLY, with ALOT of flux and a hot air station just flooding it with flux and heating the chip to the solder melting point MIGHT break the bridge between 2 pads since (being liquid) the surface tension of the flux will push the solder back into place and separate the bridge. but the only SURE way to fix is reball as you said.
This is why Sony uses a foam spacer on the PS5 to prevent their pool of LM from leaking out. Also prevents dry-out as well.
I've put LM on 6600XT Red Devil in my case. Got like 15*C lower hotspot, 500 RPM less with same power consumption and GPU is way quieter and that's what i was going for :D
Right now with 2x P12 Slim under GPU whole thing is 0 RPM for most of the time, fans start only during gaming or work in some programs and it's still, only for a while due to lower temperatures. Yeah, i don't get better performance, yeah i don't try to push clock as much as possible, because every game runs really nice on 1440p/144Hz (Not FO4 or Ryse obviosly :V) and every work which need to handled is done fast for me. If you want quieter PC then i think it's worth doing so if you get like 300-400 RPM less and lower temps. Why? Becase you can lower them even more and get same temperature with better acoustics which for many is pretty important. Yeah, it's not a level of custom loop with 2x360 45 mm with pull fans with great static pressure but for cheap i think it's great. LM + some nail polish is like 20$ for the whole thing or even less if you get your mom, girlfriend of sistem nail polish :D Yeah, you void your warranty (i live in EU) but if neccesary you can open GPU without destroying seals
Clear fingernail polish is what I used for my cpu.
LM is relatively safe when applied properly. Together with what was on the cooler this GPU got drowned in 10-20 times the ammount of LM that was needed. Twice the ammount of regular TIM you would've needed to re-paste this card.
"That's it for the video."
Goes on for another 3 minutes lol
On modern GPUs you have the shitass autoOC, that's EXTREMELY sensitive to temperature. And you can still do voltage curve corrections, which is also helped by dies temperature reduction.
And noise reduction.
With modern autoOC do you need dynamic clocks or actual one click OC like in radeon settings? Because dynamic clock speek is a great advancement since the pascal days.
the auto OC adds like 10MHz per 10C temp improvement. Unless you're going for high scores it doesn't make a difference and if you are going for high scores just use dry ice or LN2.
@@fVNzO Dynamic clock speed screws over voltage control and can unreasonably make your clock jump around for no reason at all, including frequency jumping above a stable point "because it can and fk you". Which happens all the time, if you don't fix the voltage curve to a flat line after a frequency point.
And Radeon Settings don't have one click OC, you still need to keep minimum and maximum clock 100 MHz away from each other for best results (close to no frequency dips and solid clock stability).
@@PolskiJaszczomb yes, dynamic clock speed with manual tuning is excessively annoying. I meant under stock but maybe I'm misinterpreting you.
Radeon settings not giving two haystacks and a pair of nailclippers what the fuck you actually tell it to clock is indeed very annoying.
This guy put like 6 gpus worth of liquid metal on that core lol
Personally I covered my caps and core edges with (nonconductive) thermal paste. As for any gains, whole 3 C drop under water cooling (and 6 C on hotspot), compared to 6 C on CPU and quite a bit more on hotspot. It's less than worth it, unless you're running thermally close to the edge. And for memory, really extra not worth it due to lower thermal density per chip and the risk of BGA being shorted. (Unless you put thermal paste or lacquer or epoxy on the edges.)
Using it just on core has very low likelihood of a spill as long as when you take it apart, you do so carefully. Using it on memory... well... it will go everywhere.
Additional recommendation is to also put it on the heatsink as a very thin layer as well, the surface tension will glue both surfaces together and vastly reduce the chances it gets squeezed out.
Using thermal paste, in addition to being thermally conductive, forms a semi-liquid gasket.
conformal coating + high temp electrical tape is what I use to protect around the gpu die.
This is the best example of "how not to do it". Do your homework before messing with liquid metal.
Conductonaut is some good shit. And the missus' clear nail varnish. - 30c on 5700xt hotspot but card was used and likely had other issues as it bounced off 110c junction when i got it.
I prefer using a colored nail polish to easily see if it's fully covered hahahah
@@kekoraaaa not a fan of her color palette, needs a bottle of plastidip orange lol
Would definitely help tho
2:42 There are voltage curve controls in MSI Afterburner. Liquid metal allows me to get more out of the voltage curve that the stock thermal compound wouldn't. 🤷♂️
And AMD more power tool
also with voltage curve i can raise the voltage lets stay from 1.05 max to 1.075 or 1.09
Id say the best way to clean up bridging would be to tempt it out with something the liquid metal would find tasty.
Frankly, that poor Saphire card looked like the guy didnt know what he was doing when he applied the TIM. Ive been running LM on all my GPUs for almost 5 years and havent had a single problem. I isolate the most endangered parts around the chip and only use a minimum of LM.
That guy has SO much spill it couldnt have been applied correctly.
There is not even small protection on SMD :P I applied it first time in my life on a GPU and done like 5-6 layers on everything around die just to be sure ^^
Liquid metal is something that requires user skill and depending on any coatings present it takes time as well, Ive applied it 4 times to my PC at this point and the minimum time ive spent to properly get it to spread on the IHS and Cooler was 40 minutes
Its infinitely easier to apply to a CPU vs a GPU because of the annoyance of most GPU coolers and how you mount them
I use Liquid metal on my CPU because i need the maximum possible heat transfer i can get because of the ridiculously small cooler i am forced to use because small cases that are thin enough to fit where i need my PC they all have limited CPU cooler space
And i have a 2 slot 3090 and i can say for sure that liquid metal on my card would not be good because the GPU stands vertically and chances of it pooling and dripping out are too high for a 2k card thats £2000 not $ or anything else (i was not going to wait any longer than i did)
But even then my 2 slot 3090 is cooler than most of the rest of the 3090’s out there, it came stock with thermal-right pads and paste, tho i changed out the rear pads that where silicon and replaced the front pads anyway because i didn’t realise till the card was open so had the pads anyway
40 minutes to get it to spread? That means you are being way too delicate with it. Use a bit of force to press on the cotton tips and it will take 2-4minutes.
Start from the center and use circular motions. You can do back and forth motions at the edges, but slow down at the corners.
@@Tealc2323 press to hard the cotton swap slips and your utterly FK’ed
Press to lightly it will not stick
But for me its more because dried out liquid metal doesn’t play nice with still liquid liquid-metal (at least not for me)
But regardless it mostly comes down to full coverage and making damn well sure you dont have to much
As well as testing and applying more if needed or adjusting the mounting pressure of each corner
I do not allow for cores to have a temperature difference at all, them must be the same temperature or the liquid metal is not doing its job correctly
Also once it dries out if the cooler shifts your shit out of luck and have to replace it
80c becomes 110c for me if it dries and unwelds itself (i say liquid metal welds things together when dry)
Vertical should be fine, you sound careful with the applications. I ran l vertical 9 months with rear IO at top of case and since February vertical standing up in a bench. As well as a few months in a stock case layout, its had plenty of opportunity to move round.
You'd probably need to massively overdo the application for it to be an issue.
Admittedly not on a 3090.
This is why you'd have to coat the entire board in a protective coating. Personally I'll stick to paste as having to double mate the surface and then add a fresh layer again a few months later sounds a bit much for my taste.
I think I get your point: the liquid metal under soldered components is basically a dead card (unless reballing etc...), but it is probably less likely to happen if someone does something cleanly (which is, I agree, not always easy but still), whereas the chip shortage problem is (to me) more likely to happen, but is "easier" to fix (not for the majority of people though). So I do not think putting protection on the exposed SMDs is that useless, given how easy it is to put too much liquid metal and get some runoff. And also, replacing SMDs is not an easy task to perform when not having spares available and the skills to diagnose, and perform the replacement.
Liquid metal was fun, but after killing a $1000 card with it, I'm out. It wasn't spilling that was the problem, but the waterblock becoming so strongly attached to the die, that removing the waterblock ripped a chunk of silicon from the middle of the die. This is what always happens on pure copper surfaces, the liquid metal hardens in about a year, maybe sooner. Nickel plated surfaces might be fine though.
I delidded my 7700K and put liquid metal under the IHS, I used nail polish to cover the gold pins on the PCB and everything works great but now im worried that the liquid metal will leak over time and gets out from under the IHS. I only glued the four corners of the IHS.
"rather short video" video is over 17 minutes long... Never change
I have Conductonaut on my 5700xt with a Morpheus cooler and my wife's 5700xt sapphire pulse. I use conformal coating, lots of layers. And I'm very careful with my application. Just did the sapphire card a couple of months ago, mine has been well over a year.
The sapphire pulse dropped 6-7 degrees, while gaining 66mhz and 50mV on the core with the same voltage/clock curve in furmark.
Worth it for the laymen? No. Fun for an enthusiast? Definitely. My main reason for it is to achieve lower noise levels. I hate screaming GPU fans.
Isn't there a clear coat over the SMDs surrounding GPU? When I apply LM to GPU I use Tolulene free nail polish just as an extra, and while apply I kinda scrub it onto surface and kinda sticks on rubbed surfaces(cooler and silicon). I also assume due to capillary action and being rubbed on, it doesn't ball up unless it excessive and overflows. Good suggestion to protect mem and chip as extra!
This also looks like a case of using WAY too much liquid metal. It doesn't make a mess when you use the correct amount.
About Gasket. I had discussion with one user on YT under similar video (about LM and how danger it is for newbies in PC). And look's like we have found solution. There are companies which make rubber\silicone o-rings. You need to measure how tall GPU\CPU die from PCB to top edge. This will be absolute minimum size of o-ring. You need just a bit thicker (0.1 or 0.2). Too thick - excessive pressure on PCB (we don't need that!). Diameter should be higher then CPU\GPU die size (no stretch!). When o-ring is installed and LM is applied just place heatsink on it. Pressure will squize o-ring and cover all gaps around CPU\GPU die. Flip it and screw bolts. Technicly it should work. Just need to find specific size and also it should hold AT LEAST 120 degree by C!
There liquid gasket products for automotive engines that replaced rubber gaskets. Can be picked up cheaply at auto parts stores. Much easier than making gaskets.
Simple reason why I put liquid metal on my GPU: because I had the tube around anyway
I actually reduced Junction Temps of my 7900 xtx on water from 71C to 59C. Benchmarks got only slightly better but I'm getting close to 4090 stock benchmarks
The Intel ihs has superior nickel plating to cpu and gpu coolers, using liquid metal against an average cooler will cause the gallium to harden and damage the cooler, or at least require sanding/lapping
I did 4 layers of clear nail polish on my 3070 laptop GPU,applied LM on only 70% in center of chip and heat sink.Half a year good so far temps are same as I initially did it just have to keeps fans clean,over 10 degree cooler than paste
First: the mouth breather that spilled that much LM should not touch any more hardware. Way too much LM and looks like Hellen Keller applied it.
6900xt/6950xt say differently. The difference in LM vs any paste is massive for hotspot temps.
They do have voltage control.
They do scale with voltage up 1.35vgpu.
I had LM under a memory chip on a 1070M, caused code43
Blobs are easy to deal with, just suck up everything you can with the conductonaut tube
Pure isopropyl makes LM able to be picked up by paper towel thats what I use to pick up most of it
Pure isopropyl and blasting from an air compressor (90psi) got rid of it from under the VRAM chips, fixed the card
Liquid electrical tape (you could call it plastidip) is awesome
What did you use the liquid electrical tape on?
@@UnluckyDomino Things like the outer edge of the die where the MLCCs are.
when you said using liquid metal on cpu, other than a delid. Would you use it on a nickel plated cpu mono block?
Would it be a good idea to use nail polish as a layer of protection? Generally, the things you would normally use in waterproofing.
i used some when i delided my i7 7700k its still running on it now as we speak
Compressed air. I have a commercial air compressor, you’ve gotta be careful though, it will blow the liquid metal out from between the cpu and the heat sink.
you need to know what to do with liquid metal in gpu and you are wrong is totaly worth the lm in gpu im using lm at everyone gpu i have if you use it correct hof 2080ti is around 10c lower than the best thermal paste out there
I feel i should mention that most Commercial applications use leaded solder for three main reasons better electrical conductivity, lower melt temperature reducing chance of component damage and Cost, Leaded solder is 1/3rd the price of the cheepest lead-free solder
This is why I use Arctic Silver Céramique 2 on everything. Not the best thermal paste in performance but it's both non-conductive and non-capacitive. Meaning you can spill it everywhere and it won't cause any problems. I'm not going to risk very expensive electronics to win a few degrees.
Hot glue ftw lol. Cmdr had a cool video about liquid metal damage recently too. Makes more sense after hearing u explain what the issues are.
The safest way to do this is to put the LM first into a (clean) container, and then use the cotton bud to move it from there onto the die.
Don't use the syringe directly on the die.
Running a GPU in a pool would be pretty interesting. Very insane and impractical but fun to watch
I got a 1080 ti that is dead (got it to fix it) and it have visual damage on the chip by liquid metal, makes no sense to do LM on GPU as it won't do much on temps anyways.
Petroleum jelly makes a decent barrier you can remove
a good way to build a barrier around you liquid metal applications is thermal paste
I also don't think it worth too. Unlike CPUs, on GPUs the cooler is already in direct contact with the chip, there is no IHS in between to affect thermal transfer. Apart from the fact that GPU's are much more locked to overclock than CPU's with minimal gains and also much easier to detect anything to deny the warranty later. In the same way that selling sealed GPU's is by far a lot easier than opened ones. For things like nitrogen, we use usual paste anyway.
"There is no IHS in between to affect thermal transfer..."
Ah so the LM will be even more effective! Yay. (j/k i know what you mean)
I am in a 50/50 situation where my vertically mounted GPU dripped towards the bottom.
Just cleaned it up, fingers crossed.
When I turned my PC on, it flashed on and off 3 times and then nothing.
Onboard works and GPU does not detect.
Only dripped on the little things on the inside of the metal GPU chip surround.
No burnt smell or obvious damage so I am somewhat hopeful.
I only used it because I had no paste and now I am mad at myself for not waiting for it to arrive.
Update 3 - Everything seemed ok so shut down and restarted with the hdmi attached to the 6800 and it is all back to normal.
Never will I use liquid metal again....
Just checked prices and where I live a 6800 is still over $500 Aus and it is water cooled meaning the block would be a waste as well.
I am so exceptionally happy I did not destroy the GPU.
There's nothing like being super screwed.
"rather short"
The paste on the VRAM is a yikes, too. Not really worthwhile, coverage is bad, and with the wrong z-stackup it's not conducting to the heatsink well and/or it's bending the pcb to make contact.
You can get gasket making materials from any Auto parts store that would work for this
once I had LM inside an AM3+ socket, after "cleaning" a bit was still there, but worked without issues lol
LETS DO THIS
I remember you mentioning this on one of your streams, and after explaining it, I understand why it now appears to be a bad idea. I was seriously considering it for my gtx 690 and recently sourced Radeon 7990, admittedly. For the former, because the reference cooler seems like a good candidate for liquid metal and thermal efficiency improvements. For the latter, because it's a hot running bastard at load, even after replacing thermal pads across the card and applying Arctic mx-5 on the dies.
Unfortunately, sourcing a water block for either is a challenge; and the only Arctic branded cooler I could find was for the 690, which made it unwieldy and huge. No such luck finding a replacement cooler for the 7990. :(
Im using LM on my RTX 2080 and I have no problem , Because the die is on top the LM Can't spill IF you reassemble the card with the cooler always on the down and you put the GPU on top. This guy OBVIOULY had no Idea what he was doing. The application was wrong and the reassembly also Wrong.
careful with silicone, as its acidic you'll corrode/oxidise a lot of the metals(and cause shorts), so you'll need food safe or non acid curing silicone just fyi
Can liquid metal be removed with a ultrasonic cleaner?
Isn't liquid metal gallium? Like on thermal grizzly conductonaut stuff. I've been putting it on shunts for shunt modding and have yet to have one get the solder dissolved. Also liquid metal on GPU can still drop 10C or more if you have a good enough cooler, which is great for reducing fan speeds.
Not mostly gallium, but gallium is the lowest-melting component, so it's what causes the dissolution, yes. As for noise, well, the channel's not "Actually Quieter Computing".
@@concinnus for sure but he's added capacitors for 10-20mhz extra clocks which can be done with liquid metal on these new cards lol.
It’s a mixture of gallium, indium and tin. It’s a so called eutectic mixture which means that it „is a homogeneous mixture of substances that melts or solidifies at a single temperature that is lower than the melting point of any of the constituent“
Indium has a melting point of 156 degrees, gallium 30 degrees and tin 230 degrees. The eutectic mixture brings the melting point below 30 degrees Celsius (to around -19 degrees) otherwise the liquid metal paste would not be liquid at room temperature. I suppose they also add stuff to lower the surface tension of the mixture because the surface tension of Galinstan (which thus mixture is called) is about ten times higher than water, which is why it forms those tiny beads or balls that roll or flow around everywhere and why Lm is such a bitch to clean up because it doesn’t easily stick to anything
@@Numfuddle I'm familiar with eutectic mixtures, but I don't think it's relevant (as long as the gallium is liquid). Pure gallium will destroy aluminum and embrittle steel, for instance.
@@concinnus it’s relevant because if it weren’t an eutectic mixture the melting point would be higher and the LM would be solid at room temperature
Always use Silicone Modified Conformal Coating, never had an issue with shorts on GPU liquid metal applications.
Don't use standard hot glue. Use high temperature hot glue.
In one video, you offhandedly said that ddr5 2x8gb kits are trash. Could you go into more detail if all I care about is running an XMP profile and gaming, how much of a difference would there be between a 2x8 vs 2x16gb kit of 5200 cl40? Thinking of building a new pc as AM5 drops and I don't really need 32gb of ram.
How the F do you apply that much LM. Seriously. If you can't handle it don't use it. But this is not in reverse a non recommendation like BZ says in my point.
Just didn't RTFM. Used it like normal thermal paste, i guess 🤔
That'll do it! At least something was learned.
Sadly a hardware component died for it. Don't mind spending hundreds of dollardos or what you have on something you are excited about, but don't take time to read up how to prepare for application of LM. But than again some will only learn a lesson the hard way.
If i were to use liquid metal i probably would also use thermoelectric grease around the chips to protect solder balls and components.
Liquid metal is fine, if you like to live on the edge / are careful with it's application. Really, you don't need that much.
As for making a gasket, hot-glue could work with a card like this, sure it would not be perfect but better than nothing.
Or you can just use nail polish over the components around the die...
ive used LM on my 1660ti and it worked great due to the small die area as it was relatively flat. LM is amazing but as you get in with big dies, 2070+ 3070+ you see that LM is actually a bad idea. its too thin, like water. there will be INTENSE hotspots that the gpu cant measure. i saw artifacts on my 2080s and my 3080ti with LM and i promise you, i know how to apply it properly. if you want LM to work PROPERLY on GPU dies you need to LAP the die AND the Cooler. (goodluck) the die is not flat, and neither is the cooler. since then, for GPU's i use Kingpins Extreme paste, it works flawlessly. for CPU's, i Delid, Lap the IHS, and use LM on both sides. IHS side and Die side. works flawlessly. dont use it on large die GPU's.
I put LM on my GTX 780 ages ago when I was young and dumb (still dumb, but less so now). Had it on my Naked Die 3770k too so I thought, hey, why not. Ended up removing it after not-so-great results and its only much later now that I realise how fucking lucky I was to not kill that card. Didnt protect any of the components, only put it on the GPU die (not the cooler too, as you're supposed to) and probably a ton of other dumb things i'm forgetting. I really wanna put LM on my vega just to see how it is, though I think I already know. Took me many attempts to get a good mount with good hotspot temps, yet I still wanna try it.
Given Vega has problems with package height (the HBM and the Core on the same substrate), probably not a good idea.
Liquid Metal isn't as thick as thermal paste and works best in between really flat things.
*cough* Radeon VII thermal pad
@@InternetEntity Still gonna try it one day regardless, maybe soon. I know vega likes thicker thermal paste like kryonaut, which is what I have now. My card has a molded package by the way, HBM and core are the same height....in theory. Hotspot still scares the shit out of me.
Derbauer put a motherboard in the dishwasher to clean of vaseline, would it help on this card?
Does LM dissolve with alcohol, would it make sense to dump it in a tub of Iso alcohol?
put liquid metal on my Vega64 when i put on a waterblock. varnished everything around it. been running fine for 3+years never goes over 45degrees and getting over 150mhz higher. ( runs up to 1800mhz though hits 1750 average.)
Arctic seems to have spring sale now, including thermal pads.
What is this about not being able to change the voltage on GPUs? is this something with RTX 3000 and RX 6000 specifically? I remember tuning voltage on 10 series and a Radeon 7
I found out that and went overboard on the voltage of the 6950xt with lowering the voltage I can have a stable clockspeed of 2550mfgz on 1120 mv with a hotspot of 87 Celsius
liquid metal on a gpu is like water in your oil you dont want to see it
Call me lucky, but I managed to remove LM from under my laptop GPU memory chip and it's still working normally. Was not easy. I then went all out with conformal coating and reapplied, all good. Only GPU I'd apply LM is a laptop's anyways, different scenario.
As someone who had a go at Liquid Metal:
Its kinda not worth it.
Good thermal paste and good contact gets you within a few degrees of Liquid Metal in my experience.
I also had issues with large temp deltas between Hotspot and Die Temp aka it sometimes wasnt easy to have good contact with the super thin Liquid Metal film instead of a blob of thermal paste.
"short" video 😂
Interesting 😁
I got about 5 or maybe 10c reduction on my flashed and watercooled 980....it really helped it clock higher. I guess that's an exception :)
Well, it's a video about shorts 😅
First system (laptop) I used LM on I was able to overclock a 680M up from 620Mhz stock to 700ish on AS5 to beyond 1GHz, without any further mods
In laptops where radiator volume and airflow is limited, LM is AMAZING
Buildzoid do you agree on the ridiculous price they are trying to launch the 6950xt? I mean will +1500 MT/s justify a 2.5x rise in price?
I wonder if an ultrasonic cleaner could get the liquid metal out from beneath the memory/core? Because we already saw with the smd components that it didn’t really deteriorate them, so maybe there’s a chance all you need to do is clean under the chips without removing them.
Yes an ultrasonic cleaner will work...People think these things are more fragile than they actually are...I have literally put Motherboards and GPUs in the dishwasher before...As long as you don't use soap and you let it dry completely afterwards it shouldn't hurt it at all...I usually give it a 99% alcohol bath also just to be sure
"That doesn't mean you shouldn't do it, I'm just saying you're be wasting your time." hehehehe
cooler, quieter. simple as that. which is a huge deal for many!
Yoo nice levelling up the thumbnail game
I run vega 56 (with blower cooler) sitting on 64s bios. So it has unlocked extra clock speeds on core and memory. And with Vegas the overall temps are kinda irrelevant becasue Hotspot temp exists. For preference overall which is shown is Radeon software 70C meanwhile hotspot is 110C. Which already is degrading chip slowly and also thermal throttles. So putting LM saved me basicly 10C in winters im living warm with my heatspreader but in summers im better touching grass outside. Temps go crazy even tho i have undervolted core and memory a bit.
The only thing that helped my Vega 64 was using kryonaut and really butter the biscuit (use a lot).
Igors lab writes that the screws opposite the PCI should be tightened down first, I do not know if it makes a difference, but I followed the guide.
I also used washers to tighten the mounting pressure, but it did not change the temps.
I removed the fans and shroud so only the heatsink was left and installed 2 noctua 140mm fans on the card instead.
Managed to keep the hotspot in the 90es after all that work.
I was not able to change the clocks in the AMD driver, the Asus bios somehow blocked it, downclocking 10MHz or something stupid like that, made the card unstable.
Installing the Asus software to adjust fans and clocks made a conflict with the AMD driver and the GPU fans turned off until the card thermal protection kicked in and ran fans at very high speed, the fans turned off at 65 degrees again and everything started over.
Wanted to try and bios flash to non Asus bios, the program denied it and I was not brave enough to push on with more tinkering after that.
I messed up witk LM..got some spreading of it ..tried to secure with spraypaint ..now when starting upp , it works normally so you can see startupp ,and sometimes the windows loadup icon ..then the screen wents black ...so i have supposed its toasted 🤷
Very insightful, thanks ! Learned a lot :-)
"GPUs don't have voltage control" well morepowertool exists and goc/classified exists on 3090/ti xoc cards.
ultrasonic cleaner could maybe clean between the mem. and PCB
Liquid metal does dry and when it does it’s the worst thing on gods green.
I LMd my 3060 Ti but it had no improvement compared to a paste.
How often to replace LM is my question?? I have 2 laptops and two delided chips I might have to re-apply soon.
Once or twice after a few months if the cooler is bare copper. If it's nickel plated, usually no additional application is needed.
If you benchmarked and have a reference you only need to see if their are out of order changes. LM doesn't degrade like other thermal paste, unless it dissolves into the metal you applied it to, mostly as eMpTy said copper comes to mind. So to speak for your delided chips if you applied the LM between chip and IHS you won't need a change anyway. For your laptop application it depends on the heat sink material and your degradation. If there is no big gap in temps over what you got when first applied you not have to change anything. If the temps were to go up you might go for a reaplication. Never change a running system. If nothing changes it stays the same.
@@Airwave2k2 Thats what I thought. Everything is good except the macbook and it could be that I don't use it enough to know if the temps went up. Thanks!
Nice, been thinking about this since I started the video lol. Nickel plated GPU cooler and temps are the same after 14 months. Had to flash a custom bios just to change fan activation and cut off as the GPU was so "cold" the fans weren't cooling the bloody GDDR6.
Doesn't msi afterburner allow you to enforce clocks @ mv
There's repair channels that basically don't work on any boards that are covered in liquid metal because there's a very high chance you end up in a never-ending rabbit hole of trying to figure out which component is dead or whatever.
I had a pretty weird liquid metal mishap a while back, I killed a 980 even though there was no liquid metal on the SMDs or anywhere else on the card.. core coverage was good but it still died leaving a scorch mark on the cooler surface. I know the liquid metal killed it I just have no idea how 🤔
If I was in the US I'd send it to you just to satisfy my curiosity.
Pretty sure Buildzoid is in Europe not in the US
@@DJKr15py UK
Was it an EVGA ACX card? My B-stock 980Ti's VRMs went up in smoke and I found out a while later that cooler did not do a good job cooling VRMs. Also, use nail polish and brush it onto the SMDs to protect them, colored if you can because you can easily see if it covers fully
@@kekoraaaa It's an MSI Gaming 4G, there's no visible carnage on the VRM or anywhere else on the board though. I had it in SLI (wtf was I thinking ^_^) with an identical card on stock paste and that card was fine. I think actually had somebody check it and they said the core is dead which sounds about right given the edge of the chip left a mark on the cooler like it ate through the nickel plating... must've been one hell of a short. RIP 980, expensive lesson learned 😁
@@kekoraaaa Yep and mine died as a result at the height of graphics card pricing which is why I have sworn off EVGA half-assed quality.
damn, that's quite a big mess, anyway for the price of liquid metal I really wonder why can't they just always ship a protective liquid like just a few small amount of grams specially made to protect stuff around, I mean the liquid metal is not exactly that expensive from what I saw people doing similar mixes and we pay a lot for 1gr so it should have come with a protective specially made product too.
People usually use nail polish, however in the past many years ago I used nail polish many times for other random stuff and for me after some time it always got fully cracked and most didn't even had heat into the equation so I actually wonder if nail polish is actually any good as a permanent protective solution against liquid metal cause I'm planning to delid a CPU, but never touch it again for the next decade unless temps really start to get strange...