In order to appreciate LM, a set fan speed needs to be set. Otherwise fan rpm scales to thermals. Clean off the contact points around the GPU and apply about 3 coats of nail hardener to act as a barrier to prevent inevitable LM spillage. I have been thinking about your parents and their house Greg. My thoughts and prayers to their devastation.
I think the variable fan speed thing was accounted for with the graph showing decibel values. The thermals should be better, the fans should be spinning slower, and the noise should be lower also
@@OTechnology I have absolutly had little beads of runoff around the die. Its nearly impossible to get the coldplate lined up exact when reassembling the card so there is potential for it to run off when you slide the plate into place. Not only that but if you have a drop (I'm talking tiny almost microscopic drops) that pool to the edge of the die vibration over time i.e. from installation, cleaning, fans running, ect can cause that bead to role into other components. I know because I have 4 gpus, 2 laptops, a ps4 pro, and 4 cpus under my belt now and I have had this happen. Luckily I cought it before it was a problem, but your statement is a little misleading and there are to many variable to account for and nail hardner or a very thin high heat tape causes no harm and takes a minute to apply.
I don't know about you but I've used liquid metal on plenty of GPUs too and never killed one or even have excess squirt out onto the substrate. I think that you're probably applying too much. I agree though for someone who's less than confident on their application that might be a good idea.
11:23 --> That's becouse you "cleaned" (more like, just wiped) the heatsink so bad that there was still a visible layer of thermal grease on it (that's why you couldn't aply th LM to the heatsink), then the heat transfer was bottlenecked by the thermal grease residue. I'm pretty sure that the results would be different if you cleaned it properly. Nice video tho.
@@knightnxk2906 Cause he obviously didn't know shit about what he was doing and even stated such.....but continued to state his findings as accurate. Botched test as Gabriel pointed out why. I also got 10c better temps with LM on a EVGA 3090 FTW3 Ultra with better OC as a result.
@@VinceCASPERPimentel Did you have to Nickel-plate it? I can't find if Nvidia GPUs (Gigabyte) have a Nickel-plated or an Aluminum base. Can you please advise? I believe it is Nickel plated but I don't want to risk it.
@@millanferende6723 should be able to tell by looking at it, nickel plating looks very different than aluminum. Don't risk it if not sure. LM will kill aluminum.
liquid metal doesn't really "run" like it looks like it would. if you use too much, sure, youre pretty much boned. but if you have a few brain cells left upstairs to bounce ideas off each other, generally you wont need to protect the smt next to the die. in the cases where you do, clear nail polish works well, but so does black high-temp silicon. that's what I used to protect the smt around the die of all the intel chips ive delidded. (LOTS of those hot bastards) and then you can use the same stuff as a thin sealant around the IHS and re-lid the thing so it looks stock and performs like a delid.
Just to clarify something: Thermal conductivity is thermal conductivity is thermal conductivity. There is no such thing as "good conductor, bad dissipator". Conduction is the process of absorbing OR dissipating heat to or from the surrounding environment, and if something is good at one, it is equally good at the other. Saying that aluminum is a better heat dissipator than copper is like saying the my well-insulated house will be better at letting out the cold since it's good at keeping the warmth in. No. It does both jobs exactly the same. The only thing other than conductivity that determines heat transfer is surface area and convection (in this case forced convection via airflow). This means that an optimal heatsink will be made from monocrystalline diamond, then comes silver, then copper, somewhere down the line; aluminum. (Maybe there are some new fangled sciency things that are better but I haven't heard of them) Aluminum is just cheaper than copper. That's the crux of it. Graphics cards manufacturers like high profit margins, and making a bigger aluminum heatsink is the way to go for that.
Hi, engineer here. Copper is a better "dissipator" of heat, heat dissipation is directly related to heat conduction; it is not itself a distinct physical property of any material. The reason fins are aluminium is because it's cheaper and lighter, while still being almost as efficient as copper. Copper and aluminium at the same temperature will "dissipate" heat at the same energy rate given identical geometries, the difference being that copper is quicker to accept energy from a source than aluminium is. So the temperature difference between the heat exchange device and the cooling fluid rises quicker in response to an input of energy from the source.
applying some nail polish over the parts you want to protect works well just incase some liquid metal spills over during the reapplication of the heatsink
@Mykel Hardin Well I didn't know. So there you have it. Just because you know doesn't mean everybody else does. And you're the only asshole around here.
I may have missed something, but we're clocks monitored/controlled? Maybe the LM created some thermal headroom which gpu boost took advantage of to boost higher? This in theory could have the same temps as noise as the stock tim
You are correct. Nvidia cards automatically boost their core clocks when they have thermal headroom via GPU Boost 3.0. So considering this he has better thermals + more performance without even realizing.
I have a watercooled 1080TI FE at @2.1GHz and got a 5C drop in temps with liquid metal. I think the reason why you didn't get a drop in temperature is because your cooler is the limiting factor. The RTX 2080 has a rather large die and can transfer the heat over a large area (unlike mainstream Intel CPUs). It could also be the fan speed, unless you monitored it and set it to the same speed for all tests. These cards have a target temp and will adjust the fan speed accordingly.
@@SemuckiProductions No he's not an mvp, all of you just upvoted him without watching the video, but it's actually BS. Applying liquid metal makes no difference.
@@Szilagyicsabaistvan youre an Idiot, not all cards are the same, for some cards the thermal bottleneck might actually be on the thermal conductivity between die and cooler, on this one it dosnt seem so, but that dosnt have to apply to every card out there
@@Szilagyicsabaistvan that guy did some realy bad job applying it. i also noticed a signifikant temperature drop on my delided cpu and a biosflashed gtx 780ti
So the clickbait screencap literally says "Bad Idea", but then the first think he mentions is that its not actually a bad idea.. And then says its actually fine.
It really depends on the graphic card. I did apply conductonaut on my MSI Sea Hawk 2080Ti and I dropped temps by 10 degrees versus factory thermal paste. Even versus kryonaut, I have 5degrees less with Conductonaut.
Just a little tip on applying liquid metal, heat up the die surface 1st then apply liquid metal. It helps in spreading it much more evenly and its easier.
btw thats not ur fault 8:18 gigbyte cards are all like that,it runs fine but the liquidmetal dose move more ,,best advice is electrical tape (33+)around the parts around the dye
I'll be honest, I do apply liquid metal on every graphics card I can. The reason? It's much quieter, and it usually boosts more. I do apply some regular thermal paste (and then spread it to avoid disturbing the LM application on the cooler) on the resistors or small components around the die to avoid shortages or spillage. It has always worked for me. Another thing to point out is that I usually sand out the cooler, so the LM can have a much better contact.
A little useful trick I found to assure the safety of the component your applying liquid metal to, make a tight border of electrical tape around the die. You can buy a wider electrical tape which is what I used.
Made a custom water loop with ~400W of dissipation from radiators for my new 3800X @ 4.3 GHz all core slightly undervolted PC. Used an EK water block and conductonaut on my Gigabyte Windforce GTX 1080 OC. Protected some of the exposed contacts with PRF 202 conformal coating spray. GPU temps went from 75°C to 45°C (in a ~28°C room, ~38°C water temp) during prolonged use with clocks from factory OC of ~1860 MHz to 2050-2063 MHz as the factory boost settings seem to have a strange effect on MSI Afterburner's offsets at times. GPU is at default voltage settings and could probably get to a stable 2100 MHz with a little bit of extra voltage, but the performance gains aren't worth it. There are hardly any temperature fluctuations as the heat is dissipated so fast from the GPU die.
i already have conductonout on my 1060 for almost a year now lol :P Using it with an Arctic Accelero Mono plus because the stock cooler absolutely sucked and started ratteling. Runns powermodded at a constant 2.1ghz 1.093 volt without any downclock ever at 64°c max. And yes its pretty silent also And the turing has pretty big die, more contact area makes using liquidmetal not worth it
Without a test with other thermal paste testing he is making an assumption based non the way the paste looks. Science Studio needs to try different thermal paste to verify.
Not really. This card is not ideal for using LM. Replacing the paste with Kryonaut would have definitely resulted in lower temps. The guy is a noob. He has absolutely no idea what he was doing.
I think you'll find the real reason aluminium is used in heatsinks is not because it dissipates heat better than copper (because it doesn't). The primary reason is cost. Copper is roughly 3.3x more dense and roughly 3.5x the cost, meaning that 1 cubic metre of copper costs roughly 13x more than the same volume of aluminium. There are of course additional benefits in that aluminium is much lighter and it's also easier to work with.
For me I go with liquid metal via a custom water cooling loop, once I've swapped from an aluminum kit to copper based and seen temps go from 60°C-68°C down to 38°C-45°C, my guess is the aluminum swap to copper helped a lot, but I still say liquid metal played a big part in this change of temps
I see a couple of issues with the testing and application here. 1. Liquid metal will stick to copper, you just have to clean it really well with alcohol first, and apply the metal as soon as it is dry. It will take longer to work out to the proper size, but it will eventually stick. 2. You need to set your fans to a constant speed, just high enough to not throttle when under load for the tradition TIM. Then do your bench mark, then use the same exact fan speed for the second test after application. Your fans may be kicking on later with liquid metal because the stock curve allows it to still get to the same temp as tradition TIM before running. The reason you may have only seen a 1c drop might be because the fans didn't need to work as hard to maintain the same temp. If you did do this, then it should be pointed out in the video.
If it wasn't properly contacting the cooler side, temps would be insanely high. Ever run a bare GPU through a stress test without proper cooler contact? Not pretty :-)
I did not protect my graphics card with nail polish prior to LM application. It was a disaster, the card ended up shorting the card and the LM was a huge pain to clean up. The card eventually worked after hours of cleaning with toothbrush, but its overclocking capacity took a huge hit.
The Issue with liquid thermal paste is that you have to apply on both sides the paste. Otherwise it can't move the heat. Therefore you have to understand the property's of the metal you apply it to. Aluminium as we all know starts to oxidize when it comes in contact with liquid thermal paste (gallium). Copper on the other hand has also an oxide layer like aluminium but It only absorbens the paste. That's why you need to renew it after some time. Try to sand away the oxide layer of cooper, clean it with ipa ,so u have really the bare metal and then it would stick to it. Not even to consider what else you have done wrong, it should really show a significant difference. I have done a lot with liquid thermal paste but with a lot more effort. Nickel plating then sandblasting gives the ultimate finish.
I applied LM to my 2080 Ti XC Ultra and temps under load went down by 8C. The XC Ultra has a very beefy heatsink on it already, the LM has some more room to stretch before the heatsink is saturated. 3dmark timespy stress test with fresh regular Arctic Silver 5 TM had final temps peak at 73C @ 130% TDP +140Core/+900 memory. Admittedly this is pretty outstanding for an air cooler already. Once LM was applied the temps peaked at 65C. Fan speed was set too 100% for both tests.
The silicon die has 3D letters printed on it. This prevents a flat contact. And throws any benefit of Liquid Metal out of the window sometimes depending on the actual GPU and cooler cold plate style. He also did not prime his heatsink with LM too. So the copper cold plate would eventually soak any LM up over time. And the GPU would slowly get hotter and hotter until it throttled.
you can take some painter's tape and lightly tape around the edges to catch any accidental spillage and make sure the liquid metal doesn't hit anything it's not supposed to.
It will apply to copper if it's properly cleaned, the pores still contain the residue from the previous compound. You must apply some press and lightly "scrub" it in the same spot to get it to adhere. I use 6000 then 8000 grit wet sand paper as the 2nd to last step when cleaning. Last step is last cleaning the copper with isopropyl alcohol. (Same same paper I used for die sanding the i9-9900KF)
@1:38. That doesn’t sound right. How can copper be a better conductor and yet not conduct heat to air better than aluminum? Only the thermal conductivity should matter for heat transfer
Im using conductonaut on my heavily overclocked 290x with accelero iv cooler, instantly dropped 13 degrees vs arctic's own supplied mx-4. Pretty worth it if you ask me
"Now if you have an old card and it's just got really chalky thermal paste and you know it needs to be replaced..." --> Sounds like an idea for a video.
Wrong. When you apply LM, you need to readjust fan curve. To compare temps you need to set fans manually to constant value, then you can see the difference.
For future reference, a little bit of sanding might be necessary (1500 grid, light) to help liquid metal stick onto copper heatpipes. Direct contact heatpipes are always tricky.
The one thing people don't understand about liquid metal cooling is that you need the correct metals to be in contact on both sides. As you've already found out copper doesn't do well with this particular liquid metal application. The reason why copper is used, is that it will wick away heat faster than aluminum but the flip side of the problem is that copper holds on to thermal energy long than aluminum. Thermal conductivity vs. thermal convection. The only thing I can think of that is great in both different ways is graphene but that's uber expensive, exotic and down right a real pain in the ass to actually use on a GPU or CPU. I can get it to work in a lab but it's not great for an every day user (yet).
I just ended scraping dried out grizzly liquid metal with razerblade from my cpu radiator just after 6 months after application. I trusted youtube videos praising its performance, nobody said it dry out and stop working altogether after three months. Repasting usually takes 10 minutes, this took over an hour. Highly NOT recommended.
This contradicts my results using LM on a 970 and rx 580. I got a 4C drop on the 970 and about a 9C drop on the 580. In both cases I saw nothing wrong with the existing stock paste. I don't remember having the same trouble spreading conductonaut on either card. Both cards have bare copper base plates so maybe the alloy used in those heat pipes is different and prevents adhesion. Also, I've found that temps with conductonaut rise over the course of several months. I don't have the same problem with coollaboratory pro (the blue cap stuff). My hypothesis is that thermal expansion and contraction from temperature cycles makes conductonaut migrate away from hot spots on the die. Coollaboratory liquid pro has microbeads that may mitigate the migration by breaking up the surface tension. So far I think this is specific to GPUs since nothing is said from people doing cpu delids, the larger die size of GPUs may give the LM enough space to clump and dry in certain areas. This is the first time I've seen someone get a 0C delta.
Just tip here, if you can't apply LM on ihs or die, try use watercolor brush apply, after apply go for another 1 or 2 drop on ihs or die middle of surface (do not spread), done.
1. The 20xx GPUs are, according to most, power limited, not thermally limited. You may get better performance with LM on a 2080 due to better efficiency at lower temperatures, but I don't think this will be much of a difference. I was able to improve performance on 10xx and 9xx GPUs by using LM because those platforms are often thermally limited. 2. Try just using masking tape around the die when applying LM. This is what I do, and it has worked great for me in a bunch of GPU and CPU applications. It's simple, clean, and works great.
Dude. Conformal coating or clear nail polish painted over the capacitors and some captains tape (the yellow transparent tape, will help keep LM off the components if it squeezes out. Also, put the drop of liquid metal onto the swab away from the card, not directly onto the die. You save yourself the risk of squirting it out all over the card
I was seeing 5c vs mx-4 on my PC PCS+ R9-390. That heat sink was flat and also the gpu ran hotter then the 2080. funny thang is it was only 1c on both of my RX-580s vs mx-4. LM does spread on bare copper, you just have to clean it really good.
Well it kinda depends, my card (an ancient XFX R7800 Series Tahiti LE) ran (with many different pastes) at an avarage of 100-105°c in CS:GO with a clock of 1100MHz (100+ OC). This was while using a custom fan profile with the fans hitting 100% at 80°c. I then replaced the paste with LM Conductonaut and set the fanspeed to auto, and now my temps, while pushing higher clocks, land at about 65-78°c during the same loads. It does make a difference, it's just not the "be all end all" solution. It is going to have a greater impact on some things/products, while giving little to no change with others.
Little tip, you can use regular thermal paste to insulate the small components, keeps them cooler too. It’s not perfect but it’s way better than leaving them naked.
I did this with a gigabyte 1080ti .. Temps actually got worse with the LM. the contact surface between cooler and die was very wide. so contact with the LM was nill. Using A lot of thermal paste corrected the issue, but cleaning off the LM was a PITA.
You have to also realise that the gpu die is huge compared to a cpu die. Having a tiny surface area on a cpu die means you need something more thermally conductive to really get a good temp drop. It would also be good to pull the card apart to verify the heatsink contact, because if the contact isn't good then it defeats the purpose of using something more thermally conductive. TL;DR- GPUs produce heat over a larger surface, and heat transfer capability of paste isn't saturated.
Also it won't break the card permanently if you get LM on the caps around the die, just spray it off good with some brake clean and it will work just fine. It happened to me on the 780, and it didn't post, so I just sprayed it down pretty well, let it dry, and then it worked fine.
You don't gain that much by applying LM on GPU because there is no die cover. Cooler has a direct contact with the die. My temps went down by 3 deg Celcius on a 1080ti (ref) with a Accelero Extreme IV. More can be gained by optimising the case air intake
Author: It does actually work. Try it on a Clevo laptop with MXM or any other GPU that has a Flat block on heatsink. The tube design may "work" but that is a ridiculously bad and cheap design for a heat spreader. Notice the contact pattern when you take it apart remaining in the factory TIM. Not ideal
I'm not a expert on this, but I think the squished copper heatpipes don't exactly help the Liquid Metal application, because it can't fill the gaps like thermal paste would.
Nail polish isn't made for high temp enviroments. That card will scorch it and it will flake off. Use UV mask. Just a little advice from a micro solder tech who sees people who tried to do this and ruined their card or spent a lot of money for me to chase the short down. I would personally, STRONGLY suggest you NOT do this. But it's your money :)
@@gamerelated3887 No need. The coldplate covers the resistors too if the thermal pads are making good pressured contact with both, and the face of the CPU is mating well with the block even better. Done it loads of times to laptops, complete PC's, audio amps, and even routers.
@@JuxZeil Right on, just posting my experiences. been micro soldering for some time now and I've seen quite a few circumstances where liquid metal got all over the boards. I would probably not recommend this to the general public. I don't doubt that you're sharp enough to make sure that it's safe for you but most people aren't.
@@gamerelated3887 True, but there's a lot of 'hands on' learners like myself out there too, and I like to help them pluck up the courage to use their hands and noggins as well as multiple tutorials. In case any budding DIY'ers are reading this comment, that includes reading the f'in manual. Don't be too prideful like the dad that has 4 screws left over and managed to break a fitting all because he didn't think he needed to read the instructions. 😔
My old 980Tis dropped 10 degrees after applying LM, this was super usefull in SLI. My current 1080Ti is also running on LM (+custom Loop) and is always running on liquid temp instead of a delta of 5C, which is also something.
Instead of going with liquid metal right away, try just putting on new thermal paste. Temps get worse are thermal paste ages and the factory applied thermal paste often isn't great either. My 1080 Ti mini dropped 5°C just by putting on new thermal paste.
That GAMING OC cooler is so nice, I got the 2080Ti and it's so quiet, looks simple yet stylish, getting around 2145MHz on the core. I've had the cooler off (so easy) and it's very beefy, the only complaint is the copper pipes (contact) have a rough finish and those gaps, but still it gets around 65C max in AC Odyssey. Will be making my own liquid cooling solution for it soon. I've been running LM on my 1080Ti FE Hybrid for over half a year and it's been running pretty damn well, saw 1-2 °C drop. I was bored and had some LM from my 8700K delid and decided to try it out! lol But AS5 is just as good and cheaper and FAR easier to apply.
I never done this but my logic says if LM doesn't make contact when you try apply with the buds(heatsink copper) . Then, the LM won't connect properly when you attach it to the gpu. Maybe that's why you don't see any thermal improvement. You might want to redo this properly and clean thoroughly before applying to both side.
I literally just tried putting liquid metal on my graphics card for the first time (R9 390) to see if it made a difference. I'm already using an nzxt kraken G10 adapter to mount a corsair H60 for cooling, so arguably I should've seen less of an improvement as I'm already using a better-than-stock cooling setup, so I wasn't expecting to see a giant leap in temp improvement from doing this. I prepared the copper heatplate surface with the steel wool that came with the LM (even wiping the TIM clean normally will leave a microscopic layer, preventing the LM from adhering, as we just saw in this video) and applied clear nail varnish to the surface mount compenents surrounding it, safety first! Just like in the video example I only applied it to the graphics core rather than applying it to the cooler as well, and I used the same fan settings (all to 100% throughout) on the same stress test. On 1080p + 8xAA furmark, max temp dropped by 15C. Did I do it perfectly and achieve the optimal contact? Almost certainly not, as I've never used liquid metal before. Did I even have the perfect mounting pressure between the waterblock and the die? Also, probably not, as you just have to take your best guess with this setup. Do I still have a lot to learn in this area? Of course, I am by no means an expert. Was it worth trying? Absolutely yes
Back when I bought my Asus GTX 970, it was definitely needed to replace it as the brand new card had a dried up thermal paste. Running with MX-4 since then with no issue.
LM has never been worth it for GPU's. They use Direct Die cooling, whereas Intel CPU's tend to get better results due to the Additional Thermal interface with the Lid. But alas its a learning experience.
I used LM on my old gtx 780 lightning, the stock thermal paste was similar to the stuff Intel puts on their CPUs. It dropped temps about 10c for me overclocked, and I was able to push it further than stock paste with the unlocked bios. However, I just bought a 980 ti and judging by its already really low temps I probably won't benefit from this much.
@@MarshalBeard Tried it on a 290x lightning, gigabyte 780 ghz edition, zotac 980, and evga 980 ti. Every time it was about a 5 degree drop over factory paste, but a good paste like gc extreme or nt-h1 would probably work just as well.
it is, the difference is you apply it once and you're done with it while paste dries up after a while and you have to redo it a couple of times in its lifetime
A follow up, I would also suggest using liquid cooling with a block that doesn't resist the liquid metal. I'm going to bet a big help when liquid cooling.
Liquid cooling my Titan X Pascal brought it from 60C over ambient and throttling down to 10-12C over ambient using Kryonaut. Swapping to Conductonaut brought it down by 2 degrees. This is while applying a +250MHz overclock. Idle temperature is 1C over ambient. Old idle temperatures were 10-15C over ambient. Noise levels also went down from 60dBa at max fan speed on the stock cooler to 42 dBa at max fan speed on the water cooler. Only problem is that with the radiator at the front, it dumps a crazy amount of heat into the case.
I do agree that it doesn't make sense BUT you could argue that if you put a waterblock so the cooler itself has much more capacity to transfer heat you may start to see some differences between normal paste and liquid metal....It's gonna be minor so it still makes no sense but I think you can at least see some difference
Also with a gpu there is a much larger surface area on the die allowing for better heat dissipation. As well, it's already direct die cooling compared to having a extra thermal barrier like a IHS
Always put a little bit of liquid metal to the q-tips supplied. Then smear it to the surface. It makes it much easier and safer do do that kind of operations.
I have a tube of conductonaut on the shelf already from my older intel systems. gives me about 5-8c difference on my 2700x using a h100i. ill put it on the rtx 2080 when the stock paste starts getting bad. I usually don't let stock paste stick around for more than a year or so. im at about 7-8 months on the 2080.
The better the cooler, the greater the improvement will be with LM. If the cooler has reached its max heat dissipation ability using standard TIM, switching to a better TIM will not help. However, on coolers that have their heat dissipation bottle-necked at the TIM, switching to LM will help. IE., you would see a larger difference on a GTX 1080 Ti FTW3, as that cooler has more ability to dissipate heat than the stock TIM can transfer to it. On the Gigabyte card you used, the cooler was already at its heat dissipation limit, and the stock TIM was good enough to meet that limit.
did this to my R9 390 I dropped 7 degrees and noticed it took marginally longer to heat up and cooled down a little faster then before, cant say its worth it for most but for me its made a noticeable difference in noise at least.
It's not that it ain't any better but the bigger difference is that you don't have to reapply as often and probably not once in its entire lifetime while with paste you have to redo it a couple of times
I use LM (Conductonaut) on my custom water cooled cards. On my RTX 2080 (EVGA XC Gaming with an EK water block) it made an 8°C difference in temperatures vs using Kryonaut. With Kryonaut, I was getting 24°C over ambient and with LM I am getting 16°C over ambient with a 100% sustained load at 2130Mhz on the core. Ambient temp in my office stays around 72°F (22°C). I saw similar results with my GTX 1080 Ti and I saw about a 6°C difference with Vega 64, though the Vega card is installed in a different system with less cooling capacity in general.
The problem with Graphics cards is that the VRMs are locked and you will not be able to push them to the point where the TIM is going to make a difference. The risk of it going over the edge of the die and shorting the capacitors just makes it not worth it. Also do not use liquid metal to shunt mod the power regulation resistors because the gallium in the liquid metal will break the solder down causing the resistors to fall off later. If you are going to shunt mod please just solder a resister in parallel.
Can't believe you did this without any research in regards to the fact that you should have used something like clear nail polish to paint over the small bits around the die to keep the LM from possibly spreading out from the die and making contact with them and causing a possible short. Also, I think there is more benefit to using it on the GPU in a laptop. I have an MSi GE-73 VR with the 6 core 8750 and a 1070 with the cpu and gpu under LM and never see temps higher than 70-73 for either the cpu of gpu which is excellent temps for a laptop.
I have LM on my watercooled 1080, temps are just awesome. I will check back in a few months when getting a RTX but so far just for the temps I Always would do it again.
In order to appreciate LM, a set fan speed needs to be set. Otherwise fan rpm scales to thermals.
Clean off the contact points around the GPU and apply about 3 coats of nail hardener to act as a barrier to prevent inevitable LM spillage.
I have been thinking about your parents and their house Greg. My thoughts and prayers to their devastation.
You don't really need to protect the SMDs from liquid metal as once you apply it if you do it properly it won't go anywhere.
I think the variable fan speed thing was accounted for with the graph showing decibel values.
The thermals should be better, the fans should be spinning slower, and the noise should be lower also
Thanks, Bob! Fan speed was a variable we monitored and represented with our sound tests. There was no significant change.
@@OTechnology I have absolutly had little beads of runoff around the die. Its nearly impossible to get the coldplate lined up exact when reassembling the card so there is potential for it to run off when you slide the plate into place. Not only that but if you have a drop (I'm talking tiny almost microscopic drops) that pool to the edge of the die vibration over time i.e. from installation, cleaning, fans running, ect can cause that bead to role into other components. I know because I have 4 gpus, 2 laptops, a ps4 pro, and 4 cpus under my belt now and I have had this happen. Luckily I cought it before it was a problem, but your statement is a little misleading and there are to many variable to account for and nail hardner or a very thin high heat tape causes no harm and takes a minute to apply.
I don't know about you but I've used liquid metal on plenty of GPUs too and never killed one or even have excess squirt out onto the substrate. I think that you're probably applying too much. I agree though for someone who's less than confident on their application that might be a good idea.
11:23 --> That's becouse you "cleaned" (more like, just wiped) the heatsink so bad that there was still a visible layer of thermal grease on it (that's why you couldn't aply th LM to the heatsink), then the heat transfer was bottlenecked by the thermal grease residue.
I'm pretty sure that the results would be different if you cleaned it properly.
Nice video tho.
@@F0x9o how can this is be a bs video if he has less that 70 degress on the gpu? haha
@@knightnxk2906 Cause he obviously didn't know shit about what he was doing and even stated such.....but continued to state his findings as accurate. Botched test as Gabriel pointed out why. I also got 10c better temps with LM on a EVGA 3090 FTW3 Ultra with better OC as a result.
@@VinceCASPERPimentel Did you have to Nickel-plate it? I can't find if Nvidia GPUs (Gigabyte) have a Nickel-plated or an Aluminum base. Can you please advise? I believe it is Nickel plated but I don't want to risk it.
@@millanferende6723 should be able to tell by looking at it, nickel plating looks very different than aluminum. Don't risk it if not sure. LM will kill aluminum.
No nail polish? This is dangerous and I LOVE IT!
My thoughts on listening to him talk about how close everything was to the die. XD
His hands still looked well manicured though.
liquid metal doesn't really "run" like it looks like it would. if you use too much, sure, youre pretty much boned. but if you have a few brain cells left upstairs to bounce ideas off each other, generally you wont need to protect the smt next to the die. in the cases where you do, clear nail polish works well, but so does black high-temp silicon. that's what I used to protect the smt around the die of all the intel chips ive delidded. (LOTS of those hot bastards) and then you can use the same stuff as a thin sealant around the IHS and re-lid the thing so it looks stock and performs like a delid.
@@andrewstubblefield4483 you're absolutely right. I thought the same. He fried the chip, and gave us those shity stats without proof 😂
yes clear nail polish is your friend
Just to clarify something: Thermal conductivity is thermal conductivity is thermal conductivity. There is no such thing as "good conductor, bad dissipator". Conduction is the process of absorbing OR dissipating heat to or from the surrounding environment, and if something is good at one, it is equally good at the other. Saying that aluminum is a better heat dissipator than copper is like saying the my well-insulated house will be better at letting out the cold since it's good at keeping the warmth in. No. It does both jobs exactly the same. The only thing other than conductivity that determines heat transfer is surface area and convection (in this case forced convection via airflow).
This means that an optimal heatsink will be made from monocrystalline diamond, then comes silver, then copper, somewhere down the line; aluminum. (Maybe there are some new fangled sciency things that are better but I haven't heard of them)
Aluminum is just cheaper than copper. That's the crux of it. Graphics cards manufacturers like high profit margins, and making a bigger aluminum heatsink is the way to go for that.
Hi, engineer here.
Copper is a better "dissipator" of heat, heat dissipation is directly related to heat conduction; it is not itself a distinct physical property of any material. The reason fins are aluminium is because it's cheaper and lighter, while still being almost as efficient as copper.
Copper and aluminium at the same temperature will "dissipate" heat at the same energy rate given identical geometries, the difference being that copper is quicker to accept energy from a source than aluminium is. So the temperature difference between the heat exchange device and the cooling fluid rises quicker in response to an input of energy from the source.
applying some nail polish over the parts you want to protect works well just incase some liquid metal spills over during the reapplication of the heatsink
@Mykel Hardin Well I didn't know. So there you have it. Just because you know doesn't mean everybody else does. And you're the only asshole around here.
this is what i was looking for lol
@@KyranSparda You have to be careful about the type of nail polish used. Forgot the reason though.
hi-temp black silicon works well too
If I put nail polish on my GTX 970, my whole house will burn in no time.
I may have missed something, but we're clocks monitored/controlled? Maybe the LM created some thermal headroom which gpu boost took advantage of to boost higher? This in theory could have the same temps as noise as the stock tim
true :)
You are correct. Nvidia cards automatically boost their core clocks when they have thermal headroom via GPU Boost 3.0. So considering this he has better thermals + more performance without even realizing.
I don't think any of these actually thermal throttle below 80°C. 🤔
Aww, Pepsi just wants to help!
he was just curious
Gj you did it wrong and assumed it didnt do anything... Thats pro for yall
I have a watercooled 1080TI FE at @2.1GHz and got a 5C drop in temps with liquid metal. I think the reason why you didn't get a drop in temperature is because your cooler is the limiting factor. The RTX 2080 has a rather large die and can transfer the heat over a large area (unlike mainstream Intel CPUs). It could also be the fan speed, unless you monitored it and set it to the same speed for all tests. These cards have a target temp and will adjust the fan speed accordingly.
EVGA 1070 FTW
Before: 78
After: 70
mvp
@@SemuckiProductions No he's not an mvp, all of you just upvoted him without watching the video, but it's actually BS. Applying liquid metal makes no difference.
@@Szilagyicsabaistvan youre an Idiot, not all cards are the same, for some cards the thermal bottleneck might actually be on the thermal conductivity between die and cooler, on this one it dosnt seem so, but that dosnt have to apply to every card out there
@@Szilagyicsabaistvan that guy did some realy bad job applying it. i also noticed a signifikant temperature drop on my delided cpu and a biosflashed gtx 780ti
So the clickbait screencap literally says "Bad Idea", but then the first think he mentions is that its not actually a bad idea.. And then says its actually fine.
*Liquid Metal on a Power Supply?*
Great Video!
Liquid metal on RAM heat spreaders!
Liquid metal on your liquid metal for MAXIMUM cooling
Transferring power using liquid metal cables could work
Not getting it, liquid metal on vrm cooling.
Liquid metal in a transformer?? Now that is a video to shoot
I've ran liquid metal on my GPU for over a year, 1080Ti, no temp increases overtime
It really depends on the graphic card. I did apply conductonaut on my MSI Sea Hawk 2080Ti and I dropped temps by 10 degrees versus factory thermal paste. Even versus kryonaut, I have 5degrees less with Conductonaut.
Just a little tip on applying liquid metal, heat up the die surface 1st then apply liquid metal. It helps in spreading it much more evenly and its easier.
next time use conformal coating to prevent shorting.otherwise good video
Is nail polish okay?
ventisca they are similar
Clear nail polish specifically right? I dont think you can use any nail polish, not sure tho.
btw thats not ur fault 8:18 gigbyte cards are all like that,it runs fine but the liquidmetal dose move more ,,best advice is electrical tape (33+)around the parts around the dye
I'll be honest, I do apply liquid metal on every graphics card I can.
The reason? It's much quieter, and it usually boosts more.
I do apply some regular thermal paste (and then spread it to avoid disturbing the LM application on the cooler) on the resistors or small components around the die to avoid shortages or spillage. It has always worked for me.
Another thing to point out is that I usually sand out the cooler, so the LM can have a much better contact.
Interesting.
A little useful trick I found to assure the safety of the component your applying liquid metal to, make a tight border of electrical tape around the die. You can buy a wider electrical tape which is what I used.
Very bad idea. I had to oven the gpu because of artifact and the nail polish destroyed it
best upgrade for a gpu would be to put a watercooled block on it. My gtx 1080 FE have gone from non-OC - 92C to OC +140/+500 - 57C.
Made a custom water loop with ~400W of dissipation from radiators for my new 3800X @ 4.3 GHz all core slightly undervolted PC. Used an EK water block and conductonaut on my Gigabyte Windforce GTX 1080 OC. Protected some of the exposed contacts with PRF 202 conformal coating spray.
GPU temps went from 75°C to 45°C (in a ~28°C room, ~38°C water temp) during prolonged use with clocks from factory OC of ~1860 MHz to 2050-2063 MHz as the factory boost settings seem to have a strange effect on MSI Afterburner's offsets at times. GPU is at default voltage settings and could probably get to a stable 2100 MHz with a little bit of extra voltage, but the performance gains aren't worth it. There are hardly any temperature fluctuations as the heat is dissipated so fast from the GPU die.
Great reply. I've got my 1080 ti on a kraken x62 with the g12 mount, wondering if LM would make a difference on it. Your comment convinced me it will.
Don't be afraid to use nail polish on the sides of the die. It prevents short crcuits with the little components.
Use nail polish or acrylic conformal coating onto all the little surface mount transitors and resistors
Liquid metal as lube
Possibly_A_Weeb It is fantastic as lube. I get 4 whole masturbates out of one application :)
My only worry is about galvanic corrosion after long term use
That's how the liquid metal terminator was made
@@RobyLV just the tip
@@brandonkinsman1270 terminator silver tip
i already have conductonout on my 1060 for almost a year now lol :P
Using it with an Arctic Accelero Mono plus because the stock cooler absolutely sucked and started ratteling.
Runns powermodded at a constant 2.1ghz 1.093 volt without any downclock ever at 64°c max.
And yes its pretty silent also
And the turing has pretty big die, more contact area makes using liquidmetal not worth it
Can we just acknowledge that the stock paste job on the card was really, really good? I was pleasantly impressed frankly.
Without a test with other thermal paste testing he is making an assumption based non the way the paste looks. Science Studio needs to try different thermal paste to verify.
Not really. This card is not ideal for using LM. Replacing the paste with Kryonaut would have definitely resulted in lower temps. The guy is a noob. He has absolutely no idea what he was doing.
Using LM on my 1080ti w/Kraken x31 lost about 5-10° from x31 original paste still working great after over 1 year now
I think you'll find the real reason aluminium is used in heatsinks is not because it dissipates heat better than copper (because it doesn't). The primary reason is cost. Copper is roughly 3.3x more dense and roughly 3.5x the cost, meaning that 1 cubic metre of copper costs roughly 13x more than the same volume of aluminium. There are of course additional benefits in that aluminium is much lighter and it's also easier to work with.
For me I go with liquid metal via a custom water cooling loop, once I've swapped from an aluminum kit to copper based and seen temps go from 60°C-68°C down to 38°C-45°C, my guess is the aluminum swap to copper helped a lot, but I still say liquid metal played a big part in this change of temps
I see a couple of issues with the testing and application here.
1. Liquid metal will stick to copper, you just have to clean it really well with alcohol first, and apply the metal as soon as it is dry. It will take longer to work out to the proper size, but it will eventually stick.
2. You need to set your fans to a constant speed, just high enough to not throttle when under load for the tradition TIM. Then do your bench mark, then use the same exact fan speed for the second test after application. Your fans may be kicking on later with liquid metal because the stock curve allows it to still get to the same temp as tradition TIM before running.
The reason you may have only seen a 1c drop might be because the fans didn't need to work as hard to maintain the same temp. If you did do this, then it should be pointed out in the video.
I think you got this kind of result because LM is not PROPERLY contacting to the cooler(copper) side.
If it wasn't properly contacting the cooler side, temps would be insanely high. Ever run a bare GPU through a stress test without proper cooler contact? Not pretty :-)
I did not protect my graphics card with nail polish prior to LM application. It was a disaster, the card ended up shorting the card and the LM was a huge pain to clean up. The card eventually worked after hours of cleaning with toothbrush, but its overclocking capacity took a huge hit.
The Issue with liquid thermal paste is that you have to apply on both sides the paste. Otherwise it can't move the heat. Therefore you have to understand the property's of the metal you apply it to. Aluminium as we all know starts to oxidize when it comes in contact with liquid thermal paste (gallium). Copper on the other hand has also an oxide layer like aluminium but It only absorbens the paste. That's why you need to renew it after some time. Try to sand away the oxide layer of cooper, clean it with ipa ,so u have really the bare metal and then it would stick to it. Not even to consider what else you have done wrong, it should really show a significant difference. I have done a lot with liquid thermal paste but with a lot more effort. Nickel plating then sandblasting gives the ultimate finish.
I applied LM to my 2080 Ti XC Ultra and temps under load went down by 8C. The XC Ultra has a very beefy heatsink on it already, the LM has some more room to stretch before the heatsink is saturated. 3dmark timespy stress test with fresh regular Arctic Silver 5 TM had final temps peak at 73C @ 130% TDP +140Core/+900 memory. Admittedly this is pretty outstanding for an air cooler already. Once LM was applied the temps peaked at 65C. Fan speed was set too 100% for both tests.
The silicon die has 3D letters printed on it. This prevents a flat contact. And throws any benefit of Liquid Metal out of the window sometimes depending on the actual GPU and cooler cold plate style. He also did not prime his heatsink with LM too. So the copper cold plate would eventually soak any LM up over time. And the GPU would slowly get hotter and hotter until it throttled.
you can take some painter's tape and lightly tape around the edges to catch any accidental spillage and make sure the liquid metal doesn't hit anything it's not supposed to.
FWIW Liquid Metal will also stain copper. It won't damage it, but it will change it's color.
It will apply to copper if it's properly cleaned, the pores still contain the residue from the previous compound. You must apply some press and lightly "scrub" it in the same spot to get it to adhere. I use 6000 then 8000 grit wet sand paper as the 2nd to last step when cleaning. Last step is last cleaning the copper with isopropyl alcohol. (Same same paper I used for die sanding the i9-9900KF)
@1:38. That doesn’t sound right. How can copper be a better conductor and yet not conduct heat to air better than aluminum? Only the thermal conductivity should matter for heat transfer
Im using conductonaut on my heavily overclocked 290x with accelero iv cooler, instantly dropped 13 degrees vs arctic's own supplied mx-4. Pretty worth it if you ask me
"Now if you have an old card and it's just got really chalky thermal paste and you know it needs to be replaced..." --> Sounds like an idea for a video.
Love the content SS!
You ran the the graphics card on auto-fan with liquid metal didn't you? That's the reason there was virtually no change in temperature.
Wrong. When you apply LM, you need to readjust fan curve. To compare temps you need to set fans manually to constant value, then you can see the difference.
It works way better on a watercooled card since on an air cooled card the bottleneck then shifts to the cooler which is undersized on most GPUs.
For future reference, a little bit of sanding might be necessary (1500 grid, light) to help liquid metal stick onto copper heatpipes. Direct contact heatpipes are always tricky.
The one thing people don't understand about liquid metal cooling is that you need the correct metals to be in contact on both sides. As you've already found out copper doesn't do well with this particular liquid metal application. The reason why copper is used, is that it will wick away heat faster than aluminum but the flip side of the problem is that copper holds on to thermal energy long than aluminum.
Thermal conductivity vs. thermal convection. The only thing I can think of that is great in both different ways is graphene but that's uber expensive, exotic and down right a real pain in the ass to actually use on a GPU or CPU. I can get it to work in a lab but it's not great for an every day user (yet).
I just ended scraping dried out grizzly liquid metal with razerblade from my cpu radiator just after 6 months after application. I trusted youtube videos praising its performance, nobody said it dry out and stop working altogether after three months. Repasting usually takes 10 minutes, this took over an hour. Highly NOT recommended.
This contradicts my results using LM on a 970 and rx 580. I got a 4C drop on the 970 and about a 9C drop on the 580. In both cases I saw nothing wrong with the existing stock paste. I don't remember having the same trouble spreading conductonaut on either card. Both cards have bare copper base plates so maybe the alloy used in those heat pipes is different and prevents adhesion. Also, I've found that temps with conductonaut rise over the course of several months. I don't have the same problem with coollaboratory pro (the blue cap stuff). My hypothesis is that thermal expansion and contraction from temperature cycles makes conductonaut migrate away from hot spots on the die. Coollaboratory liquid pro has microbeads that may mitigate the migration by breaking up the surface tension. So far I think this is specific to GPUs since nothing is said from people doing cpu delids, the larger die size of GPUs may give the LM enough space to clump and dry in certain areas. This is the first time I've seen someone get a 0C delta.
Just tip here, if you can't apply LM on ihs or die, try use watercolor brush apply, after apply go for another 1 or 2 drop on ihs or die middle of surface (do not spread), done.
Nooooo. You need to wet the copper. You absolutely CAN wet copper with LM. It must be dirty/oily. Clean it with solvents/alcohol.
1. The 20xx GPUs are, according to most, power limited, not thermally limited. You may get better performance with LM on a 2080 due to better efficiency at lower temperatures, but I don't think this will be much of a difference. I was able to improve performance on 10xx and 9xx GPUs by using LM because those platforms are often thermally limited.
2. Try just using masking tape around the die when applying LM. This is what I do, and it has worked great for me in a bunch of GPU and CPU applications. It's simple, clean, and works great.
Dude. Conformal coating or clear nail polish painted over the capacitors and some captains tape (the yellow transparent tape, will help keep LM off the components if it squeezes out.
Also, put the drop of liquid metal onto the swab away from the card, not directly onto the die. You save yourself the risk of squirting it out all over the card
8:32 I do the same I am glad you added that part. Showed it also. That will save me some time.
You need to add 3 layers of clear nail polish to the pcb around the die. That will protect the "little bits" from shorting.
I was seeing 5c vs mx-4 on my PC PCS+ R9-390. That heat sink was flat and also the gpu ran hotter then the 2080. funny thang is it was only 1c on both of my RX-580s vs mx-4. LM does spread on bare copper, you just have to clean it really good.
It was scrubbed with isopropyl.
I did as well. I had to scrub it 3 times before it would spread, i think i will try 100% acetone next time.
Well it kinda depends, my card (an ancient XFX R7800 Series Tahiti LE) ran (with many different pastes) at an avarage of 100-105°c in CS:GO with a clock of 1100MHz (100+ OC). This was while using a custom fan profile with the fans hitting 100% at 80°c.
I then replaced the paste with LM Conductonaut and set the fanspeed to auto, and now my temps, while pushing higher clocks, land at about 65-78°c during the same loads.
It does make a difference, it's just not the "be all end all" solution. It is going to have a greater impact on some things/products, while giving little to no change with others.
The fuck do you mean i already did it and BOOM EXTRA 87 FPS
Instructions unclear
Killed rtx 2080 ti
Little tip, you can use regular thermal paste to insulate the small components, keeps them cooler too. It’s not perfect but it’s way better than leaving them naked.
I did this with a gigabyte 1080ti .. Temps actually got worse with the LM. the contact surface between cooler and die was very wide. so contact with the LM was nill. Using A lot of thermal paste corrected the issue, but cleaning off the LM was a PITA.
You have to also realise that the gpu die is huge compared to a cpu die. Having a tiny surface area on a cpu die means you need something more thermally conductive to really get a good temp drop. It would also be good to pull the card apart to verify the heatsink contact, because if the contact isn't good then it defeats the purpose of using something more thermally conductive.
TL;DR- GPUs produce heat over a larger surface, and heat transfer capability of paste isn't saturated.
Also it won't break the card permanently if you get LM on the caps around the die, just spray it off good with some brake clean and it will work just fine. It happened to me on the 780, and it didn't post, so I just sprayed it down pretty well, let it dry, and then it worked fine.
please show GPUZ sensor data. Your temps are only core temps, not useful at all
Core temps aren't useful at all? What rubbish is this?
@@GregSalazar Need to see core temp and hot spot temp.
just the core temp is not enough.
You don't gain that much by applying LM on GPU because there is no die cover. Cooler has a direct contact with the die. My temps went down by 3 deg Celcius on a 1080ti (ref) with a Accelero Extreme IV. More can be gained by optimising the case air intake
Author: It does actually work. Try it on a Clevo laptop with MXM or any other GPU that has a Flat block on heatsink. The tube design may "work" but that is a ridiculously bad and cheap design for a heat spreader. Notice the contact pattern when you take it apart remaining in the factory TIM. Not ideal
I'm not a expert on this, but I think the squished copper heatpipes don't exactly help the Liquid Metal application, because it can't fill the gaps like thermal paste would.
I would have added thermal pads or nail pollish over the resistors myself, but I admire the balls it took to experiment on that card. ^_^
Nail polish isn't made for high temp enviroments. That card will scorch it and it will flake off. Use UV mask. Just a little advice from a micro solder tech who sees people who tried to do this and ruined their card or spent a lot of money for me to chase the short down. I would personally, STRONGLY suggest you NOT do this. But it's your money :)
@@gamerelated3887 No need. The coldplate covers the resistors too if the thermal pads are making good pressured contact with both, and the face of the CPU is mating well with the block even better.
Done it loads of times to laptops, complete PC's, audio amps, and even routers.
@@JuxZeil Right on, just posting my experiences. been micro soldering for some time now and I've seen quite a few circumstances where liquid metal got all over the boards. I would probably not recommend this to the general public. I don't doubt that you're sharp enough to make sure that it's safe for you but most people aren't.
@@gamerelated3887 True, but there's a lot of 'hands on' learners like myself out there too, and I like to help them pluck up the courage to use their hands and noggins as well as multiple tutorials.
In case any budding DIY'ers are reading this comment, that includes reading the f'in manual. Don't be too prideful like the dad that has 4 screws left over and managed to break a fitting all because he didn't think he needed to read the instructions. 😔
My old 980Tis dropped 10 degrees after applying LM, this was super usefull in SLI. My current 1080Ti is also running on LM (+custom Loop) and is always running on liquid temp instead of a delta of 5C, which is also something.
Instead of going with liquid metal right away, try just putting on new thermal paste. Temps get worse are thermal paste ages and the factory applied thermal paste often isn't great either. My 1080 Ti mini dropped 5°C just by putting on new thermal paste.
i thought the video was messed up with the black screen at the start.....refreshed my page like 3x lol
That GAMING OC cooler is so nice, I got the 2080Ti and it's so quiet, looks simple yet stylish, getting around 2145MHz on the core.
I've had the cooler off (so easy) and it's very beefy, the only complaint is the copper pipes (contact) have a rough finish and those gaps, but still it gets around 65C max in AC Odyssey.
Will be making my own liquid cooling solution for it soon. I've been running LM on my 1080Ti FE Hybrid for over half a year and it's been running pretty damn well, saw 1-2 °C drop.
I was bored and had some LM from my 8700K delid and decided to try it out! lol But AS5 is just as good and cheaper and FAR easier to apply.
I never done this but my logic says if LM doesn't make contact when you try apply with the buds(heatsink copper) . Then, the LM won't connect properly when you attach it to the gpu. Maybe that's why you don't see any thermal improvement.
You might want to redo this properly and clean thoroughly before applying to both side.
0:59 which case is that in the corner by chance?
I literally just tried putting liquid metal on my graphics card for the first time (R9 390) to see if it made a difference. I'm already using an nzxt kraken G10 adapter to mount a corsair H60 for cooling, so arguably I should've seen less of an improvement as I'm already using a better-than-stock cooling setup, so I wasn't expecting to see a giant leap in temp improvement from doing this. I prepared the copper heatplate surface with the steel wool that came with the LM (even wiping the TIM clean normally will leave a microscopic layer, preventing the LM from adhering, as we just saw in this video) and applied clear nail varnish to the surface mount compenents surrounding it, safety first! Just like in the video example I only applied it to the graphics core rather than applying it to the cooler as well, and I used the same fan settings (all to 100% throughout) on the same stress test. On 1080p + 8xAA furmark, max temp dropped by 15C. Did I do it perfectly and achieve the optimal contact? Almost certainly not, as I've never used liquid metal before. Did I even have the perfect mounting pressure between the waterblock and the die? Also, probably not, as you just have to take your best guess with this setup. Do I still have a lot to learn in this area? Of course, I am by no means an expert. Was it worth trying? Absolutely yes
Der 8auer used nail polish and applied that to the smaller bits around the GPU die. That way they're protected, just in case.
Back when I bought my Asus GTX 970, it was definitely needed to replace it as the brand new card had a dried up thermal paste. Running with MX-4 since then with no issue.
I've tried it before and was kind of unimpressed. Little messy and didn't drop temps as much as I hoped
LM has never been worth it for GPU's. They use Direct Die cooling, whereas Intel CPU's tend to get better results due to the Additional Thermal interface with the Lid. But alas its a learning experience.
I used LM on my old gtx 780 lightning, the stock thermal paste was similar to the stuff Intel puts on their CPUs. It dropped temps about 10c for me overclocked, and I was able to push it further than stock paste with the unlocked bios. However, I just bought a 980 ti and judging by its already really low temps I probably won't benefit from this much.
@@MarshalBeard Tried it on a 290x lightning, gigabyte 780 ghz edition, zotac 980, and evga 980 ti. Every time it was about a 5 degree drop over factory paste, but a good paste like gc extreme or nt-h1 would probably work just as well.
i used LM on my r9 390. it dropped from 87 degrees under load to 69.
it is, the difference is you apply it once and you're done with it while paste dries up after a while and you have to redo it a couple of times in its lifetime
A follow up, I would also suggest using liquid cooling with a block that doesn't resist the liquid metal. I'm going to bet a big help when liquid cooling.
Liquid cooling my Titan X Pascal brought it from 60C over ambient and throttling down to 10-12C over ambient using Kryonaut. Swapping to Conductonaut brought it down by 2 degrees. This is while applying a +250MHz overclock.
Idle temperature is 1C over ambient. Old idle temperatures were 10-15C over ambient.
Noise levels also went down from 60dBa at max fan speed on the stock cooler to 42 dBa at max fan speed on the water cooler. Only problem is that with the radiator at the front, it dumps a crazy amount of heat into the case.
I do agree that it doesn't make sense BUT you could argue that if you put a waterblock so the cooler itself has much more capacity to transfer heat you may start to see some differences between normal paste and liquid metal....It's gonna be minor so it still makes no sense but I think you can at least see some difference
I'm watching this with the mantra "conformal coating, use conformal coating" going through my head...
Also with a gpu there is a much larger surface area on the die allowing for better heat dissipation. As well, it's already direct die cooling compared to having a extra thermal barrier like a IHS
Always put a little bit of liquid metal to the q-tips supplied. Then smear it to the surface. It makes it much easier and safer do do that kind of operations.
I have a tube of conductonaut on the shelf already from my older intel systems. gives me about 5-8c difference on my 2700x using a h100i. ill put it on the rtx 2080 when the stock paste starts getting bad. I usually don't let stock paste stick around for more than a year or so. im at about 7-8 months on the 2080.
The better the cooler, the greater the improvement will be with LM. If the cooler has reached its max heat dissipation ability using standard TIM, switching to a better TIM will not help. However, on coolers that have their heat dissipation bottle-necked at the TIM, switching to LM will help. IE., you would see a larger difference on a GTX 1080 Ti FTW3, as that cooler has more ability to dissipate heat than the stock TIM can transfer to it. On the Gigabyte card you used, the cooler was already at its heat dissipation limit, and the stock TIM was good enough to meet that limit.
The proper way to make sure liquid metal doesn't spread over the die and onto the little smds is to apy some form of conformal coating to it.
did this to my R9 390 I dropped 7 degrees and noticed it took marginally longer to heat up and cooled down a little faster then before, cant say its worth it for most but for me its made a noticeable difference in noise at least.
It's not that it ain't any better but the bigger difference is that you don't have to reapply as often and probably not once in its entire lifetime while with paste you have to redo it a couple of times
that moderboard lighting is awsome ,what procesor there is in ?
I use LM (Conductonaut) on my custom water cooled cards. On my RTX 2080 (EVGA XC Gaming with an EK water block) it made an 8°C difference in temperatures vs using Kryonaut. With Kryonaut, I was getting 24°C over ambient and with LM I am getting 16°C over ambient with a 100% sustained load at 2130Mhz on the core. Ambient temp in my office stays around 72°F (22°C). I saw similar results with my GTX 1080 Ti and I saw about a 6°C difference with Vega 64, though the Vega card is installed in a different system with less cooling capacity in general.
The problem with Graphics cards is that the VRMs are locked and you will not be able to push them to the point where the TIM is going to make a difference. The risk of it going over the edge of the die and shorting the capacitors just makes it not worth it. Also do not use liquid metal to shunt mod the power regulation resistors because the gallium in the liquid metal will break the solder down causing the resistors to fall off later. If you are going to shunt mod please just solder a resister in parallel.
The point about thermal paste is that it's non reactive. Galinstan metals seem more trouble than they're worth.
You didn't check to see what the frequency graph was. Could be possible its boosting a bit higher negating and temperature decrease
If anyone does this, put some electricians tape around those smaller surrounding components while you apply the metal.
Or nail polish. Like I did in my other GPU LM vids.
Can't believe you did this without any research in regards to the fact that you should have used something like clear nail polish to paint over the small bits around the die to keep the LM from possibly spreading out from the die and making contact with them and causing a possible short. Also, I think there is more benefit to using it on the GPU in a laptop. I have an MSi GE-73 VR with the 6 core 8750 and a 1070 with the cpu and gpu under LM and never see temps higher than 70-73 for either the cpu of gpu which is excellent temps for a laptop.
What about clockspeeds? Have they stayed the same before and after?
Yes. Kept pivoting between 1875 and 1890.
I have LM on my watercooled 1080, temps are just awesome. I will check back in a few months when getting a RTX but so far just for the temps I Always would do it again.
2 years later....
When he put the liquid metal on the copper it looked like flubber xD