You’re a great content creator. I think I speak for most people when I say I appreciate your apology. It speaks volumes that you realize a mess up and actually take accountability. No one does that now. You’re a real stand up guy.
@@dasiro The detective shows like to say, "enhance" the image, but that's not really a thing. All they can do is zoom in and out. Family Guy pointed this out in one of its episodes.
And what have I learned today? That scanning electron microsopes are the size of desktop printers now. Albeit desktop printers from the early 2Ks. That is absolutely wild to me. The last one I used was in college and it took up floor to ceiling
TBF the whole thing is a bit bigger than it may appear in that shot. The white cabinet and section above are all part of it but it's still pretty wild that they're this convenient now.
There is no need to feel bad for being human, or having an off day. Glad to hear the viewing community was constructive enough to point you in the right direction. That saved you a lot of headache, heartache, and time. You give us so much. I think it is great everryone tried to give back. And am very happy to hear you got straight to the actual problem. Not so sure that "fix" is good enough to last for long, but it certainly proved the point. Nice to have closure. But WOW, an SEM? What a toy to get to play with. Please don't mind me, I'm just jealous!
Would have been really interesting to see performance difference between putting it back together with liquid metal & if you machine the IHS to clean up the corroded surface!
Don't think it matters much, I had a similar issue, the ihs had black spots, cleaned the best I can and then applied new liquid metal, temps seemed about the same when I applied liquid metal for the 1st time. And 1 year after nothing wrong with it.
If you machine the IHS you will need to do the same with the edges or the distance between it and the CPU cores will be to much and then you will have to make a whole new hold down mechanism since the IHS will be thinne and you won't get enough preassure to be able to get good connection with the conntacts on the motherboard. As default, you have to screw down the hold down mechanism with a specific torque value so a modified IHS would not work directly. Though they could maybe use something else than liquid metal to avoid it degrading more.
It would be nice to see a side-by-side comparison with a part of the heatspreader that was not affected by the liquid metal, because on that scale almost every material looks weird. Also: this would be a very good example of why NOT to LM your CPU if you want to use it more than a few years or sell it afterwards.
If the cpu is soldered, yes I agree. But if you have an older CPU that used thermal paste like 8th gen Intel, LM will not react with the leftovers of the soldering process, because there's none.
@TigTex It's with the IHS, doesn't matter if soldered or TM. Anyone else reading this take note of OPs solution to just sell it off and pass the problem to someone else.
Never put much thought into it , but the micro scope views and analysis of what the black material was actual composed of really impressed me .Clearly i must watch more videos. Well done .
I want to thank you for your fantastic Kryosheet! Today i applied the Kryosheet to my PNY 4070 Ti XLR8, and the max temps drops from 83C (Hotspot 95C) to 74C (Hotspot 84C). It already had the pumpout issue after less than a year from the factory.
@@offspringfan89 Temps could probably be better, seems my coldplate surface is flaking nickel or something, the affected area is larger than the actuall GPU die, so no idea. But the temps are just fine as is now, and i never have to look at them again :)
I cleaned a de-lidded 7950x3d and the Myrco cooling block and applied conduction air extreme to both and had to grab something else form the tool cart and my cat jumped her the case and looked down inside, I ran over and chased him off and put it all together, not seeing the small fluff he left had drifted to my chip. I was extremely surprised when my rig thermalled down before it completely booted and took awhile to figure out. Finally took apart to re paste and there was a small dark wad in the middle.
The integrity here is immense, thank you for your honesty even if it's something a trivial as not getting to the bottom of why the CPU was overheating. You could see how genuine you were being, and seeing this for a mere RUclips video inspires great confidence in the Thermal Grizzly brand. I'm never had any issues and the performance to price difference is crazy. This man is on a mission to hold the PC industry to a new standard!
"So we put the sample in our SEM..." hol up. THIS is the kind of content I LOVE to watch. You need to find a 13900KS that has problems and do a microscope look with the SEM at the on-die cache and just a general look at the VIAs. It would be amazing to see... and see if corrosion is partly to blame.
Ahh good to see getting to the bottom of the issue... Now the IHS ... if liquid metal caused the damage to the IHS by eating into the metals and amalgamating with it. Wouldnt putting more gallium liquid metel onto it just make it worse? The damage has been done but the process was likely slowed down by the gallium already amalgamating itself where it ate into it. But by cleaning it, you just added fresh gallium onto an already corroded part, tis just going to eat deeper and quicker into the metal. Wouldnt it be better to try and resolder the IHS back on? Also ever since liquid metal interfaces were being popularised, this amalgamation issue with gallium was always on my mind, and I didnt trust its use... I guess this issue youve come into is just proof that liquid metal isnt a long term solution to cpu cooling.
Yes, in the long term this will be a problem again. At which point you scrub off all the oxide again and do the whole process again. The gamble now is that the useful life for this CPU is short enough that the dealing with the metal corrosion takes less time than trying to resolder the IHS. Honestly, that sounds really hard and I have no idea how feasible it is.
yes, it needs to be replaced with a direct die solution or another ihs, gallium will do this again or speed up the destruction procesd, in a year or two it will make craters on that ihs
This is a complicated topic but even on bare copper after a few applications the amalgamation slows down considerably. I'd expect a similar outcome in most cases. What's notable here is the ammount of oxidation and the interaction with the gold layer. It looks this amalgamation caused this layer to weaken and seperate from surface below, which I assume is the same nickel plating around it. I'd imagine sealing off the IHS can help minimize oxidation problems, which Roman didn't seem to do and the area is larger than most CPUs. However I've used LM under an unsealed IHS for many years and they weren't without some issues but this is quite extreme. The oxidation reaction of galinstan can effectively become non-self-limiting with temperature in certain conditions that are above my head and I'm guessing the interaction with the gold played played a role.
The Ga forms an amalgam about 80-100um into the copper. This loss of Ga changes the composition of the LM left behind and it looks "dry" and chunky. I clean this off with isopropyl and reapply fresh LM which dissolves any residue left into it (i.e. no need to be abrasive on it). Then the application lasts years (Been using LM on laptop copper heat sinks for a decade now). This does not happen as much with Nickel plated heatsinks But whenever there is too big a gap, LM can definitely oxidise and "dry out" like this
Admittedly, I was a bit disappointed when I saw the last video blow off the issue rather than go into a deep investigation. This seems much more on track with your channel history. Thank you for digging deeper into your own issue 😂 we the community greatly appreciate the quality content
Hardware Unboxed as well. They and Gamers Nexus have a pretty solid relationship. Basically use each other as a check on results, which results in greater reliability of their findings. If one reports on something, you can bet the other is not far behind.
tbh Roman, everyone is guilty of making a mistake! but taking accountability and admitting it was laziness that caused the issue is something that you don't see often so I commend you for your honesty :D In addition, a mistake is often a very good teacher and something which imo gives many of us viewers interesting behind the scenes look at the nerdy tech stuff that many of us want to see and can only dream of seeing in person one day :D
The magnification your SEM pulled off is impressive and history channel might use it on ET theories and proof of life on Mars and others. I hope someday we come across kryosheet like material for delidded CPU and IHS instead of liquid metal or liquid solder. After using carbonaut on my laptops and non messy solution.
Excellent follow-up, Roman! 👍 That's what we are here for - not for an always flawless presentation (such a thing does not exist, even the best of us make mistakes), but for owning it up, being honest about is ans then following up with a corrected and/or extended analysis.
Anyone remembers Coollaboratory Liquid Ultra paste that used mercury early on? That thing used to eat tops of chips and heatspreaders even faster than Gallium based paste.
I wouldn't beat yourself up too much about it, because it was still a decent video showing different investigative steps. It just stopped too early, which this follow-up resolves!
It's good that you found the problem. This might be why my 9900KS slowly is starting to get worse and worse temperatures as well. It's being replaced in 1-2 months with a new AM5 system, so I haven't really spent alot of time trying to fix it. It's close to 5 years old at this point, and the LM application might need a little refresh. I will take a look when the new system is up and running.
Google says you have to reapply heat paste every "few" years. And, if your temps are rising, it's probably time. I have a 9th gen i7 - no problems so far.
Der8auer 2018 "Gallium in general also reacts with copper and leaves it pitted on the very top layer, but the reaction is not heavy so doesn't really matter to Thermal Grizzly. Nickel has the function of a diffusion barrier and doesn't react with the Gallium, which makes it a lot more suitable for the application" What metal does AMD use for its IHS? Seems like a hefty tradeoff to swap soldered indium for gallium. Scraping the solder is going to also create gaps in the nickle plating. I'm surprised the entire IHS didn't crumble.
supposedly is copper, wity a layer of nickel to avoid these chemical reqctions, sanding it to remove the gold and the old solder didnt ended in the best results it seems 😅
You're misunderstanding. The extreme failure you're referring to is a phenomenon that occurs primarily with aluminum. All IHS are nickel-plated copper and always have been. The issues comes down to the surface finish, scratches, metal impurities, oxidation, etc. All of which can be mitigated to some degree. In this case the gold layer also caused problems and no attempts were made to seal off the IHS from the surrounding air. It was never in danger of losing structure stability. Anyone wanting to rely on LM long term needs to understand some maintenance will likely be required at some point. In good conditions it could be years before notable problems develop.
@@bsz6328 What do you mean by eaten away? Discoloring the nickel is common and there's certainly copper underneath otherwise you wouldn't have a cooler at this point.
@@JJFX- I mean the flat surface is very far from flat now. LM did not just discolor the surface, but I got an about half a mm deep, ~50mm wide crater on the coldplate, bare copper everywhere. It's an el cheapo block, mayb its nickel plating wasn't the best. On my GPUblock there aren't any visible mechanical changes. As far as my experience goes...
Really interesting to see the long term effects of liquid metal. This makes me wonder if the PS5 is going to suffer the same fate down the line as Sony used LM instead of thermal paste for their console.
the ps5 never had a gold layer thst was scrapped to put liquid metal, the problem here was that the surface was not meant to enter in contact with liquid metal pst being repaired on alpt of repair channels do not show any of the problems shown here the ihs on these cpys came soldered, to remove it it had to be heat to melting point of the metals used to solder it on place, those metals were removed but made the surface far from ideal for this liquid metal
The main potential issue with the PS5 is leaking but that's far more likely to be caused by the user opening it. The last drama around that was overblown. The console community just doesn't have as much experience in these topics. Not only was the heatsink intended for this use case, they applied a thick foam gasket around it to minimize the chance of leaks and oxidation. Sure, it may need reapplying at some point down the road but what you're seeing here shouldn't happen.
thanks for the follow up! the video about that cpu was... weird, for your standards. you surely have your reasons but after the current one it feels complete. you're the guy who actually looks into problems and i love you for that!
It's not chemical reaction..gallium still diffuses into the copper through the nickel plating because it's so thin (electrochemically plated on IHS) so it diffuses and leaves the indium behind
I just wonder… why delidding in the first place? Pure indium has better thermal conductivity than galinstan (sic!). Also, being solid, there is less mess with diffusion and accidental spillage. Do I miss something?
I applied liquid metal to my delidded 4570K back then. The first application of the stuff seems to age within a couple of months, as part of it oxidizes and another diffuses into the IHS. When you clean the surfaces and reapply some fresh LM after that thas happened, it stays long-term stable. I still got the CPU and MB, still have great temps.
My experience exactly The initial application changes composition and loses effectiveness as the Ga is lost into the heatsink surface Had some installs last over 3 years with no temp degrading, when the fit is really good (my experience is mostly with laptops)
I use Flitz polish to clean LM off copper and nickel. When I reapply conductonaut every 2 years or so, Flitz is INCREDIBLE. It's also great for getting LM off AIO cold plates.
Those scans were really cool! I wish I had tools like that to play with...Every single Cpu block or IHS I have ever put liquid metal on has done this exact thing...I normally just lap the metal a bit with 2000grit sand paper and reapply LM...The better temps are worth the hassle to me
G'day Roman, Thanks for showing us what the cause actually was because it is really interesting & educational to see long term scenarios of products like Thermal Paste/Liquid Metal
I notice elevators have certificates showing when they were inspected. That should be done for PC, although perhaps not on a government regulated level. An info card, maybe printed in business-card style with the dates of what services had been done and when those services were done, with a windowed slot for that card on the case, could be extremely useful. Why remember when it can be written down? It was interesting to see the effects of oxidization and alloy formation in the liquid metal through a microscope. I'll never mess with it.
Good job on the proper check up! I think you went above and beyond! I remember wondering about the TIM but then excluding it as a cause assuming it was soldered and wouldnt explain the raised power draw. Now we know the cause, and I still find it curious that the power draw is that high. AMD cpus take into account their temp and with higher temp comes more required Wattage for the same workload. Ironically this leads to a downward spiral. Seems 'naked' chips overcompensate for the lack of cooling.
I saw this same result for long-term use - 5 years - of conductonaut, even on standard nickel-plated copper blocks, it's just slower, and any mechanical wear from mount/unmount cycles will greatly accelerate it. The top of my Der8auer Edition 8086k looks similar, the degradation is just not as severe. The block it was mated to looks very close. That's why for long-term deployment I only use PTM7950 now, and keep a big syringe of kryonaut for bench testing.
I wondered about longevity of devices that use liquid metal especially with it entering the mainstream with Sony using liquid metal on Playstations. Even Asus has been using it in their laptops. I always thought Sony used it to get a slight bit more performance out of their silicon until the process improved and they would revert back to thermal paste and those early devices they didn't expect to survive long enough for corrosion to become a problem.
@Javadamutt Do not buy the devices, they will fail after warranty due to thermal issues unless remedied. While not to the extent of the capacitor plague, it's going to continue to grow in impact in the coming years.
I added liquid metal interfaces to a bunch of my machines when I first got it, since the thermals were so good, but I keep seeing stuff like this and dreading that I've got bad times in my future.
Idk why, but having LM on copper heatsink for over 2 years, the temperatures are still superb to any thermal paste, even with corrosion. I basically cleaned and polished the copper twice, but temps didn't change at all.
I am not a technical tech person, I video edit, yes I trouble shoot as much as I can, and learn a lot about from trusted RUclips channels, this one, jayz2cents , gamers Nexus and some others. Just kinda wanted to say thanks for us less knowledgeable folks that just want to build good stable systems for what we need them to do. Also at fair pricing. You guys are like my personal tech team. Everyone of you has gotten me through some problem
As always a great video! I wonder if a pass with a laser ablation machine is doable on the interior side of the IHS to restore it to peak condition, i guess it will depend on the depth of that infiltration
Looks to me that the uneven IHS caused air gap which allowed the LM to corrode and stop transferring heat. You need to clean the glue off the PCB corners, I'd take the IHS lip down a tiny bit to close the gap to the dies further.
So gluing back the IHS is a rather good advise, as this would limit the amount of oxygen to enter the liquid metal. Which would slowdown the oxidation process by some years maybe. Now we need a glue that is a very good barrier against oxygen.
This. Whatever the chemical reaction was exactly, anytime you have oxidation occur it's because air was introduced. The IHS coming apart was hint 1. Seeing the black oxidized LM was hint 2. Those 2 things alone are bad but there could be others. Easy to see/understand with hindsight. lol But still an interesting video and nice to see Roman get his rig back. Assuming he finds a way to seal everything tighter/more permanently.
The PS5 uses liquid metal with a foam barrier just to avoid shorts. Glue and airtightness seems unnecessary. Maybe just a smaller fraction of gallium would be ideal. You could raise the melting point to 90C so it flows during initial burn in and then never again, sort of like Intel's STIM, but hopefully thinner.
You are too self-critical and apologize too many times, we are not used to this on RUclips! You are truly one of the best and most responsible and at the same time modest content creators on RUclips, continue with the excellent work you are doing!
I've had this happen with my Titan xP. I use an aftermarket cooler and initially used LM. After a while, it started running super hot. Took it apart, saw the state of the heatsink, got rid of the liquid metal and used some Thermal Grizzly instead.
if you ever doing a lot of "soldering" , and if you have sandpapered the copper nail on the solder tools so much that it has no more plating left on its surface(naked copper) you will notice that when you do melt a lot of tin, the chemical substance in the tin would react to the naked copper and form black residues that inhibit heat transfers. that doesnt happen when your solder tool's copper nail still well plated.
Well done. I had the same problem with after 2 years of operation with bare copper water block on a Radeon 290X. I always thought it was the (very poorly) home made LQ that caused it. But now it seems it just sped up the reaction. But this dosen´t explain why the outlet water temperature was more than 30 degrees higher than coming in. The thermal images clearly show a massive difference. Did you forget to connect the pump on that one occasion?
this was actually REALLY interesting! I had NO IDEA about the embrittlement of nickel ... I figured it was immune. but I now realize that AMD uses INDIUM and not gallium on their IHS... but I know that some aftermarket liquid metal is an indium-gallium alloy... I wonder if that still causes nickel to break down over time too.... AMD is able to put that gold layer to prevent this, but what are people supposed to do at home???
I was not expecting that to be the issue, like most of us I Completely missed that the CPU had been de-lidded. I'm surprised that there still isn't a long term fix for liquid metal corrosion, I'd thought nickel plating was enough. Makes me worry about the long term future of products like the PS5, I'm guessing they definitely won't end up common retro gaming consoles like all the previous generations, since the majority will kill themselves over time.
@der8auer-en I read through most of the comments on the last video when you uploaded, and while there were a lot of incorrect suggestions, several of us thought this had to be the delidded die. Diagnostics often result in unexpected faults, continue the great work, no apology needed. Gallium etching has its applications, unfortunately in this case unwanted.
I am surprised that he is surprised that liquid metal react to other metals like that. Even Linus from LTT warned people that liquid metal can react but didn't know how much.
Glad to see it was a "simple" fix. I love seeing this kind of stuff, zooming into the cracks and see them turn into canyons, valleys and mountains is ... ominous. Which makes honey i shrunk the kids even more terrifying lol. Laziness will always bite us :D What are your plans to fix this? Make a delid bracket for the threadripper?
Very interesting video thanks a lot for doing this, it shows nickel is not enough for unlimited use some day it will fail to stop gallium, and the package power decreases for some reason when there is less heat is also very goofy.
I can relax, now that there's a resolution. Phew! I really needed that, after being left up in the air since last week. Incidentally, I've never heard the word: 'embrittlement' come out of the mouth of a native English speaker, which is not to say that it's incorrect, rather, that your English seems better than that of most people I know. Kudos!
Watching fixer upper videos keeps reminding me i got take MB out & fix a few pins not all ram slots work Led magnifying glass lamp is all i have to see with 3x 10x cant use my cheap microscope plugs into Pc .
I had almost exactly the same thing on my 8700k, the HS had sort of bonded with the liquid metal, and had a lot of buildup and pitting on it. Took close to 5 years for it to get that bad and run into too high temps/instability for me.
I don't think you messed up. I quite enjoyed the "here's the problem, here's some theories, I'll get back to you with what it was later." It was fun to theorize about what was going on! Which means I'm still super interested in if both AIOs had the temperature difference across the input/output that you saw with the one that was originally on it? The second one you swapped out to test if it was an AIO fault might have been fine as the CPU was the issue, but was the first AIO pumping really slow ASWELL?! I just can't make the maths on the physics make sense unless the flow rate is
Your entitled to be a little bit lazy you upload in 2 languages bro. I appreciate your honesty
Seeing the gallium dendrites that had got into the micro-cracks into the IHS and pushed them apart was wild.
It's the whiskers channel. Cat whiskers, metal whiskers...
@@elgonzo7239 Yep, that's the term we use in relation to titanium whiskers.
You’re a great content creator. I think I speak for most people when I say I appreciate your apology. It speaks volumes that you realize a mess up and actually take accountability. No one does that now.
You’re a real stand up guy.
He does everything!! From being a content creator to a engineer!! As well as manufacturing!!
He was sitting down in the video though.
I'll see myself out.
7:10 enhance, Enhance, ENHANCE!!
except this is not image enhancement, but REAL zoom
@@dasiro The detective shows like to say, "enhance" the image, but that's not really a thing. All they can do is zoom in and out. Family Guy pointed this out in one of its episodes.
@@briankleinschmidt3664 the movie 'Super troopers' enhance scene
Sum1 watches SMA
They can enhance a 4-pixel photo into a regognizable face. I've seen it on TV!
I appreciate the follow up and the science. Fixing a PC that has been giving you grief is a great feeling.
And what have I learned today? That scanning electron microsopes are the size of desktop printers now. Albeit desktop printers from the early 2Ks. That is absolutely wild to me. The last one I used was in college and it took up floor to ceiling
TBF the whole thing is a bit bigger than it may appear in that shot. The white cabinet and section above are all part of it but it's still pretty wild that they're this convenient now.
Depends on scAnning or tinneling
There is no need to feel bad for being human, or having an off day. Glad to hear the viewing community was constructive enough to point you in the right direction. That saved you a lot of headache, heartache, and time. You give us so much. I think it is great everryone tried to give back. And am very happy to hear you got straight to the actual problem.
Not so sure that "fix" is good enough to last for long, but it certainly proved the point. Nice to have closure.
But WOW, an SEM? What a toy to get to play with. Please don't mind me, I'm just jealous!
I love the honesty. Others could/should learn from you.
Your fave toy is a Scanning Electron Microscope. Awesome.
Absolutely, complete pro.
Would have been really interesting to see performance difference between putting it back together with liquid metal & if you machine the IHS to clean up the corroded surface!
seeing a funny moments youtuber commenting on a derbauer video was the last thing i was expecting to see today, but im happy to see it.
Don't think it matters much, I had a similar issue, the ihs had black spots, cleaned the best I can and then applied new liquid metal, temps seemed about the same when I applied liquid metal for the 1st time. And 1 year after nothing wrong with it.
@@aFeectgod damn, GTA SA
The issue is that the oxidation-resistant layer is destroyed. You could polish it away, but would expose the bare copper in the process.
If you machine the IHS you will need to do the same with the edges or the distance between it and the CPU cores will be to much and then you will have to make a whole new hold down mechanism since the IHS will be thinne and you won't get enough preassure to be able to get good connection with the conntacts on the motherboard. As default, you have to screw down the hold down mechanism with a specific torque value so a modified IHS would not work directly. Though they could maybe use something else than liquid metal to avoid it degrading more.
It would be nice to see a side-by-side comparison with a part of the heatspreader that was not affected by the liquid metal, because on that scale almost every material looks weird.
Also: this would be a very good example of why NOT to LM your CPU if you want to use it more than a few years or sell it afterwards.
If the cpu is soldered, yes I agree. But if you have an older CPU that used thermal paste like 8th gen Intel, LM will not react with the leftovers of the soldering process, because there's none.
@TigTex It's with the IHS, doesn't matter if soldered or TM.
Anyone else reading this take note of OPs solution to just sell it off and pass the problem to someone else.
Never put much thought into it , but the micro scope views and analysis of what the black material was actual composed of really impressed me .Clearly i must watch more videos. Well done .
I want to thank you for your fantastic Kryosheet!
Today i applied the Kryosheet to my PNY 4070 Ti XLR8, and the max temps drops from 83C (Hotspot 95C) to 74C (Hotspot 84C). It already had the pumpout issue after less than a year from the factory.
11°C drop in junction temperature is huge.
@@offspringfan89 Temps could probably be better, seems my coldplate surface is flaking nickel or something, the affected area is larger than the actuall GPU die, so no idea. But the temps are just fine as is now, and i never have to look at them again :)
Those microscope scans would make amazing desktop background pics.
It is cat's fault... cat hair underneath, making CPU not having a proper contact with the water block!
I have had this happen. Seriously. 😅
Oof of the floof.
That has actually happened to me!
I cleaned a de-lidded 7950x3d and the Myrco cooling block and applied conduction air extreme to both and had to grab something else form the tool cart and my cat jumped her the case and looked down inside, I ran over and chased him off and put it all together, not seeing the small fluff he left had drifted to my chip. I was extremely surprised when my rig thermalled down before it completely booted and took awhile to figure out. Finally took apart to re paste and there was a small dark wad in the middle.
Conductonaut
The integrity here is immense, thank you for your honesty even if it's something a trivial as not getting to the bottom of why the CPU was overheating. You could see how genuine you were being, and seeing this for a mere RUclips video inspires great confidence in the Thermal Grizzly brand. I'm never had any issues and the performance to price difference is crazy. This man is on a mission to hold the PC industry to a new standard!
"So we put the sample in our SEM..." hol up. THIS is the kind of content I LOVE to watch. You need to find a 13900KS that has problems and do a microscope look with the SEM at the on-die cache and just a general look at the VIAs. It would be amazing to see... and see if corrosion is partly to blame.
Ahh good to see getting to the bottom of the issue...
Now the IHS ... if liquid metal caused the damage to the IHS by eating into the metals and amalgamating with it. Wouldnt putting more gallium liquid metel onto it just make it worse?
The damage has been done but the process was likely slowed down by the gallium already amalgamating itself where it ate into it. But by cleaning it, you just added fresh gallium onto an already corroded part, tis just going to eat deeper and quicker into the metal.
Wouldnt it be better to try and resolder the IHS back on?
Also ever since liquid metal interfaces were being popularised, this amalgamation issue with gallium was always on my mind, and I didnt trust its use... I guess this issue youve come into is just proof that liquid metal isnt a long term solution to cpu cooling.
Yes, in the long term this will be a problem again. At which point you scrub off all the oxide again and do the whole process again.
The gamble now is that the useful life for this CPU is short enough that the dealing with the metal corrosion takes less time than trying to resolder the IHS. Honestly, that sounds really hard and I have no idea how feasible it is.
yes, it needs to be replaced with a direct die solution or another ihs, gallium will do this again or speed up the destruction procesd, in a year or two it will make craters on that ihs
@@MD_Builds RIP the days when Intel used thermal paste instead of solder. The change must have caused a lot of problems for TG.
This is a complicated topic but even on bare copper after a few applications the amalgamation slows down considerably. I'd expect a similar outcome in most cases.
What's notable here is the ammount of oxidation and the interaction with the gold layer. It looks this amalgamation caused this layer to weaken and seperate from surface below, which I assume is the same nickel plating around it.
I'd imagine sealing off the IHS can help minimize oxidation problems, which Roman didn't seem to do and the area is larger than most CPUs. However I've used LM under an unsealed IHS for many years and they weren't without some issues but this is quite extreme. The oxidation reaction of galinstan can effectively become non-self-limiting with temperature in certain conditions that are above my head and I'm guessing the interaction with the gold played played a role.
The Ga forms an amalgam about 80-100um into the copper. This loss of Ga changes the composition of the LM left behind and it looks "dry" and chunky. I clean this off with isopropyl and reapply fresh LM which dissolves any residue left into it (i.e. no need to be abrasive on it). Then the application lasts years
(Been using LM on laptop copper heat sinks for a decade now).
This does not happen as much with Nickel plated heatsinks
But whenever there is too big a gap, LM can definitely oxidise and "dry out" like this
Admittedly, I was a bit disappointed when I saw the last video blow off the issue rather than go into a deep investigation. This seems much more on track with your channel history. Thank you for digging deeper into your own issue 😂 we the community greatly appreciate the quality content
I call those fake sticks "dust covers"
I call them "heat insulators"
Or dummy ram, i have too, 2 sticks to cover 2 slots within my 4 slots, since i have dual channel only
I call them "A Proletarian's Pride"
I call them "First world problems"
Corsair money burn pro xtreeeeam xt +
Just switch to a 14900k. I heard those are rock solid...
🤣
🤣🤣🤣🤣🤣🤣
Well they do a really good impersonation of a Brick I'm told......
Yeah they are solid rocks
I love Roman's intellectual honesty. This and Gamer's Nexus are the only channels I "trust" for hardware.
Hardware Unboxed as well. They and Gamers Nexus have a pretty solid relationship. Basically use each other as a check on results, which results in greater reliability of their findings. If one reports on something, you can bet the other is not far behind.
@@richardbeckenbaugh1805I was just about to say the same thing...Hardware Unboxed is definitely on that list for me!
HOLY CRAP!!!! You have an SEM?!?!?! BRO!!!!! WOW!!! How are you not playing with all those toys EVERY DAY?!?!?!? lol
Glad you figured it all out :)
Electron scan everything from now on, just to be sure 🤩
A hardness or toughness test of the dark areas would be interesting compared to normal areas.
A real man is not known by how he starts, but how he finishes.
the microscope picture is gorgeous, keep that as a backround
tbh Roman, everyone is guilty of making a mistake! but taking accountability and admitting it was laziness that caused the issue is something that you don't see often so I commend you for your honesty :D
In addition, a mistake is often a very good teacher and something which imo gives many of us viewers interesting behind the scenes look at the nerdy tech stuff that many of us want to see and can only dream of seeing in person one day :D
LM is quite unpredictable. I stopped using it because of that.
The magnification your SEM pulled off is impressive and history channel might use it on ET theories and proof of life on Mars and others.
I hope someday we come across kryosheet like material for delidded CPU and IHS instead of liquid metal or liquid solder. After using carbonaut on my laptops and non messy solution.
Excellent follow-up, Roman! 👍
That's what we are here for - not for an always flawless presentation (such a thing does not exist, even the best of us make mistakes), but for owning it up, being honest about is ans then following up with a corrected and/or extended analysis.
Anyone remembers Coollaboratory Liquid Ultra paste that used mercury early on? That thing used to eat tops of chips and heatspreaders even faster than Gallium based paste.
I wouldn't beat yourself up too much about it, because it was still a decent video showing different investigative steps. It just stopped too early, which this follow-up resolves!
It's good that you found the problem. This might be why my 9900KS slowly is starting to get worse and worse temperatures as well. It's being replaced in 1-2 months with a new AM5 system, so I haven't really spent alot of time trying to fix it. It's close to 5 years old at this point, and the LM application might need a little refresh. I will take a look when the new system is up and running.
Google says you have to reapply heat paste every "few" years. And, if your temps are rising, it's probably time. I have a 9th gen i7 - no problems so far.
Der8auer 2018 "Gallium in general also reacts with copper and leaves it pitted on the very top layer, but the reaction is not heavy so doesn't really matter to Thermal Grizzly. Nickel has the function of a diffusion barrier and doesn't react with the Gallium, which makes it a lot more suitable for the application"
What metal does AMD use for its IHS? Seems like a hefty tradeoff to swap soldered indium for gallium. Scraping the solder is going to also create gaps in the nickle plating. I'm surprised the entire IHS didn't crumble.
supposedly is copper, wity a layer of nickel to avoid these chemical reqctions, sanding it to remove the gold and the old solder didnt ended in the best results it seems 😅
You're misunderstanding. The extreme failure you're referring to is a phenomenon that occurs primarily with aluminum. All IHS are nickel-plated copper and always have been.
The issues comes down to the surface finish, scratches, metal impurities, oxidation, etc. All of which can be mitigated to some degree. In this case the gold layer also caused problems and no attempts were made to seal off the IHS from the surrounding air. It was never in danger of losing structure stability.
Anyone wanting to rely on LM long term needs to understand some maintenance will likely be required at some point. In good conditions it could be years before notable problems develop.
Well, my coldplate was nickel plated in theory, but now nickel got eaten away. Interestingly my gpublock is intact sortof.
@@bsz6328 What do you mean by eaten away? Discoloring the nickel is common and there's certainly copper underneath otherwise you wouldn't have a cooler at this point.
@@JJFX- I mean the flat surface is very far from flat now. LM did not just discolor the surface, but I got an about half a mm deep, ~50mm wide crater on the coldplate, bare copper everywhere. It's an el cheapo block, mayb its nickel plating wasn't the best.
On my GPUblock there aren't any visible mechanical changes.
As far as my experience goes...
Really interesting to see the long term effects of liquid metal. This makes me wonder if the PS5 is going to suffer the same fate down the line as Sony used LM instead of thermal paste for their console.
Probably, and it's probably done with intent, for applying LM is a much more intricate process than TP application
the ps5 never had a gold layer thst was scrapped to put liquid metal, the problem here was that the surface was not meant to enter in contact with liquid metal
pst being repaired on alpt of repair channels do not show any of the problems shown here
the ihs on these cpys came soldered, to remove it it had to be heat to melting point of the metals used to solder it on place, those metals were removed but made the surface far from ideal for this liquid metal
The main potential issue with the PS5 is leaking but that's far more likely to be caused by the user opening it. The last drama around that was overblown. The console community just doesn't have as much experience in these topics.
Not only was the heatsink intended for this use case, they applied a thick foam gasket around it to minimize the chance of leaks and oxidation. Sure, it may need reapplying at some point down the road but what you're seeing here shouldn't happen.
thanks for the follow up! the video about that cpu was... weird, for your standards. you surely have your reasons but after the current one it feels complete. you're the guy who actually looks into problems and i love you for that!
Integrity & honesty is hard to find on the internet. Appreciate all of your work and content.
It's not chemical reaction..gallium still diffuses into the copper through the nickel plating because it's so thin (electrochemically plated on IHS) so it diffuses and leaves the indium behind
I just wonder… why delidding in the first place? Pure indium has better thermal conductivity than galinstan (sic!). Also, being solid, there is less mess with diffusion and accidental spillage. Do I miss something?
@@argoneumHe probably delidded it to see inside for a video originally then used it for an editing rig after rather than sitting on a shelf.
@@v3xx3r Thanks :)
@v3xx3r It was to promote the delidding process and sell the tools and LM products needed to perform the modification.
@@argoneumif cpu is soldered, delliding is only worth it if you are going to use a cooler on the die directly, otherwise it's stupid
I applied liquid metal to my delidded 4570K back then. The first application of the stuff seems to age within a couple of months, as part of it oxidizes and another diffuses into the IHS. When you clean the surfaces and reapply some fresh LM after that thas happened, it stays long-term stable. I still got the CPU and MB, still have great temps.
My experience exactly
The initial application changes composition and loses effectiveness as the Ga is lost into the heatsink surface
Had some installs last over 3 years with no temp degrading, when the fit is really good (my experience is mostly with laptops)
I use Flitz polish to clean LM off copper and nickel. When I reapply conductonaut every 2 years or so, Flitz is INCREDIBLE. It's also great for getting LM off AIO cold plates.
I mean, it happens. You're stressed and can't remember something from 4 years ago. But you owned this and gave us a result. Well done
Glad that you revisited this and analyzed what went wrong.
Those scans were really cool! I wish I had tools like that to play with...Every single Cpu block or IHS I have ever put liquid metal on has done this exact thing...I normally just lap the metal a bit with 2000grit sand paper and reapply LM...The better temps are worth the hassle to me
G'day Roman,
Thanks for showing us what the cause actually was because it is really interesting & educational to see long term scenarios of products like Thermal Paste/Liquid Metal
I notice elevators have certificates showing when they were inspected. That should be done for PC, although perhaps not on a government regulated level. An info card, maybe printed in business-card style with the dates of what services had been done and when those services were done, with a windowed slot for that card on the case, could be extremely useful. Why remember when it can be written down? It was interesting to see the effects of oxidization and alloy formation in the liquid metal through a microscope. I'll never mess with it.
Good job on the proper check up! I think you went above and beyond!
I remember wondering about the TIM but then excluding it as a cause assuming it was soldered and wouldnt explain the raised power draw.
Now we know the cause, and I still find it curious that the power draw is that high.
AMD cpus take into account their temp and with higher temp comes more required Wattage for the same workload. Ironically this leads to a downward spiral. Seems 'naked' chips overcompensate for the lack of cooling.
I saw this same result for long-term use - 5 years - of conductonaut, even on standard nickel-plated copper blocks, it's just slower, and any mechanical wear from mount/unmount cycles will greatly accelerate it.
The top of my Der8auer Edition 8086k looks similar, the degradation is just not as severe. The block it was mated to looks very close.
That's why for long-term deployment I only use PTM7950 now, and keep a big syringe of kryonaut for bench testing.
I wondered about longevity of devices that use liquid metal especially with it entering the mainstream with Sony using liquid metal on Playstations. Even Asus has been using it in their laptops.
I always thought Sony used it to get a slight bit more performance out of their silicon until the process improved and they would revert back to thermal paste and those early devices they didn't expect to survive long enough for corrosion to become a problem.
@Javadamutt Do not buy the devices, they will fail after warranty due to thermal issues unless remedied. While not to the extent of the capacitor plague, it's going to continue to grow in impact in the coming years.
I had completely forgotten you delidded that CPU. Glad people pointed it out because it was cool looking at all this stuff.
Everybody has their slip-ups. Glad to see it was a simple fix and not an actual issue with the CPU or motherboard.
I added liquid metal interfaces to a bunch of my machines when I first got it, since the thermals were so good, but I keep seeing stuff like this and dreading that I've got bad times in my future.
not gonna lie, I started sweating when you took out the threadripper cpu *while* filming
Idk why, but having LM on copper heatsink for over 2 years, the temperatures are still superb to any thermal paste, even with corrosion.
I basically cleaned and polished the copper twice, but temps didn't change at all.
I am not a technical tech person, I video edit, yes I trouble shoot as much as I can, and learn a lot about from trusted RUclips channels, this one, jayz2cents , gamers Nexus and some others.
Just kinda wanted to say thanks for us less knowledgeable folks that just want to build good stable systems for what we need them to do. Also at fair pricing. You guys are like my personal tech team. Everyone of you has gotten me through some problem
As always a great video! I wonder if a pass with a laser ablation machine is doable on the interior side of the IHS to restore it to peak condition, i guess it will depend on the depth of that infiltration
Humility is so comforting.
the 10K magnification of the IHS would make a neat background
Looks to me that the uneven IHS caused air gap which allowed the LM to corrode and stop transferring heat. You need to clean the glue off the PCB corners, I'd take the IHS lip down a tiny bit to close the gap to the dies further.
Great recovery and work on this derBauer.
So gluing back the IHS is a rather good advise, as this would limit the amount of oxygen to enter the liquid metal.
Which would slowdown the oxidation process by some years maybe. Now we need a glue that is a very good barrier against oxygen.
This.
Whatever the chemical reaction was exactly, anytime you have oxidation occur it's because air was introduced.
The IHS coming apart was hint 1. Seeing the black oxidized LM was hint 2. Those 2 things alone are bad but there could be others. Easy to see/understand with hindsight. lol
But still an interesting video and nice to see Roman get his rig back. Assuming he finds a way to seal everything tighter/more permanently.
@@shawnduffy279 5:37 It was really an eyeopening revelation that the oxygen levels (88%) were so super high!
Ooo, interesting idea. Could also use nitrogen as a purge gas before glueing it down...
The PS5 uses liquid metal with a foam barrier just to avoid shorts. Glue and airtightness seems unnecessary. Maybe just a smaller fraction of gallium would be ideal. You could raise the melting point to 90C so it flows during initial burn in and then never again, sort of like Intel's STIM, but hopefully thinner.
You are too self-critical and apologize too many times, we are not used to this on RUclips! You are truly one of the best and most responsible and at the same time modest content creators on RUclips, continue with the excellent work you are doing!
I've had this happen with my Titan xP. I use an aftermarket cooler and initially used LM. After a while, it started running super hot. Took it apart, saw the state of the heatsink, got rid of the liquid metal and used some Thermal Grizzly instead.
Very interesting and satisfying conclusion to this issue. File this under "There are no miracles."
i feel bad for that guy that thought that the cpu was pulling more than 280w or over 400amps (in bios)
Those zoomed pictures look amazing!
Very interesting! Love getting to see these more unusual failures, the electron microscope shots looked incredible! 😅
if you ever doing a lot of "soldering" , and if you have sandpapered the copper nail on the solder tools so much that it has no more plating left on its surface(naked copper) you will notice that when you do melt a lot of tin, the chemical substance in the tin would react to the naked copper and form black residues that inhibit heat transfers.
that doesnt happen when your solder tool's copper nail still well plated.
I have an i7 8700k delided by you (bought in early 2018 on caseking) still going strong to this day. Never reopened
around 4.5y the liquid metal would be absorbed in the nickel in the IHS. You'll have to reapply it.
I love pictures from scanning electron microscopes. Its always so neat, its just too bad they're so expensive. So, thank you for that 😄
Amazing how small & affordable Electron Microscopes have gotten since I first heard of them in the 80s!
Well done. I had the same problem with after 2 years of operation with bare copper water block on a Radeon 290X. I always thought it was the (very poorly) home made LQ that caused it. But now it seems it just sped up the reaction.
But this dosen´t explain why the outlet water temperature was more than 30 degrees higher than coming in. The thermal images clearly show a massive difference. Did you forget to connect the pump on that one occasion?
I thought things were slowly heating up, I never felt so close - I couldn't feel us drifting apart.
Hmm maybe silicone based thermal paste isn’t so bad after all...
that and not delid, wonderful results that do not eat the ihs 😂
I known something that oxidize more than this IHS......intel.
this was actually REALLY interesting! I had NO IDEA about the embrittlement of nickel ... I figured it was immune. but I now realize that AMD uses INDIUM and not gallium on their IHS... but I know that some aftermarket liquid metal is an indium-gallium alloy... I wonder if that still causes nickel to break down over time too.... AMD is able to put that gold layer to prevent this, but what are people supposed to do at home???
I was not expecting that to be the issue, like most of us I Completely missed that the CPU had been de-lidded. I'm surprised that there still isn't a long term fix for liquid metal corrosion, I'd thought nickel plating was enough. Makes me worry about the long term future of products like the PS5, I'm guessing they definitely won't end up common retro gaming consoles like all the previous generations, since the majority will kill themselves over time.
Working on SEM was great and super interesting. I'd love to do it couple more times
@der8auer-en I read through most of the comments on the last video when you uploaded, and while there were a lot of incorrect suggestions, several of us thought this had to be the delidded die.
Diagnostics often result in unexpected faults, continue the great work, no apology needed.
Gallium etching has its applications, unfortunately in this case unwanted.
I haven't bothered with LEDs in memory since Crucial discontinued their 'Tracer' line.
I am surprised that he is surprised that liquid metal react to other metals like that. Even Linus from LTT warned people that liquid metal can react but didn't know how much.
I'm glad you listen to the comments but there's no need to apologise Roman, great content as always, that's why we love you!
??? Whenever a person f*cks up they MUST apologize.
I disagree. It's important to stay humble. If you don't do it voluntarily, the world will do it for you.
Whhoooo boy, leaving out the german techsphere... that must have caused a full blown panic sh*tstorm at AMD HQ.
You're awesome! Great to see some "proper" engineering/science. It's rare anyone does anything like this.
Glad to see it was a "simple" fix. I love seeing this kind of stuff, zooming into the cracks and see them turn into canyons, valleys and mountains is ... ominous. Which makes honey i shrunk the kids even more terrifying lol.
Laziness will always bite us :D What are your plans to fix this? Make a delid bracket for the threadripper?
Very interesting video thanks a lot for doing this, it shows nickel is not enough for unlimited use some day it will fail to stop gallium, and the package power decreases for some reason when there is less heat is also very goofy.
Once again, you show how to own up to your mistakes. Props, bud.
I can relax, now that there's a resolution.
Phew!
I really needed that, after being left up in the air since last week.
Incidentally, I've never heard the word: 'embrittlement' come out of the mouth of a native English speaker,
which is not to say that it's incorrect,
rather, that your English seems better than that of most people I know.
Kudos!
I hear it used relativelt often when talking in a more technical setting.
I love the 7950 Gx2 in the back! I have one in 2007
Thank you for investigating it further. Really good video!
My guts told me it was a cooling issue, but not knowing the previous delidding I had no idea what it could be ..
Watching fixer upper videos keeps reminding me i got take MB out & fix a few pins not all ram slots work
Led magnifying glass lamp is all i have to see with 3x 10x cant use my cheap microscope plugs into Pc .
Incredible views from the electron microscope! Very helpful and educational. Now i got more reasons to avoid liquid metal
any video with electron microscope is worth watching :-)
Wow, fantastic! Ive never seen a SEM cpu analysis
It's almost always user error with these weird things (hard on my ego when I am the user). Good job on finding the issue.
Thx for the update, I was losing sleep over this. Not really but I did think about it a few times past week!
If the Nickel is eaten away I would did surface polishing and put the Kryosheet and forget the CPU second time.
I had almost exactly the same thing on my 8700k, the HS had sort of bonded with the liquid metal, and had a lot of buildup and pitting on it. Took close to 5 years for it to get that bad and run into too high temps/instability for me.
I don't think you messed up. I quite enjoyed the "here's the problem, here's some theories, I'll get back to you with what it was later." It was fun to theorize about what was going on!
Which means I'm still super interested in if both AIOs had the temperature difference across the input/output that you saw with the one that was originally on it? The second one you swapped out to test if it was an AIO fault might have been fine as the CPU was the issue, but was the first AIO pumping really slow ASWELL?! I just can't make the maths on the physics make sense unless the flow rate is
owning your mistake and putting it online.. respect