Gadgetory

one second guys we need to get the CPU out we keep it in here but there's a there's a security hole so I got to get it out of there today we're doing a little bit of a looser format for a video the plan is to talk about what we've learned from d-lighting at this point a pretty good amount of processors including 379 80 xc70 900 X so 960 X the 80s 700 K 7700 K and a couple others for other youtubers as well so there's a lot of small stuff we've learned here primarily centering around resealing in a normal user environment you're probably going to reseal the lid to the CPU somehow after you delete it because otherwise you can't turn it vertically and you have to hold on to the IHS against the substrate all the time which doesn't really user-friendly so we talked about things we've learned there and also going through a couple of thermal notes on performance when deleting helps when it doesn't and what you can expect overall before that this video is brought to you by thermal Grizzly makers of the conductor not liquid metal that we recently used to drop 20 degrees off of our coffee leak temperatures thermal grizzly also makes traditional thermal compounds we use on top of the IHS like cryo not and hydro not pastes learn more at the link below so this isn't a deal it tutorial so much as it is the things do after you've taken the lid off and what you need to know couple things here with d-lighting there's a lot of rhetoric around this process where it's kind of gotten a bit extreme where because people see media outlets doing it and you see sometimes 20 degrees Celsius improvements you'll see you need to deal in to overclock or it has to be deleted to perform well thermally that's simply not true for the current generation of CPUs Intel can absolutely do better with how they're transferring their energy from the die to the IHS but you don't actually have to deal it in 8700 K it's got HCC CPU thermal paste on there now so the high core count stuff from the 79 80 X II is on the 8700 K as far as we're aware and that helps a lot with thermal performance but you do want to do later once you start pushing higher overclock so as you get towards for example 4.6 4.7 gigahertz on the i9 series it's kind of impossible to keep it cool without either extreme cooling or a deal it and if the D letting gets you anywhere from 10 to 20 degrees it's instantly worth it it's also helpful because if you're trying to keep lower noise levels by lowering the fan speed you can help that process by deleting the CPU the thermal transfer is a lot better so it gets rid of the heat more efficiently and you can run lower fan RPMs so that's another benefit so you don't strictly need to deal it but if you're pushing higher over clocks you want lower fan noise or you want to buy a smaller cooler then there's advantage to D letting the trick though is what to do sort of after you've deleted the processor because applying liquid metal isn't too hard but it is something that I struggled with the first time I did it and it takes a bit of a practiced hand or eye to really figure out and there not a lot of great YouTube videos out there to show the amount of liquid metal to use without it becoming just a pool of electrically conductive potentially damaging liquid metal so we'll show some of that today this is an i-9 7900 X CPU that I previously deleted it has never been resealed so after we deleted it we left the the silicon adhesive off of it and it's as simple as just like physically pulling the IHS off to open it up again so for this one because I never resealed it I can just open it like that and this one hasn't been open since it was sealed this is pretty normal for how it looks after liquid metal has been applied and you've removed the eye adjust again so we have liquid metal on both sides here things to know so this processor we only scraped off a very little amount of this silicon adhesive and as we iterated our process for deleting things kind of learned a few tricks about this the silicon adhesive in this case doesn't really impact they're almost to negatively there's not a lot of it it's not too gunked-up and ultimately what you're dealing with is a CP that's going to be left loose we're not we never resealed it so it wasn't a big deal but if you want to improve well two things here if you have to reseal it because you're planning to use it in a computer or transport it or anything like that where it might be vertical you don't want to worry about it we've been using this jb weld it's just a silicon adhesive they're not a sponsor or anything I just bought it on Amazon we previously used UHU high-temperature sealant the same thing couple problems with this stuff it likes to explode out of the back so it's kind of annoying to use resealing it's pretty annoying too because it gets all gunked up but it comes out like a slope black glue basically and that's what they use for the most part it's a high temperature silicone adhesive and to get it on there you really want to get some of this old stuff off so you can use your fingernail for that you can grab something that's capable of scraping and just kind of pick at it get it out of there the trouble is when you do this you really need to be careful of a few things one of them is the small surface mount components around the parameter of the CPU so we have these capacitors and resistors around the edges and though is if if you're going out this with the scraper with your fingernail or whatever and you pull against one of those by accident it's very likely you're gonna rip it off there's it's not a lot of holding them on there so be careful when doing that you can see some of the silicone adhesive right here is right up against the component this is an area I would leave it alone or do my best to avoid coming into contact with those so that's one major thing to look out for the most in the delayed process the most likely thing to get damaged is one of these small components and the most likely cause of that damage is you not the Delian tool so be careful scraping stuff off and also when you're dealing with small components don't panic if one comes off chances are it's a capacitor if it's a capacitor the processor will still work almost definitely you might lose some ripple suppression or some overclocking head room but it'll still worry it's not the end of the world if you if you pull something else off there might be over but don't panic too much if something does come off so that's some of the basics in an unstructured way of what to be careful of and things like that we'll get to applying the adhesive in a moment but first let's look at the liquid metal side of things if you want to remove liquid metal conduct or not the thermal grisly stuff comes with these black q-tips that are actually really good they're the cotton schemes whatever reason the cotton seems tighter wound on these than what we normally buy so we can find a q-tip with like a tightly wound whatever it is on the end there that's not gonna come off and cotton candy style that would be ideal and you can just put some rubbing alcohol on the end of it so we're just gonna go right down the middle and take some of this off and you'll see that it's pretty cooperative you can actually see some of the liquid metal still shiny on there this is stuff you don't really want to get it on your skin it is an irritant it's probably I've gotten it on my skin I've been fine how your skin reacts is different potentially but you know just try and keep it off as in general as rule another thing is if you have blobs of this stuff you can actually siphon it back up into the tube so we can demo that here just by pulling the the putting the nozzle on so let's try and save from this blob there it goes it's gone got that one too so it's pretty easy to get it back in there if you needed to but we don't we don't need to reclaim this we have plenty more conduct or not so I'm just gonna wipe this off with a paper towel and this stuff can really if it's been on there awhile like this has it can stick a bit so sometimes I'll take the IHS and just run it under water for a second and dry it thoroughly afterwards just a high pressure water we'll get the rest off with some manual scrubbing with your thumbs or something or a sponge but you can see how we've got some residual stuff still on there and it will come off it's just a matter of how much effort it is to get that back off okay so if we wanted to really be thorough on this I'd go blast it with some water and dry it off and clean it a bit more but the purpose isn't to reseal it right now it's to do demos so what I'm going to do next let's just kind of pretend I wash the rest of that off I want to do next to show you how much liquid metal to apply from a lot of testing where we've done a lot of trial and error I've had deleted CPUs that just simply would not perform well they would perform worse than stock and that's because I often didn't have enough liquid metal on there so you need liquid metal on both sides you need it on the IHS and on the die and that's because of the surface tension of liquid metal blob just fell happy that's because of the surface tension of the liquid metal you'll need on both sides otherwise it's just not going to make contact it won't transfer it'll be worse than Tim next thing is so we're gonna to apply some liquid metal all I'm gonna do is take the syringe and just kind of apply some downward force on it like that and eventually you'll get a drop that comes out so we just have one drop on the end there I don't need to use the plunger just kind of apply gravity and that's going to be pretty close to enough for and i7 kb lake style cpu where you're dealing with a much smaller surface area for this we're gonna do a little bit more because it is an i-9 CPU so we're gonna do two drops for this one as opposed to just maybe one of those on KB Lake and then to spread the stuff you can use the same q-tip I was recommending a moment ago and I actually to show you kind of how long this stays good I've had this liquid metal on here for probably two months at this point and it's still like I can still apply it to services it goes on like pain just like that and it's still perfectly good that's been on there for at least a month at this point so you get use out of these there's a lot of discussion online about liquid metal evaporating that's not really what happens some types of liquid metal can crack over time conductor not specifically should be good for about at least two years we have some stuff in long term thermal testing that's been going for several months now and it's thermally the same as it was when the test started so it lasts a lot longer than some of the comments would have you believe but it depends a lot on what you use so yeah this is we've got some applied here and when you want to move it around it's just like paint and because of the way because of the surface tension of liquid metal if you have some on the tip of the spreading surface it'll behave differently than if you don't sometimes it'll pull the liquid metal back up sometimes it'll spread it out so this is pretty good right here and you're going to want to apply it in a way that it's just over the dye obviously and when you REBOUND this IHS it will have to go in the same orientation as when it came off so a good way to tell that on I nine CPUs is to look at the RFID chip in the corner here and that RFID chip has a cutout in the IHS for clearance right here in this corner it's the only corner with a cutout so you would just align this one on here like that and that'd be the end of that the KB Lake and other CPUs have other indicators you'll have to look at those on a per CPU basis but this is a pretty good amount of spread I would be happy with this it would perform well if you kind of we can try and get some shots of it it's not liquidy enough to drip if I hold it vertically it's not gonna drip down but it is liquidy enough that you can kind of see it shimmer a bit on the service as opposed to a bunch of streaks which would be too thin would be the opposite problem you don't have enough on there and I'm going to show you what we don't want on this CPU so there's kind of a pool in the middle here and this is just left over from when we pulled it off that pool would be too much it'll spread out it'll come out around the edges of the dye and then you risk shorting components around them to prevent shorting you can grab a nail polish there are certain types of nail polish you don't want to use we have some information and some of our old articles I'll link one below if I remember too that will tell you what to look for in a nail polish for this you'll sound a bit weird if you go to a beauty salon and ask for a nail polish that is not electrically conductive or something that won't corrode your substrate but if you use our information it should help out so you would take a little a tiny amount of this we already have some on there too much honestly I put too much on there because this is the first one I ever did but you take a tiny amount of this and you would find whatever component you want to cover up and you just apply it like that that's it's kind of on the heavy side but the point of this is to build a bit of a shield and encasing or a moat around electrically around components that are electrically active and then if the liquid metal spills over the side it will never get to those components it'll contact the the nail polish and as long as you buy something that can withstand a high temperature which this stuff can you'll be good to go so that's a tip to protect the components and then at this point you would just spread the rest of it but we don't need to do that of course you don't have to use a liquid metal you can use something like cryo knot or some other MX whatever five some other compound that's not electrically conductive it'd just be a paste but the problem then is you don't get as much benefit so how do you get benefit out of D Lydian if you're just using thermal paste well ideally you scrub a bunch of this adhesive off because this adhesive creates space in between the IHS and the die and that's not gonna help you out thermally it's actually gonna make transfer more difficult and will either require more compound or well yeah it'll just straight require more compound and more interface the transfer through is worse you want as little paste or liquid metal on there as possible because ideally you just want die to copper contact not die to compound that's a tenth of a millimetre thick to copper contact or whatever so in an ideal world you'd have no compound but it's not an ideal world so we have to have something and this stuff I tend to just scrape it off with my thumbnail because it's it's safe I can control it a little more easily and what we've learned we have some charts on this we can show too what we've learned is that you really with these I nines all I need to take off one layer and you can leave the other layer on there as a guidance for when you remount the IHS so here are some numbers for that previously we worked on Kyle's CPU bit wit Kyle we deleted his 79 80 II and worked on an r6e motherboard with 1.2 4 volts intentionally torturous over voltage really and 4.5 gigahertz clock and we observed peak temperatures of 104 degrees Celsius which is t.j.maxx when sealed with a heavy amount of silicone adhesive in fact it was actually technically worse than the Intel stock Tim and adhesive in some cases we also observed improvements upwards of 20 degrees Celsius by using a lighter seal strictly on the upper layer of the IHS for the GN 79 ad AXI we were at 75 degrees on sealed versus 94 degrees peak with the stock Tim so that's another 20 degrees difference not quite as warm as Kyle CPU but their different CPUs so there's some disparity there as well so for the adhesive we'd recommend removing just one layer leave the other as guidance you can always undo it and remove the other layer later because ultimately removing both layers we found is less than a 3 degree improvement it's it tends to be like 2 which is kind of within an error of testing so it's not really an improvement and all you're trying to do is just get enough off that you can close it without too much interference so I'm gonna scrape one corner off and show you how to apply the adhesive so this is about where I would stop for cleaning the corner we've got most of the adhesive off it's just kind of a smudge there now and then for applying new adhesive I only do it on the IHS so what we're gonna do is just kind of look at where that is that's gonna be on this corner up here so I'm going to apply some adhesive just on this boot sort of on each one of these is what I would do you can also do on the inner layer here but we've found it less important so let's get some of that going ok so I'm just gonna open this lid and stuff smells typically you would use the nozzle but we're not we're intentionally applying very little for demonstration purposes so see let's off so I don't have to smell it anymore you probably want to wear some kind of paper mask or something if you're working closely with this stuff and you're leaning over the CPU but so let's just put some on one of these corners this one's pretty scrubbed off already I'll put some there I have found that so if you're using it like a a what do they call them if you're doing like piping like like baking piping with it that's a little bit too heavy where if you kind of go along the edge with the nozzle it will actually be too thick that would be about the heavy seal what we did with Kyle's and your performance will be worse as a result I would recommend spreading it manually and I've technically I do this with my finger you really shouldn't it isn't a skin irritant I don't know what kind of effects it has but it's easier I don't recommend doing it that way though but you spread it out kind of thin on the surface and then when you reseal it you'll be just really surprised at how much that does for you I'd probably stop around there I think typically and then you just do this on each of the four corners and it's really not going to come off so what you would do next is you can use the D litter tool to reseal it often they come with a clamp that pushes down right on the middle of the top IHS what I've been doing is I actually will socket it unplug the systems you're not tons of diagonally turn it on socket it and install the cooler and just let it clamp and that should be enough to apply force let it sit depending on what kind of adhesive it is this stuff takes about an hour to get an initial dry I leave it alone for a lot longer than you need to I leave it alone for overnight basically as my rule because I've had enough problems where I've had to deal it them again and again and reapply liquid metal on it sucks so I just leave it overnight at this point to make sure it's not an issue with the sealant and don't apply more than this because if you do your performance it's gonna be worse another common problem here is hot spots on the core so it's pretty common to have an issue where you go back into test thermal performance and you see every core is at 60 degrees then you have wanted 104 so that could have happened because there's not enough liquid metal in a particular area or like we have right here you might have a dot of adhesive that was scraped off so right at the tip of the pen there there's a black dot in the pool of liquid metal so it's hard to see but that black dot is from when I was scraping off silicone adhesive so if you do that just that could definitely cause a hot spot on one of the cores because you're not getting contact anymore with something that's highly conductive this liquid metal is like seventy three watts per meter Kelvin and the silicone adhesive is temperature resistant so you don't you don't want there that cause that would cause poor performance and deleting it again and applying liquid metal will fix it for another example we recently just deleted Ed's CPU from tech source and for that one we used an ax 299 dark motherboard from EVGA so the results are different it actually pushes a lot more current through the CPU than the ram-paige six extreme and this is running blender at 4.5 gigahertz with 1.2 volts for voltage we observed about 14 degree improvements from peak core temperatures to peak core temperatures keeping in mind that we were throttling previously so we have no way of knowing how much higher the temperature would have gone had the clocks been on throttled so that's some of the practical basics another note would be clean thoroughly which I didn't do here but clean it thoroughly before you apply liquid metal use just rubbing alcohol q-tip something like that and in terms of the next step its you should do this before deal it in as well but knowing how it's a test is really important because people will say well I'm only getting a 1 degree improvement or I'm getting a 40 degree improvement or whatever and neither of those numbers should be happening it's certainly possible at least on the 1 degree side it's it's possible that the seal was too heavy or the liquid metal application was too thin or something like that but the bigger likelihood is that the testing isn't adequate there are thermal testing is really difficult to do right so I'm gonna give you some quick basics it's not enough for for what we do in terms of reporting thermal data making cards and everything but it's enough where you'll be able to get a good idea of if you're dealing work properly before testing probably is hardware info 64 open that up set it to log data don't just spot check it it's not good enough because you will bias the numbers so set it to log data over time ideally every second or something like that run an application that you know will create a pretty common or a pretty steady rather current load going into the CPU so if you have for example blender is pretty much a straight line sometimes it'll spike higher than others you have to keep an eye out on it you can clamp the the piece the EPS 12-volt cables to get a bit more accuracy so if you put a current clamp on them then you'll know if it's drawing more power than it did last time but not everyone's gonna do that so hardware info 64 will give you a wattage reading for the CPU that tends to be pretty accurate with most Intel CPUs so you can rely on that for the basics and if it's drawing way more power in one test than the other you need to rerun it a few times and get a couple different numbers because ultimately the thing that drives temperature and heat is the power consumption and sometimes some software drives more power than others you take something like an AV X low wear clothes like blender or prime95 29.2 and you compare it to a non AV x workload like prime 26.6 the temperature numbers are going to be hugely different because with a BX you might have to AVX instructions spinning off on each core because you have two threads on each core whereas with non AVX a lighter workload with lower power requirements so the point is to test you need consistency consistency you need a consistent power load blender is pretty good prime is good but it's power cycles and sometimes it'll power cycle hard so what you should do is plot the the power consumption over time if you can get it from hardware info or a current clamp and then look at the the power cycles during that test and then when you do the nut the next test later after the deal it you compare the power cycles again you pick two of the peaks or two of the lows or whatever where the power is about the same for each test and you come here those two numbers and you average that data over the period of time that it was at steady-state not the period of time the test is running because it will be ramping up in temperature but the steady state temperature if you're under something like a 280 millimeter liquid cooler let the test run for half an hour so that the liquid can fully saturate so to speak and you're at steady-state at that point with the liquid so the radiator shouldn't be changing the behavior at that point and all you have to do now is average a couple of numbers and you'll get your temperature value we would recommend checking all of the core temperatures and averaging those and then you can also check peak to peak or differences something with intel cpu with cpus in general is that the cores all behave differently so every now then you'll have one that's 20 degrees higher than another one that's okay it's not a big deal it doesn't mean you did anything wrong just make sure that they're both lower than they were when it was thermal paste it's not uncommon to see 20 degree Delta is between cores depending on what's happening on each core during the test so that's kind of normal but there's a lot of bad test data out there and you'll see users compare temperatures like they're a fire strike score like well I got 70 degrees why did you get 50 degrees not really how it works there's a lot of stuff that goes into it the software is a big part of it the liquid cooling solution or air cooling solution is the biggest part of it and so you need to have consistent testing and don't compare to other people compared only to your own numbers just collect them properly beforehand there's a lot more we can do on that but when I I loaned Paul our D leading kid I actually sent him a pretty detailed overview of how to test and how to deal with it again and reapply looking at all that stuff and the point of it was that testing these things you really only get one shot at it cuz after you've deleted it you can't go back and test it stock again so spend the time beforehand test it properly stock and then do it after and and just kind of average numbers at steady-state with an active constant load level there's a power your numbers should be good so I think abouts most the basics you might have some power from deviating for every 10 degrees Celsius that you drop CPU temperature we see about a 4% decrease in power consumption that's from power leakage its operating more efficiently and this is well plotted at this point with a lot of CPUs and tests so you'll see some change but not a ton and I think that's most my notes here it's pretty pretty disorganized but the point was here's a bunch of ideas during the process that we've encountered and problems we've encountered an experience I've gained doing the leads they're not necessary you don't need to do them but there's something to be said for an enthusiast task it does improve your temperatures significantly in a lot of cases and that can give you other benefits whether or not they are necessary or useful to you is up to you but hopefully that helps somewhat there's a lot more I could say here but we got to cut it at some point so I'll cut this one for now if there are more questions about this we'll revisit it the testings probably the most detailed part if we wanted to really go detail with it but I think that should give you some pointers to get started for now and check our D lid coverage you just search google for gamers Nexus deal it or something at this point you'll find a lot of different data on it though stuff we did for Kyle CPU and for our own 79ad XE was pretty interesting that'll get you started on the coverage side and that's it for now thank you for watching is always being subscribed for more if you want to catch the other videos go to store that guarantees access dotnet slash mod matte to pick up a mod matte like the one I was working on here and you go to patreon.com/scishow sexist helps out directly thank you for watching I'll see you all next time