Gadgetory

today we are going to need a lot of thermal paste so we are doing thermal compound application methods on thread Ripper using the 1950's specifically the video will show you using a piece of plexiglass how the throne page spreads when we depress it on to the IHS and then we're going to look at how it spreads once a cold plate is applied to it through actual mounting force as normally and finally the end of the video we're going to look through the results which one's perform better if there's any appreciable difference at all and this was partly inspired by an older video of dear bowers so we're borrowing his methodology with one of these which it's not perfect you're not using the same force to mount this thing as you would a cooler but that's why we're expanding on it by also mounting a cooler and we'll show you before and after before getting to those that this coverage is brought to you by e BGA and they're 1080p is C 2 which we've recommended fairly highly for its build quality and the icx sensors which are kind of fun to play with you can check our full FC 2 review for the 1080i if you're curious to learn more or you can click the link in the description below to find the product page for the 1080i SC 2 and because AMD gave some outlets permission to go ahead and post thermal results already we figured it'd be ok if we went ahead and posted some of ours for some of that matters quite a lot for you as a DIY builder which is thermal paste application so we're looking through that thread Ripper is interesting because of its dye layout and this is something we showed in our initial video where we show the location of the dyes and how distant they are from the center so it's not like what you'd normally expect with a traditional heat spreader and dye where it's right in the center of the thing ok so here's what we're going to go with I have 6 methods here that I've drawn on the paper and the first one number one is actually what AMD is recommending for their application so they're showing 5 kind of small to medium sized dots in the corners so you hit the dies here and remember only two dyes are active so you might have these two or these two the other two are silicon substrate interposers that are just spacers the next is a modified version of what we've been using which is just basically Road lines across the dyes and then what actually doing as this right now but that may change after these results so we're doing like a dotted road basically this is your traditional dot and X and then just ham spreading it by hand across the whole thing and just as a reminder here is the chip that we for the IHS that we drew on with dry erase marker as a reminders and some people thumb that's highly offensive uh-huh so the IHS be drawn shows you the dye location you've got two of these active and to inactive we don't know which ones and that's what we want to cover so our objective is to cover this if we flashback to the footage of the default A's attack thermal paste we already know how that applies and it only really covers I maybe like half of the dye area the cold plate covers the whole thing it barely covers these outer edges but the thermal paste did not so we're going to look into that with this video let's go ahead and start this off as you can see with this drawn on one we have the Plexiglas one or whatever it is polycarbonate whatever matching up pretty much dead on so we use this to push down and let's actually just start with method number four which is the traditional dot in the middle method and I'm just going to go ahead and put a check mark here so we're doing the Plexiglas one first so here is what we're looking at and this first one I'm going to do a big dot right there on the Z which is dead ten of everything this is a pretty traditional way to apply it their own pace so it's reasonable to expect people would continue doing this if they don't know what's underneath the IHS we're going to go a bit bigger than normally and normally you might stop somewhere around there if you're working with a smaller intel chip or something we're going to just do a glorified version of that I mean that's still pretty good amount of thermal paste let's do a little bit more okay so a giant died right in the middle or a blob rather not die and size this drawing on yeah okay so here we go just kind of line it up and not getting much here and as you forced more pressure of course it will spread a bit more which will see that once we start applying the cold plate with the socket correctly with the screws but that's what we're getting right now it's just can't spread out much more than that so I think that's pretty much where we stand so that's what the blob gets you now if we look at the dyes not covering a whole lot here we got about half of each of them that's going to be suboptimal transfer you might be fine you know if the cold plates still touching these outer areas you're still pulling heat off of them but this isn't about being fine it's about what's the most efficient that's what we're looking at you today may be the case that it's totally irrelevant we'll find out in the results section now to clean this off next one we're going to do is the let's go with the AMD recommended method so they're doing a dot right around here in the middle and let's also assume that they're putting the others in the middle of the dye location so this is going to be between the e and the N that's also pretty good size those are all pretty I think we may have done a bit more here than the Andy doesn't there their media recommended version but I think we might need more try not to apply too much torsional force here so something like that it's what we end up with which if you did a bit bigger dots you'd actually you would cover all the dyes so that might not be a bad method bigger dots than we did but overall not bad we'll test it and see how it does let's move on next one let's do the line method not the one we've been using but one pretty close to it do a line from the a to the N T to the R and this is not what are we doing that in the middle of these or what kind of the bottom of the a kind of the middle of the T okay and these are are going to stick out from under the cold plate sort of they'll they'll definitely bleed out from under the cold blade of a nice deck cooler okay here we go the real test is going to once we get the cooler down here all right so by hand that's pretty good that's about where it's about as far as we're going to get so that one looks good the best looking so far we've covered all the dyes we're not doing full i HS contact if you're mounting a smaller cooler to it anyway it doesn't matter but that's what that's what we're looking at with this one which actually this makes it kind of look like the next method is not going to matter okay next one I'm out of paste I think on the first team it's all the same pace though so next one let's do the same thing just with the dots in the middle which really almost feels pointless but pretty much the same spreads out a little farther on the sides but the two lines for the road does just as well as the dots in the middle two more methods to do with this one let's do an X all right X how's it go see I should have figured that out sooner make city makes it easier for the camera to see what's going on when I push more force onto it right okay it's not great it could work if you use more compound than I did but as you can see over here you don't have full coverage pretty good everywhere else honestly so basically the lesson is just make sure you're at the right height it's the starting point so you could bypass that by just doing a dot in the middle of the R and a dot in the middle of the end to make sure you cover everything but not bad overall okay so a bit much on the throne face yeah kind of unavoidable I guess I'm up here to probably do one of the football things but yes that's a quick look at it we're going to do something more proper now with an actual CLC apply that to it wait okay that hands okay apply an actual TLC to it with real proper mounting pressure that's more even across the even across the die area and across the IHS then we'll get into testing see how it goes with last compound to use for that one spreading method but so far what we've learned is the X method looks okay so active looking okay right here and then let's do a note good this one was okay this one not so good and this one is like excessive these are my notes for later compared to number two so that's what we're going to test let's do the cold plate stuff and then move on to testing so you finished applying all the thermal paste we did a couple different variants so there's one variation of the blob with method where basically did a small and a large and then a slightly larger large blob because those are yielding interesting results so we did those instead of the full-coverage then I'll paste spread basically remove that one from the equation and honestly I think that one will make more sense with something like the nock to a full coverage plates anyway because right now if you manually apply their own plates to cover the entire IHS with one of these days attack coolers you're not going to use a lot of the compound that's on there anyway so it just kind of sits there and not sure whether it helps or hurts but we had other things to test so act to that one but had plenty of other cool ones to work with and we've got photos comparing the different application methods after the cooler was removed to here's how it ended up looking in the photos these images show the application after the cooler was torqued down fully run through our burnin testing and then removed this has a huge your mileage may vary element to it so don't take any of these as hard fact it will always be this way take them as this is how it would be if you applied it the way we did it will obviously change with these application because you're just going to either use slightly less or slightly more thermal paste each time you apply thermal compound including variation on where that paste is actually placed on my HS the plexiglass test was a nice idea but once we actually applied real mounting force to the thermal paste application it was clear just how much more spread you get from that retention kit it pushes down a whole lot harder than you'd be able to do by hand so once in a real-world application the Plexiglas method doesn't really remain all that representative of what we're doing but it's a nice idea for on-camera demonstration of how the compound spreads outward and that's about it it's good for seeing the process of spreading but not good for seeing the final full spread which is what these photos show only a few applications show notably different spread patterns between each of the application methods and we can get into testing those next some basic on testing we use the same mounting force each time the retention kit is pretty easy to work with torque drivers and things like that also make it easier so same mounting force same cooller same thermal compound for all the tests from the same batch not that that's too relevant at that level it's not going to be visible and with our resolution anyway max speeds on the crack next sixty two and then the voltages we fixed the stock test to 1.2 volts so it's a little bit on the high side compared to out of box stock voltages but it's stable and it doesn't change at one point - it just kind of sits there which is what you want because voltage changes the power increases or decreases and you have variant we fixed it to one point three six two five for the overclock with a forty multiplier and then for logging had current log in via clamp at the EPS twelve volt voltage monitoring and of the CPU that is and then ambient all of which were aligned ii ii and spreadsheets and then we could check for consistency errors or things like that in there as I understand it threaded fur has 36 sensors in it so these things they well they have a lot more sensors than that but they have 36 relevant sensors to us and those are placed all over the dies and inside the IHS I did ask about this at the am the event if they could give some examples of where they're placed within the IHS and didn't really get any concrete examples just more of a well there's a lot of them so they're pretty much everywhere and that's fair so the sensors in here are averaged depending on how you what software you use and how it works with it and if it even recognizes a stripper with hardware info 64 which is definitely the best one right now it creates a TDI readout and TDI is a collection of all those sensors averaged so I wish we could see them regularly because then you could if you knew where they were you can tell how the coverage is in contact is for each specific dye but but this is good enough so that's what we're using we are also using prime95 28.5 29.2 had some issues we like to do 29.2 with 8k sizes for the fft but it just it there's some issue right now with either thread ripper or prime we're not really sure what so we just use 28.5 because it works and the clocks didn't change that well at all really there's power cycling and prime so we account for that in the data and the spreadsheets and an average it all out as necessary and get rid of any of the data that looks bad so we enroll into it now pretty straightforward though assuming your ultimate goal is just to cover the whole IHS it seems that application method doesn't too much matter as long as a sufficient amount of compound is located centrally our small blob tests for calling it uses a sparing amount of combat and the least out of all of these will show that on the screen now if you want to see how much it was and that managed to perform similarly to the ACE attack pre application and the double line pattern which performed a bit worse than expected honestly the double line seemed like to do it better our our X pattern gave us the most overall coverage once we corrected the lines to better intersect the dies and ended up covering the entire cold plate of the X 62 this gave us a measurable and repeatable difference of about two to three Salty's improvement depending on which numbers you're comparing this is pretty damn close to our error margins so it skates by just barely so more we're really close to being equal here it's not enough to be a revelation the only real reason the X pattern is performing well is because we've covered more of the cold plate though there's a limitation to the efficacy of a spread as you start bumping into the seal boundaries toward the screws in the plate outward in the outer edges of the coolers there's a seal and you're really not contacting anywhere there's micro fins or even liquid anymore and the very least we're preventing air pocket hot spots from forming in those screw holes which do sit right over parts of the dies so that's always a good thing following and these application method recommended we ended up closer to 44 C not that distant from the other tests and also within error margins for some of its neighbors or most of them even it could be better AMD's recommended application method actually is just fine as a method but it leaves a whole lot of room for the user to play and that's not a silly a good thing there's no reason that the five dots method recommended wouldn't work just as well as the others it's just a matter of how much pace you apply with larger dots assuming it's not excessive we should more or less match the performance of all the other application methods nearby we followed a indies guide as best we could but there's room for bigger dots as always and ultimately were just here in a video anyways they're hard to get depth there but figured odds would make it just fine the most interesting test was the heavy blob method we did two applications of this one with the first using a significant amount of compound dead-center of the z in the word risin the second test used about maybe 25 or 30% less compound and still managed to a 39.6 Celsius delta T over ambient average performance figure the larger version of this blob operated at around 35 Celsius delta T over ambient a significant performance uplift over some of our other test patterns like the v dot pattern with the smaller dots and even the A's attack pattern which is the silkscreen applied one again we think that this is just a matter of relying on the coolers own mounting pressure to spread the compound out evenly across the IHS rather than trying to manually spread the coverage the downside is that you're sort of guessing out how much you need to cover everything because it's all being applied under the hood when you're tightening things down that's the only real disadvantage to the blob method it's easier to gauge coverage with an X for example or the parallel line and these recommended v dot method is also good for this but we just have to use enough to make sure there's full spread outward with any of them and we'll be fine what's not shown here is tired tests without thermal paste which resulted in clock throttling as you might expect a very unhappy thread read for cpu but we were just curious as for overclocking results while we didn't have a good enough cooling solution to find a difference the CPU is pushing 21 to 26 amps down the EPF 12 volt cables at this point around 262 326 watts with that overclock and so was running hotter than our Kraken X 62 could reasonably handle it at stock state it could handle it it wasn't throttling at all actually but its bottleneck so to speak enough by being the weaker link than the thermal paste so we're not going to get a picture of how much they'll face impacts things without a better cooler we need a heavier duty cooler to get the thermal down to a point where that their own phase difference is visible if you're wondering why this chart doesn't have every single method we used in the other one it's because it was pretty apparent what was happening by this instance and testing no point in trying more methods when our cooler is clearly the limiter there's no perfect method the only real conclusion here that covering the whole plate or at least most of the plate is going to help out a bit but as far as which method to do it it comes down to how confident are you in the coverage of each application if you feel better about the X because that's what you use and you feel like you can gauge that its coverage and make sure it'll cover the relevant dyes then go with it but if you feel better about the blob just use a whole lot of thermal paste dead center it'll spread out and be pretty good it's just a matter again of how much compound you use there's no reason the X method or the dual lines method or AMD's recommended five dots method wouldn't be just as good as any of the other ones it's just you have to use a lot of compound so personally speaking my experience generally is with that blob method I know how much to use to spread it across the whole IHS with all the older CPUs so you just up scaly here and that worked out if I had used more compound with any of the other ones you'd probably see pretty similar for Florence but it just comes down to what you're used to so relying on the cooler to spread things out is it seems pretty good it's a lot of force you just put a big amount right in the middle and it will squish all the excess outward and cover the whole thing including the rest of the cold plate that isn't covered by the stock pay it's on the easytek coolers so we're looking at next is where it really gets interesting with the full coverage cold plates we don't have any of those yet but if you have something like the Noctua cooler with the full coverage plate or the EK water blocks one coming out soon those instances might be one where it makes sense to manually spread the compound over the hawaii HS because you actually have a plate that can contact the reported IHS whereas one we're using right now only contacts the center so if you cover the whole thing you just end up with a lot of compound on the corners that never touches any cold plate never touches any copper at all but that's what we're looking at so overall not a huge difference in performance for a lot of these it just comes all that the conclusion here is not blob is better know what it is it's more they're all taste is better within reason so if you can cover the whole thing then it looks like that'll be best for performance so one thing I'm not a hundred percent sure about but I'm kind of I suppose we could hypothesize about is that the holes the countersunk holes in the cold place being that they are deeper into the plate than the actual cold plate could be producing hot spots of air when the things heating up because they do actually sit right over the dies so that might be why we're seeing performance uplift in the scenarios where there's full spread over those holes as well but that's it for this time as always patreon.com slash gamers Nexus that was that directly you can go to gamers Nexus squarespace.com for the store where you get a shirt like this one or the one I wore in the first half of the video the day between them subscribe for more as always I'll see you all next time today we are going to need a lot a thermal paste content here did the first time