if you are consistent about sanding in one direction, then rotate by 90 deg when you change grits, then you can see when you have removed all of the scratches from the rougher grit.
Roman, you're a lifesaver. Just the experiment video I needed, as I am currently in discussion with a guy who bought a lapped (more like flattened I guess) 1366 CPU from me. About 400 grid it looks like, and the guy was upset about the CPU being scratched, saying I ruined it. This is just perfect! Herzliche Grüße von einem schönen Niederländischen Dorf nahe der Deutschen Grenze!
Direct die cooling, lapping the die and heat sink to the point where you can basically do cold welding, thus not needing and thermal compound, is basically the only thing I'd find particular interesting in this category, but of course it would both be time consuming, difficult and potentially expensive to do a proper test, so I don't expect to see any such experiments any time soon.
That would be interesting test indeed. In machining world there are gauge blocks which are lapped to be very-very flat (sub-micron over the area) and the good quality ones wring together (google for gage block wringing). Now, if both cpu and heatsink would be that flat and wrung together without thermal paste, that would probably be quite good. Although I wonder how much cpu bends with motherboard when heatsink is mounted... That would ruin the whole idea.
@@vilts Yes, the PCB flex could be a problem, but I don't think so, presuming a single die and even pressure from the cooler. As for multiple dies, that's a tougher one. I still don't think it would be a problem, but even if, it is possible to support the motherboard, I think. All in all, I think it's perfectly possible and I'm also convinced that it would work very well, not to mention that the heat conductivity shouldn't degrade over time, but to achieve this result you would basically have to do the last work, cleaning and assembly in a vacuum. Well, if I am to believe what I've read on the subject, having never attempted anything like it myself. I've machines parts to a tolerance of .001 mm by hand, but that is hardly the same. We'd need perfection, not because perfection is needed to get good result, but because we wouldn't be able to trust the results otherwise, as even the tiniest imperfection could have a significant impact.
Back in LGA775 days, I was into cpu and heatsink lapping, I did both of my P4 prescott and C2Q yorkfield. From 120grit to 2000grit sandpaper with a piece of glass underneath, lapping on both CPU IHS and heatsink. The cpu and heatsink are smooth and flat enough that when both surface touches, it stick together like a magnet. Both CPU seeing temperature drop of 5c or more. I remember doing CPU alone will get around 2-3c of temperature drop, doing both surface get 4-5c, getting both surface to near mirror finish drop another degree.
I was taught in college to get a perfectly flat and squared surface when lapping by hand was to do a minimum of ten strokes pulling towards yourself then rotate ninety degrees and repeat ten more strokes then rotate another ninety degrees for all four sides.
Jan Nowak So google it or look up GamersNexus on TH-cam then. I’m not gonna write you a dissertation on who he is or why he’s nicknamed Tech Jesus. I gave you the stepping stones, now it’s up to you to build the path to the knowledge you seek.
@Jan Nowak why did you have a conversation about not wanting to know who someone is after you asked whom that someone is, and they're not a random guy when a nickname is associated with them and you are actively talking about them and only them.... they are the subject matter and therefore you can refer to them as someone you dont know of but shouldn't continuously refer to them as a "random guy". 😂😂
Try again with some sort of sanding block to apply even pressure during sandings. Also, try some polishing before each step. Lastly, there seems to be some irregularities along the borders of the IHS. That's probably from the pressure being placed around the borders instead of evenly across the surface.
Yeah it looks like while sanding he's rounding the edges to me. Just going through the motions doesn't always get the job done. He should be inking on a surface plate and seeing what he's really doing.
I appreciate the time that each stage took to sand and test. Interesting results for #40 grit. Thanks for the tips on how to tell if your IHS is flat or not so you can judge if there would be potential for improvement or not. Seems like this is one of the last areas to try to increase performance from though: already have the fastest CPU and excellent cooler.
It would be interesting to see if lapping the heatsink along with the IHS would allow for further improvement. It seems intuitive to me that flattening down the IHS would only help up until the point that the roughness of the _heatsink_ becomes the limiting factor.
The problem is not the roughness of the surface but that the IHS gets deformed when it gets soldered. That doesn't happen to the heatsink so I would assume it's already flatter and smoother than you could make it and you would only make it worse.
They are supposed to be lapped together. Not sanded separately as shown in the video. Yes, it makes a difference. Yes, high surface roughness is a bad thing. People that think otherwise do not understand basic thermal resistance, the terms being used, or even what roughness is.
The LN2 pots are already lapped so they don’t cover it. Kingpin believes leaving a lightly scratched surface on the IHS reduces the risk of thermal paste cracking, decreasing the thermal delta, mostly only apparent in LN2 temps. BTW Saying things like people don’t understand what “roughness” is just makes you look like an Ass and makes me want to disregard everything you stated.
@@probablynotabigtoe9407 So you're a snowflake that will ignore facts if they're not stated in a way you like. Got it. And you think "paste" cracks. Hint: it is no longer a paste if it cracks and it belies a different problem with a different solution. Be technical. Otherwise, kingpin can believe whatever they like, doesn't make it correct or perform the best and if you have to bring in an edge case that has a different fundamental underlying problem to have a sort-of-not-really leg to stand on... well... you can think I'm the ass all you like; at least I'm not disingenuous. Regardless, It's still hilarious, if a bit sad, to see the blind leading the blind like this and who think sanding it lapping.
Once you are beyond the uneven surface benefits are minimal, any gains are only due to the thinner material between die and the cooler heatsink (thinner IHS). Hence why you are dealing in fractions and not drops in whole degrees.
I was looking for this comment which luckily didn't require much. Once the surface is ground down to the copper plate, every additional step in sanding actually doesn't remove much of any material by a significant amount compared to the prior. Still going from 180 ~> 1200 will result really no difference in what material is removed.
if that had been a significant factor, we would probably have observed more benefit from the higher grit sandings, as there would be less material at each step.
Hey Der8auer, Thank you for calling this sanding and not LAPPING! My feathers get ruffled when people call it Lapping. Anyways, great work you are doing. Cheers.
My first Q6600 (the less desirable original B3 stepping) had a very uneven IHS. Started with 120, then worked my way to 1500 testing between each. I got a 6C drop with the 120, then another 1C with 220. After that I didn't notice any consistent measurable difference. To date that has been the only CPU where lapping, sanding, or whatever you want to call it made more than 2C difference.
Nice one. I did lap my old CM HyperTX3 because I use it with Ryzen 2600X, though while Ryzen is under warranty gonna leave him alone. Maybe lap him after warranty ends. 600 with a small amount of water and tootpaste is what I used to make the cooler surface somewhat polished and have blurred mirror effect.
IHS are desgined with a slight profile so they're flatter at operating temp. Lapping flat at room temp (even if you could do by hand without it becoming concave) is sub-optimal. I only do this on the contact block of a cooler which has obvious deep grooves.
So.. I have a noctua NH-D15S, and it comes manufactured with a convex (top to bottom) cold plate, with noticeable circular ridges in it.. I wonder if it's best to grind these ridges out? They look significant..
I know this is an old video but in regards to the part about not knowing if the IHS or cooling block is flat you could always test fit with with marking blue/marking out dye (instead of thermal paste) on one component and see the contact pattern. That is if you are bothered to go that far down the rabbit hole...
I think you need a better method of determining flatness. It is hard to tell on video but it looks like the outside edges are rounded over to me. Just because you're sanding something does not always mean you're making it flatter.
I ve "lapped" a few CPU starting with my old Q9550 E0 - if you do it by hand you 'll always have a certain amount of "rounding" on the edges and it doesn't matter as long as the IHS is otherwise flat afterward. The area that matters is middle/where the DIE is located. The best result I achieved was with my Q9550 where the core delta shrunk from 15 kelvin down to 7-8 and overall temps went down by around 4°C. That IHS was soldered but really bent.
@@folterknecht1768 how can it not matter? You are losing the conduction wherever you have no contact. You may be better off than you were before but that still does not mean it doesn't matter.
Agreed on the rounded over edges. Basically this is because the sandpaper is lifting from the flat surface, you can see it happening for example @6:45. This can be alleviated by taping the edges of the sandpaper down so it is taught over the surface. Alternatively you can try a light spray adhesive between the glass and paper, but this could potentially add its own unevenness.
If you want to know if sanding your CPU could do a "huge" difference, just take out your CPU, clean it with alcohol, go to a glass-table and spin the CPU - with the IHS facing down of course. If the surface is flat it will not spin, if it is uneven it will twirl/spin.
You get better temps with both perfectly flat. And no ek is not perfect but they are really close, unless you are benchmarking, it isn't really needed.
I lapped a CPU using 800 - 1000 - 1500 - 2500 Grit After that I went to a Jeweller and had him buff it as much he could. We found that the metal is porous and when you get to a microbubble it creates streaks down the IHS. So the quality of the metal restricts you from intel, they are made cheap by casting.
To what I have read Intel actually stamps the IHS out of copper sheets. Though, at some point it is just easier to mill out a hole in the IHS and make direct die contact instead.
I lapped a few CPU's like this. Always had like 2 max 3 degrees C of improvement. It's not worth it unless You are going to make a heavy OC, then every single degree counts, it costs almost nothing and takes only hour to do it, so for me its obvious when it doesn't have warranty any more. It's more difficult when You lap an AMD CPU, when You need to watch out for pins, but I figured it out and used old socket from a broken motherboard to put it on pins while sanding.
Awesome findings! So bottomline, sanding from 80 to 600 will clearly bring some thermal benefits. I find it particularly interesting @6:05 that with 40 the heatspreader was too rough, even though of course it must have become a bit thinner than the original, but the roughness was too much so the overall result (thermalpaste on that thinner but too rough heatspreader) could not transfer heat as well as the thicker but smoother original. Yet interestingly enough, right with the next grain at 80 this changes, so in spite of the remaining noticeable roughness, the overall result with the thinner but still rough surface already does a bit better than the thicker but smoother stock.
I only rough up an IHS or cold plate if I'm applying liquid metal, so it sticks a bit better. Otherwise, I don't bother. IMO, people should focus on applying the right amount of TIM, properly.
I use the thin film method and I don't overclock but the temps I get on my system are great. Right now I'm running +20.0°C to +24.0°C All I have is a $10 Hyper T2 cooler. So the only thing I can chalk it up to is the TIM application. I just used the compound that came with the heatsink too. So whatever that was. Must be great stuff?
Gave mine 2600K and cooler block 5 min with 800. Got little better contact in outer corners. Together with better paste it gave better average temps overall about only 2-3 degrees, but the averages are now 1-3 degrees from core to core then before about 5 degrees.
I remember reading the results of a similar test several years ago. That test also showed that going past 400 or 600 sandpaper did not yield any improvement thermally, just cosmetically.
600 is the finest my local DIY shop had. I started with 240 to take the top surface off and then polished with the 600. I got a brushed metal effect as I was going up and down (and not in circles). Eventually the 600 paper wore all the way down to being super-smooth. I continued to polish with this until the IHS had a mirror-like effect. Overall I got a three degrees drop from stock (8700K).
Wire wool (very fine 0000) and liquid metal gives amazing results, especially with delid (don't use wire on cpu die - obviously). Ensure you have a flat surface for the heat spreader and cooler interface surface, this will minimise gaps and increase thermal transfer. i7 8086k @ 5.2GHz (1.35v) with temps in mid 60s running Prime95
The reason why your test balanced out is due to the thermal paste, when you lap your cpu, surface variations is less, thus, you need a very thin thermal compound, in fact, when i do it, i stop using thermal paste and start using oil, grease for lapping coolers under 600, and a dab of machinist oil (or even baby oil) above 1200, if you do this you will see a difference.
I had a problems with mine i7 11700k and high temps after overclocking. I always thought that 11th gen is just hot gen. Once i oced my cpu to 5ghz + 1.38 vcore, i even couldnt play games due 95+ temps and crashed after -+ 30 mins. If i tried to run prime 95, i instantly reached 100+ temps and throttling. Then i run 2 cores and 4 threads test with all enabled avx and got 98 max temps after 1 minute. One week before i saw a videos about lapping cpu and aio coolers. I decided to try. I lapped both, cpu at the end of 3000 grit and my aio at 1000 thousand grit and boom. My temps dropped by a -30%. Now with prime 2 cores + 4 thread + avx i got max temps only 70 in one minute test and even 4 cores and 8 threads candle no more then 80 in 5 mins test. Crazy. I can even turn on all threads in test and it reaches with 5 ghz and all avx on 100+ only in -+ 50 secs. About gaming totally difference. From 95+ temps, in 70. Stock temps is ridiculous. No more then 65 with all cores enabled in prime 95. Btw, i always used high power intesive tests not blended. Ofc i know that these temps go lower, because somewhere was a big hill, and that was in my aio cooler most as i noticed, it took very long time to sand it out.
I think it's well worth to just visually expect a new CPU with a razor or similar reliable straight-edge. From what I know of this topic, if IHSes are uneven then they are really bad. If you get a horrendous one then just return it for another one if it's not too huge of a hassle. If you can't see anything horribly wrong during visual inspection then there won't be a lot to gain from sanding - and the vast majority of IHSes are pretty flat. I'd probably skip sanding the iHS and instead keep the nickel-plated surface for use with liquid metal. Liquid metal will work ok with copper too so you could theoretically do both, but it's the most stable against nickel, and using liquid metal instead of a good regular paste will net you much better temperature gains than the sanding will. I think the added peace of mind and stability of the liquid metal + keeping warranty + much less risk of damage makes this a much more worthwhile approach. Also a hell of a lot easier. Sanding is tedious work, requires suitable materials and some experience to do well. I've done it a few times back in the old days, and I had improvements, but not enough to be worth it make any actual difference in OC or noise levels.
Yeah Depends on the cooler, I’ve bought some cheap air coolers that looked like using gravel might get it better (although I’m joking, a sidewalk or paved road actually would make it a lot smoother, I’ve sanded some parts on road curbs before... just not computer parts). Most of my other coolers (phase, LN2, waterblocks, AIO all look almost mirror finish without ever touching them, then it’s not needed, but air coolers with .5mm roughness is pretty bad, you might get cuts running your fingers across it fast, it was really roughly done at the factory!) I may compare this truly rough cooler before and after an FX (my hottest CPU, though currently it peaks below 0c at 1.7 Vcore on phase, lol!) My TR-1950x is just way too easy to cool, so it won’t show a big enough difference. My other channel is “Legit_Overclocking”, if someone comments on any of my currently garbage videos, I’ll make sure to do a quick 3 minute video showing the differences (differences in max OC at thermal limits, and temps compared vs same speeds). Just comment and say you want to see cooler lapping video. Preferably though, come up with something more impractical like “use a jet engine to cool a Titan”, because I’m more likely to do that (sounds more interesting and fun, I can spend $3,000 on a jet, that’s not an issue if it’s fun), I can do a cooler lapping video though, sounds intriguing enough.
I think FX is actual thermal probe, HWmonitor shows wrong though because Intel used a linear scale, and FX used a different thermocouple that is only accurate at high temps and seemed like no software company felt like correcting the algorithm for AMD chops (AMD did this because low temps don’t really matter anyways), with one of my coolers at -35c block temp under load, my FX-8350 at 1.7 Vcore (limit of Asrock 970m Pro 3 motherboard in BIOS), the core temp still shows 0c under load, which is what basically any computer part shows sub-zero. It should be 80c or 79c for max FX thermal limits, though official spec says lower like 69c, Overdrive says I had 79-80c thermal margins when the CPU thought it was at 0c, but the overdrive limit may be beyond AMDs official recommended spec anyways. The air cooler I mentioned does ok, but it seems like 5 minutes with some and paper or an angle grinder and it would be better. It was really rough... I’ve got plenty expensive coolers, but sometimes I just need one that I can slap on a test system with minimal effort and have my nice coolers free. It does ok for that, but it could probably be better.
Subscribed - I'm curious to see if the bottleneck removed by delidding would create a greater difference between results. I don't think it would make for a scientific test for the same reasons you mentioned, but what if the same test was done on a soldered IHS. We do not know if Ryzen 3000 will be soldered but if it is, it may make for an interesting part 2.
There were some mild scratches and a small pit on my 5800X IHS when removed my Wraith Prism. Not sure if the cooler itself or preapplied thermal compound caused them
you left your comment two years ago but I'm gonna ask anyways. You said your 5800 XIHS had a mild scratch or scratches on it. I too have a 5800 X with some mild scratches on it. Also the bottom right corner edge feels like there's a "burr" on it. And because the bottom right corner edge has this sharpness to it it's digging into the cold plate of the AIO. When I removed the AIO pump from the CPU after using it for a couple years I noticed the Coldplay had scratches from the corner of the CPU.
I'm about to lap my 9700k because the temperature spread from core to core is very inconsistent. Core 2 always seems to run 7-9° hotter than all other cores. The spread between the other cores are between 2-5°. I am running a custom water loop with a Thermaltake W4 cpu block, a EVGA hydro copper 2080 super, and a Thermaltake 420x64mm radiator and a EK 480x60mm radiator. I recently bought a EK velocity water block to see if that makes any difference first, then I will try lapping after that.
@der8auer, i think you should add delta per cores (coldest one and hottest one under load). Now we never know are there was any improvements. Not really important, but also may be interesting. P.S. Any news about Carbonaut? =) P.S. What if i add marking back after i lap CPU (with laser)? Will it count? =) I need my warranty back!
Carbonaut will be available in about 2 weeks :) sure if you can add it back :D Will add deltas next time but you can see difference between min and max
Vielleicht versuchst du mal die 3 Schleifstein-methode zum abrichten von heatspreader und heatsink. unter umständen würde das die performance von thermal pads stark boosten.
Could it also be that the paste is now what is causing the temps to not drop lower? From what i understand if you have 2 flat surfaces then there is no need for thermal paste. Similar to optical bonding doesn't use any glues. I wonder if Liquid metal would have given larger differences or Direct contact with the waterblock. Further Could have it been that the water block was also not flat or at the very least also needed to be sanded/lapped. I know Almost all of EKWB do not come flat. They are concave. So you will reach a wall at some point just sanding 1 of the 2 faces. To many variables here. Its pretty clear you hit a thermal wall with your results. I wonder how the results would be if you got both surface optically flat and used no thermal paste.
Would of liked to of seen difference between hottest and coldest core. Per core temp can vary by 10c or more. Would like to know if sanding would close that gap.
I hate lapping. I have done it, however, it is a lot of work to do it by hand. [EDIT: for sanding I think I used to start with 320 grit and finished with 2000 grit.]
it's a fine line between having an increase in thermal my micro scratches for surface increase and rough scratches that introduce air pockets or pockets of not full contact that essentially work like micro insulators. As seen by experiment shown there is a tipping point for a optimum.
Versuch doch mal mit diamtenpaste auf 8000 oder 10000 hinzubekommen. es gibt auch noch speziele paste die 20000 macht. Dann ist die oberfläche verspiegelt. Dann hat man den glass aufeinander effekt wenn man den lüfter auch so polieren würde, wo kaum Luft mehr zwischen ist. Somit mehr wärmefläche. Ich denke das alles zwischen 500 und 6000 fast identisch ist. den größten unterschied gibt es bei 40-200-8000-25000. Der punkt ist halt es auf einer perfekt geraden oberfläche zu machen. Also meistens Glass. In einer absolut gleichen Bewegung.
I lapped x5660 and x5675, was definitely worth it on those Xeons, dropped about 8 degrees Celsius. Though delidding is better, BUT...those chips are soldered, so lapping is all you can do
you need to lap or grind even the heatsink to get the best results... almost no paste needed better thermal conductivity. you are just perfecting 50% of the surface... take care of the heatsink to get better results
That's a great presentation, thanks! Btw, I probably have it misspelled as was hearing this at the beginning (and since I don't have that particular knowledge): what is "ignium", what soldering material are you referring to?
very intresting video i liked it video ideas: speed run cpu sanding (time to cooling point system) also would like to see a simular test on some CPU coolers like corsair h100 that generaly has a rough machined plate
I lapped my 5800x3d. I basically delidded it, then sanded down the silicon. The delid itself got me about 10 degrees and the lapping of the die got me another 3 degrees. The delid was as stupid idea, the lapping of the silicon is even dumber.
Have you found the Ryzen I heat spreaders flatter than Intel I heat spreaders? I've seen at least 2 You Tubers with expensive Intel CPUs with an uneven IHS. I would have expected a $2000 CPU to have an excellent integrated heat spreader and not spin like a top...
I get 5-6c better temps sanding my i7 6700k, one good check if you try to spin the cpu on the IHS, if cpu spins it's uneven protruding on center of the IHS so it worth it to sand if you plan OC, if not spin at all it means it's almost flat, so in theory it's not worth the time. In my case my 6700k spins like hell. I'm on 4,6 OC with 1,32v (not very lucky on silicon lottery)
Hey, why are CPU heat spreaders made that stupidly uneven in the first place? For example on Ryzen CPUs, they have the corners or entire sides, visibly way higher than the center......... How the hell are the coolers supposed to make good contact??
Has to deal with time. Kingpin went into detail over it. The only level side is the side with Die/IHS. Thermal past makes up the difference on consumer rigs. IHS's are stamped from one big sheet.
ihs are almost always made of copper..since there is no water involved, I would not be worried about. Galvanic corrosion occurs in humid ambients or inside the loop because of the liquid and the different metals, but not between the ihs (which is like 80 degrees at least and therefore not so humid) and the heatsink. I hope I have been clear enough :)
@@enricod.7198 IHS are also nickel plated too. There is also moisture in air normally. Copper exposed to air will oxidize eventually. That's why manufacturers are nickel plating their goods. They're not doing it because they just have money to burn.
@@1pcfred yeah I was referring to the ihs..oh well ok ahah didn't thought the moisture in the air may be sufficient to damage the components, my bad. So is this a real concern if we sand the ihs or with the thermal paste and heatsink applied almost no moisture gets through?
@@enricod.7198 there must be some reason why manufacturers are plating an IHS in the first place. But what that is I don't know. Well I know they're doing it to keep the copper from oxidizing. But why they care I don't know. Is it purely cosmetic, or is there some technical issue? Maybe they just do it because it seems like the right thing to do? It is an interesting question.
IHS is User friendly and has it place, but direct die will Always be king for me and doesn’t get the love it deserves. LN2 direct die doesn’t matter, but for normal people using ambient air overclocking it’s significantly better and needs a bigger spotlight. Although with modern CPU’s it is difficult to get proper cpu pressure mount, which can really fuck with your cpu if done wrong. So I guess it isn’t for the faint of heart.
Old Nvidia Gpus, Up until geforce 4 was really concave in the middle and you gained some cooling by lapping them. That was my experince from the 90ies. Alla intel p3 and AMD athlons had the bare die.. no ihs, but sometimes you cracked a corner and the cpu was dead!
I somehow expected the difference to be a little larger, but I guess it really wouldn't start to matter much unless one started to overclock by a fair bit.
@@1pcfred And it is a good video for that reason. It at least gave me a better appreciation of what is and isn't a serious scratchy surface on an IHS. Videos helping to alleviate misconceptions is good, after all, there is a lot of misconceptions floating around, like I once saw a tech channel proclaiming that quad channel memory were having negligible improvements compared to dual channel memory, based on tests on the AM4 socket. Not taking into consideration that it is a dual channel socket, some motherboards have 4 dim slots, but that isn't the same thing as 4 channels.
@@todayonthebench I think the bigger take away from this is how little is to be gained by sanding a CPU. 2 degrees at best does not seem to make any sense to pursue. In another video a guy named Kingpin says he sands the IHS so the thermal compound doesn't crack when he does LN2 runs. So in his case it's doing something for him. But that is a pretty limited use case.
@@1pcfred Yes, sanding IHSes is rather pointless for most average users. Going to a better heat sink or water cooling setup will have larger gains without voiding the warranty of the CPU. More extreme overclocking is another story, but one usually doesn't use such systems for more then a benchmark anyway.
@@todayonthebench yeah over at Gamers Nexus they talk about why Kingpin laps his IHS. It's so the heatsink compound doesn't fail when he does LN2 runs. Which is a very special use case.
Why? That mainly provides a placebo effect for the most part. The only useful purpose for that is to try to run direct contact WITHOUT Tim, and it wouldn't be capable of that high of precision.
It's curious that in "metric-land" you use grit which measures the diameter of the particles in INCHES. 100 grit is covered with 1/1000 of an inch particles, 200 grit is 1/200 grit etc.
For practical applications US customary measures work better than metric does. That's why we called out the altitudes in feet when we landed on the Moon. Because when inches count no one cares how many meters you are off the surface.
@@SuperFredAZ it is primarily due to US economic influence on consumer goods. It's why stuff like sheets of plywood come in kooky sizes. For us a sheet of plywood is 4x8. So for the rest of the world it is whatever the metric equivalent of 4x8 is. That's also why 13mm drive is so popular. It's really a half of an inch. 10mm is 3/8s 2.54mm pitch in electronics is a tenth of an inch. etc. etc. I know they make metric bricks. They are god awful ugly! We build jails out of them here. What do you expect from socialism? Freedom units FTW!
@@1pcfred Also tire sizes are all inches i.e. 17 inch 18 inch, etc. but then they go to metric as in 18V235R50, the 18 ins 18 inches, and the 235 is 235 mm, and the 50 is a ratio. Also the metric world measure their television and monitor sizes in inches i.e. 65 inch tv, you'll find them described in Europe as such.
@@SuperFredAZ often prizes do not go to the first but instead are reserved for the most popular. America did not invent tires but we popularized their use. TV on the other hand we did invent. It always annoyed me how TVs are measured on the diagonal though. There just seems to be something blatantly deceptive about that practice. Like what's the longest straight line we can draw in this rectangle? It just bugs me.
Maybe with soldered CPU could be higher with cpu vcore. I found one older test on Anandtech forum: forums.anandtech.com/threads/lapped-my-fx-8350-4-3ghz-oced-temps-lowered-by-9%C2%B0c.2300800/
yea I use whetstones for my lapping and I'd say the scratches make no difference. My 5k grit whetstone has a decent amount of impurities (I want a new one >.
I wonder what would happen if I were to grind it on my whetstones. I got a 12k whetstone. It would be an interesting test I think. But I don’t have a cpu to destroy so I’m leaving that one to someone who wants to ruin both their whetstones and their cpu
@der8auer thanks a lot. I used this method on c2d e6550 and I didn't remember the temperatures but looks cool. Can you preform this test on amd like 2600/2700 or even on a TR4 CPU? Big hug and see you soon :)
hello i just got a 10900k the temps ar semi ok but not as good as i xspected i have 3x360 radsand d5 pump + ek magnatude block after testing the cpu in glass and razor blade it is perect lat yet my ek black is rather convex. the cpu will not spin on glass but as soon as i place ontop of the water block i can spin th cpu very easy im wondrng if its worth lapping the water block
nice video :) have a difrent question thou , i bought the Lian LI o11 Dynamic case you was helping designing , but i cant get my thermals down to the lvl i have with the Fractal Design R5 i using water cooler for the cpu combine with air cooled gfx card, i have som friends telling me its becouse its the 90 Degree bend the air is need to take that it less effective cooling and Air flow , got any tips to fix that? tbh i love th design iof the o11 Dynamic case , kinda notice the mash/filters allso messing with the aif flow and termals .. im abit annoyed i cant find A water cooling block for the Gigabyte RTX 2060 Gaming OC card if you could help me out that too it would be great thanks
I wonder if switching to a different kind of thermal paste (with a smaller particle size?) might better take advantage of the finer grades of sanding.
if you are consistent about sanding in one direction, then rotate by 90 deg when you change grits, then you can see when you have removed all of the scratches from the rougher grit.
Roman, you're a lifesaver. Just the experiment video I needed, as I am currently in discussion with a guy who bought a lapped (more like flattened I guess) 1366 CPU from me. About 400 grid it looks like, and the guy was upset about the CPU being scratched, saying I ruined it. This is just perfect! Herzliche Grüße von einem schönen Niederländischen Dorf nahe der Deutschen Grenze!
Direct die cooling, lapping the die and heat sink to the point where you can basically do cold welding, thus not needing and thermal compound, is basically the only thing I'd find particular interesting in this category, but of course it would both be time consuming, difficult and potentially expensive to do a proper test, so I don't expect to see any such experiments any time soon.
That would be interesting test indeed. In machining world there are gauge blocks which are lapped to be very-very flat (sub-micron over the area) and the good quality ones wring together (google for gage block wringing). Now, if both cpu and heatsink would be that flat and wrung together without thermal paste, that would probably be quite good. Although I wonder how much cpu bends with motherboard when heatsink is mounted... That would ruin the whole idea.
@@vilts Yes, the PCB flex could be a problem, but I don't think so, presuming a single die and even pressure from the cooler. As for multiple dies, that's a tougher one. I still don't think it would be a problem, but even if, it is possible to support the motherboard, I think.
All in all, I think it's perfectly possible and I'm also convinced that it would work very well, not to mention that the heat conductivity shouldn't degrade over time, but to achieve this result you would basically have to do the last work, cleaning and assembly in a vacuum. Well, if I am to believe what I've read on the subject, having never attempted anything like it myself. I've machines parts to a tolerance of .001 mm by hand, but that is hardly the same. We'd need perfection, not because perfection is needed to get good result, but because we wouldn't be able to trust the results otherwise, as even the tiniest imperfection could have a significant impact.
Was the cooler lapped? How does cooler curvature play into this?
What about lapping both IHS and heatsink to the point where they "suck" each other like 2 plates of glass? Wouldn't thermal paste be redundant then?
Back in LGA775 days, I was into cpu and heatsink lapping, I did both of my P4 prescott and C2Q yorkfield. From 120grit to 2000grit sandpaper with a piece of glass underneath, lapping on both CPU IHS and heatsink. The cpu and heatsink are smooth and flat enough that when both surface touches, it stick together like a magnet.
Both CPU seeing temperature drop of 5c or more. I remember doing CPU alone will get around 2-3c of temperature drop, doing both surface get 4-5c, getting both surface to near mirror finish drop another degree.
that is not lapping
that is sanding
if they stick like a magnet, wouldn't it work without thermal paste?
Thanks for doing this in english. German is a beautiful language but I have yet to master it.
I was taught in college to get a perfectly flat and squared surface when lapping by hand was to do a minimum of ten strokes pulling towards yourself then rotate ninety degrees and repeat ten more strokes then rotate another ninety degrees for all four sides.
Yeh but you do this more for core to core delta than temps .
der8auer - you have to redo this entire experiment, in reverse, with novec, to see if rougher surfaces are better... ;)
Why didnt you grind the waterblock as that too being an ek is likely not flat either
Not using Delta over ambient? Tech Jesus would not be pleased.
Jan Nowak Steve from GamersNexus
Jan Nowak So google it or look up GamersNexus on TH-cam then. I’m not gonna write you a dissertation on who he is or why he’s nicknamed Tech Jesus. I gave you the stepping stones, now it’s up to you to build the path to the knowledge you seek.
@Jan Nowak why did you have a conversation about not wanting to know who someone is after you asked whom that someone is, and they're not a random guy when a nickname is associated with them and you are actively talking about them and only them.... they are the subject matter and therefore you can refer to them as someone you dont know of but shouldn't continuously refer to them as a "random guy". 😂😂
Try again with some sort of sanding block to apply even pressure during sandings. Also, try some polishing before each step. Lastly, there seems to be some irregularities along the borders of the IHS. That's probably from the pressure being placed around the borders instead of evenly across the surface.
Yeah it looks like while sanding he's rounding the edges to me. Just going through the motions doesn't always get the job done. He should be inking on a surface plate and seeing what he's really doing.
tbh the amount of people that would benefit from just remounting of heatsinks is crazy
As suspected. Great incheck as always Der8auer! Seems like making the IHS surface flat is the most important if the IHS has that flaw.
Very interesting results. I wasn't expecting it to be that close. Thank you!
I appreciate the time that each stage took to sand and test. Interesting results for #40 grit. Thanks for the tips on how to tell if your IHS is flat or not so you can judge if there would be potential for improvement or not. Seems like this is one of the last areas to try to increase performance from though: already have the fastest CPU and excellent cooler.
It would be interesting to see if lapping the heatsink along with the IHS would allow for further improvement. It seems intuitive to me that flattening down the IHS would only help up until the point that the roughness of the _heatsink_ becomes the limiting factor.
The problem is not the roughness of the surface but that the IHS gets deformed when it gets soldered. That doesn't happen to the heatsink so I would assume it's already flatter and smoother than you could make it and you would only make it worse.
They are supposed to be lapped together. Not sanded separately as shown in the video. Yes, it makes a difference. Yes, high surface roughness is a bad thing. People that think otherwise do not understand basic thermal resistance, the terms being used, or even what roughness is.
The LN2 pots are already lapped so they don’t cover it. Kingpin believes leaving a lightly scratched surface on the IHS reduces the risk of thermal paste cracking, decreasing the thermal delta, mostly only apparent in LN2 temps.
BTW Saying things like people don’t understand what “roughness” is just makes you look like an Ass and makes me want to disregard everything you stated.
@@probablynotabigtoe9407 So you're a snowflake that will ignore facts if they're not stated in a way you like. Got it. And you think "paste" cracks. Hint: it is no longer a paste if it cracks and it belies a different problem with a different solution. Be technical.
Otherwise, kingpin can believe whatever they like, doesn't make it correct or perform the best and if you have to bring in an edge case that has a different fundamental underlying problem to have a sort-of-not-really leg to stand on... well... you can think I'm the ass all you like; at least I'm not disingenuous.
Regardless, It's still hilarious, if a bit sad, to see the blind leading the blind like this and who think sanding it lapping.
hahaha no since you’re smarter than everyone, why don’t you tell us what happens to thermal paste at ln2 temperatures and why it can’t crack.
Once you are beyond the uneven surface benefits are minimal, any gains are only due to the thinner material between die and the cooler heatsink (thinner IHS). Hence why you are dealing in fractions and not drops in whole degrees.
Is it possible that the performance improvement was from the ihs thinning out and not because of the surface being shiny
I was looking for this comment which luckily didn't require much.
Once the surface is ground down to the copper plate, every additional step in sanding actually doesn't remove much of any material by a significant amount compared to the prior. Still going from 180 ~> 1200 will result really no difference in what material is removed.
if that had been a significant factor, we would probably have observed more benefit from the higher grit sandings, as there would be less material at each step.
NO, the problem is that the base was getting even moreconvex with every pass... he was polishing the ihs, not flattening it.
Now as 10900K is out, with thinner IHS, there an improvment.. so yea i guess it is possible.
Hey Der8auer, Thank you for calling this sanding and not LAPPING! My feathers get ruffled when people call it Lapping.
Anyways, great work you are doing. Cheers.
My first Q6600 (the less desirable original B3 stepping) had a very uneven IHS. Started with 120, then worked my way to 1500 testing between each. I got a 6C drop with the 120, then another 1C with 220. After that I didn't notice any consistent measurable difference. To date that has been the only CPU where lapping, sanding, or whatever you want to call it made more than 2C difference.
cee128d I had one of these and a G0. I had similar experiences with both.
Nice one. I did lap my old CM HyperTX3 because I use it with Ryzen 2600X, though while Ryzen is under warranty gonna leave him alone. Maybe lap him after warranty ends. 600 with a small amount of water and tootpaste is what I used to make the cooler surface somewhat polished and have blurred mirror effect.
Did your Ryzen 2600X have a small lip on one side? My 1700x did.
IHS are desgined with a slight profile so they're flatter at operating temp. Lapping flat at room temp (even if you could do by hand without it becoming concave) is sub-optimal. I only do this on the contact block of a cooler which has obvious deep grooves.
What if I had a small scratch on the green part of the cpu? Would it affect anything? I haven't turned it on yet because I'm scared.
So now your IHS is perfectly flat, but what about the cooler block? Is the completely flat, or would lapping that get even better results?
So.. I have a noctua NH-D15S, and it comes manufactured with a convex (top to bottom) cold plate, with noticeable circular ridges in it..
I wonder if it's best to grind these ridges out? They look significant..
I know this is an old video but in regards to the part about not knowing if the IHS or cooling block is flat you could always test fit with with marking blue/marking out dye (instead of thermal paste) on one component and see the contact pattern. That is if you are bothered to go that far down the rabbit hole...
I think you need a better method of determining flatness. It is hard to tell on video but it looks like the outside edges are rounded over to me. Just because you're sanding something does not always mean you're making it flatter.
I ve "lapped" a few CPU starting with my old Q9550 E0 - if you do it by hand you 'll always have a certain amount of "rounding" on the edges and it doesn't matter as long as the IHS is otherwise flat afterward. The area that matters is middle/where the DIE is located.
The best result I achieved was with my Q9550 where the core delta shrunk from 15 kelvin down to 7-8 and overall temps went down by around 4°C. That IHS was soldered but really bent.
@@folterknecht1768 how can it not matter? You are losing the conduction wherever you have no contact. You may be better off than you were before but that still does not mean it doesn't matter.
@@1pcfred
abload.de/img/unbenanntzfjcc.jpg
Agreed on the rounded over edges. Basically this is because the sandpaper is lifting from the flat surface, you can see it happening for example @6:45. This can be alleviated by taping the edges of the sandpaper down so it is taught over the surface. Alternatively you can try a light spray adhesive between the glass and paper, but this could potentially add its own unevenness.
If you want to know if sanding your CPU could do a "huge" difference, just take out your CPU, clean it with alcohol, go to a glass-table and spin the CPU - with the IHS facing down of course. If the surface is flat it will not spin, if it is uneven it will twirl/spin.
I'm curious about sanding the CPU *and* heat sink to ensure both are flat, for maximal contact. Are the EK water blocks already perfectly flat?
You get better temps with both perfectly flat. And no ek is not perfect but they are really close, unless you are benchmarking, it isn't really needed.
I lapped a CPU using 800 - 1000 - 1500 - 2500 Grit After that I went to a Jeweller and had him buff it as much he could. We found that the metal is porous and when you get to a microbubble it creates streaks down the IHS. So the quality of the metal restricts you from intel, they are made cheap by casting.
To what I have read Intel actually stamps the IHS out of copper sheets.
Though, at some point it is just easier to mill out a hole in the IHS and make direct die contact instead.
@@todayonthebench This process may have changed, this I doubt, the chip I "buffed" was a 775 C2Q Q6600. Will do a follow-up Video when I can
@der8auer Can I pin, please?
I lapped a few CPU's like this. Always had like 2 max 3 degrees C of improvement. It's not worth it unless You are going to make a heavy OC, then every single degree counts, it costs almost nothing and takes only hour to do it, so for me its obvious when it doesn't have warranty any more. It's more difficult when You lap an AMD CPU, when You need to watch out for pins, but I figured it out and used old socket from a broken motherboard to put it on pins while sanding.
Awesome findings! So bottomline, sanding from 80 to 600 will clearly bring some thermal benefits.
I find it particularly interesting @6:05 that with 40 the heatspreader was too rough, even though of course it must have become a bit thinner than the original, but the roughness was too much so the overall result (thermalpaste on that thinner but too rough heatspreader) could not transfer heat as well as the thicker but smoother original. Yet interestingly enough, right with the next grain at 80 this changes, so in spite of the remaining noticeable roughness, the overall result with the thinner but still rough surface already does a bit better than the thicker but smoother stock.
I only rough up an IHS or cold plate if I'm applying liquid metal, so it sticks a bit better. Otherwise, I don't bother. IMO, people should focus on applying the right amount of TIM, properly.
I use the thin film method and I don't overclock but the temps I get on my system are great. Right now I'm running +20.0°C to +24.0°C All I have is a $10 Hyper T2 cooler. So the only thing I can chalk it up to is the TIM application. I just used the compound that came with the heatsink too. So whatever that was. Must be great stuff?
So, how much of the gain stems from a less thick heat spreader after sanding, which also loweres the thermal resistance?
Gave mine 2600K and cooler block 5 min with 800. Got little better contact in outer corners. Together with better paste it gave better average temps overall about only 2-3 degrees, but the averages are now 1-3 degrees from core to core then before about 5 degrees.
Cool test! Like you said, weigh the risk vs reward :)
I remember reading the results of a similar test several years ago. That test also showed that going past 400 or 600 sandpaper did not yield any improvement thermally, just cosmetically.
600 is the finest my local DIY shop had. I started with 240 to take the top surface off and then polished with the 600. I got a brushed metal effect as I was going up and down (and not in circles). Eventually the 600 paper wore all the way down to being super-smooth. I continued to polish with this until the IHS had a mirror-like effect. Overall I got a three degrees drop from stock (8700K).
Wire wool (very fine 0000) and liquid metal gives amazing results, especially with delid (don't use wire on cpu die - obviously). Ensure you have a flat surface for the heat spreader and cooler interface surface, this will minimise gaps and increase thermal transfer. i7 8086k @ 5.2GHz (1.35v) with temps in mid 60s running Prime95
The reason why your test balanced out is due to the thermal paste, when you lap your cpu, surface variations is less, thus, you need a very thin thermal compound, in fact, when i do it, i stop using thermal paste and start using oil, grease for lapping coolers under 600, and a dab of machinist oil (or even baby oil) above 1200, if you do this you will see a difference.
I had a problems with mine i7 11700k and high temps after overclocking. I always thought that 11th gen is just hot gen. Once i oced my cpu to 5ghz + 1.38 vcore, i even couldnt play games due 95+ temps and crashed after -+ 30 mins. If i tried to run prime 95, i instantly reached 100+ temps and throttling. Then i run 2 cores and 4 threads test with all enabled avx and got 98 max temps after 1 minute. One week before i saw a videos about lapping cpu and aio coolers. I decided to try. I lapped both, cpu at the end of 3000 grit and my aio at 1000 thousand grit and boom. My temps dropped by a -30%. Now with prime 2 cores + 4 thread + avx i got max temps only 70 in one minute test and even 4 cores and 8 threads candle no more then 80 in 5 mins test. Crazy. I can even turn on all threads in test and it reaches with 5 ghz and all avx on 100+ only in -+ 50 secs. About gaming totally difference. From 95+ temps, in 70. Stock temps is ridiculous. No more then 65 with all cores enabled in prime 95. Btw, i always used high power intesive tests not blended. Ofc i know that these temps go lower, because somewhere was a big hill, and that was in my aio cooler most as i noticed, it took very long time to sand it out.
I think it's well worth to just visually expect a new CPU with a razor or similar reliable straight-edge. From what I know of this topic, if IHSes are uneven then they are really bad. If you get a horrendous one then just return it for another one if it's not too huge of a hassle. If you can't see anything horribly wrong during visual inspection then there won't be a lot to gain from sanding - and the vast majority of IHSes are pretty flat.
I'd probably skip sanding the iHS and instead keep the nickel-plated surface for use with liquid metal. Liquid metal will work ok with copper too so you could theoretically do both, but it's the most stable against nickel, and using liquid metal instead of a good regular paste will net you much better temperature gains than the sanding will. I think the added peace of mind and stability of the liquid metal + keeping warranty + much less risk of damage makes this a much more worthwhile approach. Also a hell of a lot easier. Sanding is tedious work, requires suitable materials and some experience to do well. I've done it a few times back in the old days, and I had improvements, but not enough to be worth it make any actual difference in OC or noise levels.
So lapping the cpu can squeak out some small gains.....what about the cold plate on your cooler then??
Yeah Depends on the cooler, I’ve bought some cheap air coolers that looked like using gravel might get it better (although I’m joking, a sidewalk or paved road actually would make it a lot smoother, I’ve sanded some parts on road curbs before... just not computer parts). Most of my other coolers (phase, LN2, waterblocks, AIO all look almost mirror finish without ever touching them, then it’s not needed, but air coolers with .5mm roughness is pretty bad, you might get cuts running your fingers across it fast, it was really roughly done at the factory!)
I may compare this truly rough cooler before and after an FX (my hottest CPU, though currently it peaks below 0c at 1.7 Vcore on phase, lol!) My TR-1950x is just way too easy to cool, so it won’t show a big enough difference. My other channel is “Legit_Overclocking”, if someone comments on any of my currently garbage videos, I’ll make sure to do a quick 3 minute video showing the differences (differences in max OC at thermal limits, and temps compared vs same speeds). Just comment and say you want to see cooler lapping video. Preferably though, come up with something more impractical like “use a jet engine to cool a Titan”, because I’m more likely to do that (sounds more interesting and fun, I can spend $3,000 on a jet, that’s not an issue if it’s fun), I can do a cooler lapping video though, sounds intriguing enough.
I think FX is actual thermal probe, HWmonitor shows wrong though because Intel used a linear scale, and FX used a different thermocouple that is only accurate at high temps and seemed like no software company felt like correcting the algorithm for AMD chops (AMD did this because low temps don’t really matter anyways), with one of my coolers at -35c block temp under load, my FX-8350 at 1.7 Vcore (limit of Asrock 970m Pro 3 motherboard in BIOS), the core temp still shows 0c under load, which is what basically any computer part shows sub-zero. It should be 80c or 79c for max FX thermal limits, though official spec says lower like 69c, Overdrive says I had 79-80c thermal margins when the CPU thought it was at 0c, but the overdrive limit may be beyond AMDs official recommended spec anyways.
The air cooler I mentioned does ok, but it seems like 5 minutes with some and paper or an angle grinder and it would be better. It was really rough... I’ve got plenty expensive coolers, but sometimes I just need one that I can slap on a test system with minimal effort and have my nice coolers free. It does ok for that, but it could probably be better.
Subscribed - I'm curious to see if the bottleneck removed by delidding would create a greater difference between results. I don't think it would make for a scientific test for the same reasons you mentioned, but what if the same test was done on a soldered IHS.
We do not know if Ryzen 3000 will be soldered but if it is, it may make for an interesting part 2.
Nice test, i was expecting more difference between 600 and 1200, but thinking about It thermal paste Is doing it's job.
There were some mild scratches and a small pit on my 5800X IHS when removed my Wraith Prism. Not sure if the cooler itself or preapplied thermal compound caused them
you left your comment two years ago but I'm gonna ask anyways. You said your 5800 XIHS had a mild scratch or scratches on it. I too have a 5800 X with some mild scratches on it. Also the bottom right corner edge feels like there's a "burr" on it. And because the bottom right corner edge has this sharpness to it it's digging into the cold plate of the AIO. When I removed the AIO pump from the CPU after using it for a couple years I noticed the Coldplay had scratches from the corner of the CPU.
@@philmccracken2012 yeah I've seen some IHS with that "burr" in a corner or two. Seems their machining isn't very precise
Would assume the reason any of these temps where different from eachother was because of ambient temps changing in the room
Great video sir! This video will be referenced for a long time.
You can buy aftermarket ihs's. Grind those and keep your caseking delidded cpu's warranty?
Many calories burned that day.
If you get both it and your heatsink coldplate to a mirror polish, can you do away with thermal paste and drop temps as low as possible?
Or maybe just get away with a thin coating of Conductonaut?
@@TheVillainOfTheYear You're missin' my point, I want as few layers as possible between my die and a coldplate.
I'm about to lap my 9700k because the temperature spread from core to core is very inconsistent. Core 2 always seems to run 7-9° hotter than all other cores. The spread between the other cores are between 2-5°. I am running a custom water loop with a Thermaltake W4 cpu block, a EVGA hydro copper 2080 super, and a Thermaltake 420x64mm radiator and a EK 480x60mm radiator. I recently bought a EK velocity water block to see if that makes any difference first, then I will try lapping after that.
The mirror finish should have been the last step, just to show people how it performs.
I buffed an IHS at a jeweller put a video up, check it out. Didn't do much for thermals
@der8auer, i think you should add delta per cores (coldest one and hottest one under load). Now we never know are there was any improvements. Not really important, but also may be interesting.
P.S. Any news about Carbonaut? =)
P.S. What if i add marking back after i lap CPU (with laser)? Will it count? =) I need my warranty back!
Carbonaut will be available in about 2 weeks :)
sure if you can add it back :D
Will add deltas next time but you can see difference between min and max
Vielleicht versuchst du mal die 3 Schleifstein-methode zum abrichten von heatspreader und heatsink. unter umständen würde das die performance von thermal pads stark boosten.
Could it also be that the paste is now what is causing the temps to not drop lower? From what i understand if you have 2 flat surfaces then there is no need for thermal paste. Similar to optical bonding doesn't use any glues. I wonder if Liquid metal would have given larger differences or Direct contact with the waterblock.
Further Could have it been that the water block was also not flat or at the very least also needed to be sanded/lapped. I know Almost all of EKWB do not come flat. They are concave. So you will reach a wall at some point just sanding 1 of the 2 faces.
To many variables here. Its pretty clear you hit a thermal wall with your results.
I wonder how the results would be if you got both surface optically flat and used no thermal paste.
Would of liked to of seen difference between hottest and coldest core. Per core temp can vary by 10c or more. Would like to know if sanding would close that gap.
That’s exactly what his results show... hottest vs coolest core, did you not watch? 😋
I hate lapping. I have done it, however, it is a lot of work to do it by hand. [EDIT: for sanding I think I used to start with 320 grit and finished with 2000 grit.]
realy good info, I would love to see a ryzen 2700x oc in the same test :D
WHY? This is a universally applicable test. Its ONLY testing the roughness of a lap and removing the nickle.
Kingpin said he roughens up the surface after lapping so the paste can stick better
it's a fine line between having an increase in thermal my micro scratches for surface increase and rough scratches that introduce air pockets or pockets of not full contact that essentially work like micro insulators. As seen by experiment shown there is a tipping point for a optimum.
Versuch doch mal mit diamtenpaste auf 8000 oder 10000 hinzubekommen. es gibt auch noch speziele paste die 20000 macht. Dann ist die oberfläche verspiegelt. Dann hat man den glass aufeinander effekt wenn man den lüfter auch so polieren würde, wo kaum Luft mehr zwischen ist. Somit mehr wärmefläche. Ich denke das alles zwischen 500 und 6000 fast identisch ist. den größten unterschied gibt es bei 40-200-8000-25000. Der punkt ist halt es auf einer perfekt geraden oberfläche zu machen. Also meistens Glass. In einer absolut gleichen Bewegung.
I lapped x5660 and x5675, was definitely worth it on those Xeons, dropped about 8 degrees Celsius. Though delidding is better, BUT...those chips are soldered, so lapping is all you can do
you need to lap or grind even the heatsink to get the best results... almost no paste needed better thermal conductivity. you are just perfecting 50% of the surface... take care of the heatsink to get better results
For a mirror finiah you to sand in one direction. Circles are the worst. Fine grade steel wool after the 1200 will do wonders
That's a great presentation, thanks! Btw, I probably have it misspelled as was hearing this at the beginning (and since I don't have that particular knowledge): what is "ignium", what soldering material are you referring to?
Indium is a metal have better thermal conductivity than copper so its used in CPU to attach IHS in low end CPU instead of indium is used thermal paste
I actually want to know if the IHS is very big, like the AMD TR4. Will it make a bigger difference compares to smaller AM4 IHS after sanding?
very intresting video i liked it
video ideas:
speed run cpu sanding (time to cooling point system)
also would like to see a simular test on some CPU coolers like corsair h100 that generaly has a rough machined plate
I lapped my 5800x3d. I basically delidded it, then sanded down the silicon. The delid itself got me about 10 degrees and the lapping of the die got me another 3 degrees. The delid was as stupid idea, the lapping of the silicon is even dumber.
Have you found the Ryzen I heat spreaders flatter than Intel I heat spreaders? I've seen at least 2 You Tubers with expensive Intel CPUs with an uneven IHS. I would have expected a $2000 CPU to have an excellent integrated heat spreader and not spin like a top...
I get 5-6c better temps sanding my i7 6700k, one good check if you try to spin the cpu on the IHS, if cpu spins it's uneven protruding on center of the IHS so it worth it to sand if you plan OC, if not spin at all it means it's almost flat, so in theory it's not worth the time.
In my case my 6700k spins like hell. I'm on 4,6 OC with 1,32v (not very lucky on silicon lottery)
Hey, why are CPU heat spreaders made that stupidly uneven in the first place?
For example on Ryzen CPUs, they have the corners or entire sides, visibly way higher than the center......... How the hell are the coolers supposed to make good contact??
Has to deal with time. Kingpin went into detail over it. The only level side is the side with Die/IHS. Thermal past makes up the difference on consumer rigs. IHS's are stamped from one big sheet.
@@AbsolutionArmament Oh, ok.
@@AbsolutionArmament has to do with money. Manufacturers are not going to waste money on processes that offer little return for most consumers.
Would using an aluminium heatsink on a bare copper IHS cause galvanic corrosion or would the thermal paste protect it?
ihs are almost always made of copper..since there is no water involved, I would not be worried about. Galvanic corrosion occurs in humid ambients or inside the loop because of the liquid and the different metals, but not between the ihs (which is like 80 degrees at least and therefore not so humid) and the heatsink. I hope I have been clear enough :)
@@enricod.7198 IHS are also nickel plated too. There is also moisture in air normally. Copper exposed to air will oxidize eventually. That's why manufacturers are nickel plating their goods. They're not doing it because they just have money to burn.
@@1pcfred yeah I was referring to the ihs..oh well ok ahah didn't thought the moisture in the air may be sufficient to damage the components, my bad. So is this a real concern if we sand the ihs or with the thermal paste and heatsink applied almost no moisture gets through?
@@enricod.7198 there must be some reason why manufacturers are plating an IHS in the first place. But what that is I don't know. Well I know they're doing it to keep the copper from oxidizing. But why they care I don't know. Is it purely cosmetic, or is there some technical issue? Maybe they just do it because it seems like the right thing to do? It is an interesting question.
What do you think about trying sandblasting the resulting cpu to get the same surface finish as the cpu stock?
IHS is User friendly and has it place, but direct die will Always be king for me and doesn’t get the love it deserves. LN2 direct die doesn’t matter, but for normal people using ambient air overclocking it’s significantly better and needs a bigger spotlight.
Although with modern CPU’s it is difficult to get proper cpu pressure mount, which can really fuck with your cpu if done wrong. So I guess it isn’t for the faint of heart.
aight I might try that with my delidded Pentium 4's
You should lap the silicon die as well... the less thermal material the better thermal pass through
Have you ever used Prussian Blue to help see high and low spots be removed or sanded into the ihs?
Quickshot Gaming i want a prussian blue cat
Dorian Gray one drop of the stuff on your hands and you'll have your wish.
Quickshot Gaming wow!!!!
I would have liked to see your core to core delta during the whole process.
lovely work ! ty for the video
Lapping is fully worth it, ive had great results.
I want to try it, but sadly my 4770k is bad at OC..maxes out at 4,5GHZ.
Old Nvidia Gpus, Up until geforce 4 was really concave in the middle and you gained some cooling by lapping them. That was my experince from the 90ies. Alla intel p3 and AMD athlons had the bare die.. no ihs, but sometimes you cracked a corner and the cpu was dead!
I somehow expected the difference to be a little larger, but I guess it really wouldn't start to matter much unless one started to overclock by a fair bit.
That's why he made the video to dispel an easy misconception to have.
@@1pcfred And it is a good video for that reason.
It at least gave me a better appreciation of what is and isn't a serious scratchy surface on an IHS.
Videos helping to alleviate misconceptions is good, after all, there is a lot of misconceptions floating around, like I once saw a tech channel proclaiming that quad channel memory were having negligible improvements compared to dual channel memory, based on tests on the AM4 socket. Not taking into consideration that it is a dual channel socket, some motherboards have 4 dim slots, but that isn't the same thing as 4 channels.
@@todayonthebench I think the bigger take away from this is how little is to be gained by sanding a CPU. 2 degrees at best does not seem to make any sense to pursue. In another video a guy named Kingpin says he sands the IHS so the thermal compound doesn't crack when he does LN2 runs. So in his case it's doing something for him. But that is a pretty limited use case.
@@1pcfred Yes, sanding IHSes is rather pointless for most average users.
Going to a better heat sink or water cooling setup will have larger gains without voiding the warranty of the CPU.
More extreme overclocking is another story, but one usually doesn't use such systems for more then a benchmark anyway.
@@todayonthebench yeah over at Gamers Nexus they talk about why Kingpin laps his IHS. It's so the heatsink compound doesn't fail when he does LN2 runs. Which is a very special use case.
You should get one of the machines Kingpin has for doing this.
Why? That mainly provides a placebo effect for the most part. The only useful purpose for that is to try to run direct contact WITHOUT Tim, and it wouldn't be capable of that high of precision.
That machine spins the sanding disc in one direction. You want to sand in multiple directions, like doing a figure 8, so you don't get uneven wear.
@@ahnilatedahnilated7703 watch the vid gamers nexus. Kingpin knows what hes doing
@@cee128d cause it's easier and faster. And cause shiny things are nice
i think only worth if the surface is not complete plane.When the surface is warped.
It's curious that in "metric-land" you use grit which measures the diameter of the particles in INCHES. 100 grit is covered with 1/1000 of an inch particles, 200 grit is 1/200 grit etc.
For practical applications US customary measures work better than metric does. That's why we called out the altitudes in feet when we landed on the Moon. Because when inches count no one cares how many meters you are off the surface.
@@1pcfred I was just curious why the Germans haven't changed their sandpaper ratings to something based on the meter.
@@SuperFredAZ it is primarily due to US economic influence on consumer goods. It's why stuff like sheets of plywood come in kooky sizes. For us a sheet of plywood is 4x8. So for the rest of the world it is whatever the metric equivalent of 4x8 is. That's also why 13mm drive is so popular. It's really a half of an inch. 10mm is 3/8s 2.54mm pitch in electronics is a tenth of an inch. etc. etc. I know they make metric bricks. They are god awful ugly! We build jails out of them here. What do you expect from socialism? Freedom units FTW!
@@1pcfred Also tire sizes are all inches i.e. 17 inch 18 inch, etc. but then they go to metric as in 18V235R50, the 18 ins 18 inches, and the 235 is 235 mm, and the 50 is a ratio. Also the metric world measure their television and monitor sizes in inches i.e. 65 inch tv, you'll find them described in Europe as such.
@@SuperFredAZ often prizes do not go to the first but instead are reserved for the most popular. America did not invent tires but we popularized their use. TV on the other hand we did invent. It always annoyed me how TVs are measured on the diagonal though. There just seems to be something blatantly deceptive about that practice. Like what's the longest straight line we can draw in this rectangle? It just bugs me.
Would there be a noticeable difference if you used an aftermarket/custom IHS versus the lapped stock IHS?
I want both surfaces polished down to atomic scale, and cold weld them together in a vacuum chamber. This way you don't even need any TIM.
What the fuck are you talking about? Have u been smoking crystak meth lately?
Would we need to treat the copper after polishing?
Can you lap with a monoblock?
Roman Great vid as always ! When you gonna have the Core i9-9990XE 5Gigs boost for sale on Caseking ?
Maybe with soldered CPU could be higher with cpu vcore. I found one older test on Anandtech forum:
forums.anandtech.com/threads/lapped-my-fx-8350-4-3ghz-oced-temps-lowered-by-9%C2%B0c.2300800/
yea I use whetstones for my lapping and I'd say the scratches make no difference. My 5k grit whetstone has a decent amount of impurities (I want a new one >.
I wonder what would happen if I were to grind it on my whetstones. I got a 12k whetstone. It would be an interesting test I think. But I don’t have a cpu to destroy so I’m leaving that one to someone who wants to ruin both their whetstones and their cpu
What if there is a bigger scratch on it. You can see the copper in the scratch
you should perfectly lap the cpu and the cooler and use no thermal paste at all
@der8auer thanks a lot. I used this method on c2d e6550 and I didn't remember the temperatures but looks cool. Can you preform this test on amd like 2600/2700 or even on a TR4 CPU? Big hug and see you soon :)
i don't see why you couldn't
Thank you as always!
der8auer really loves 8 (the pattern he move the CPU on the sand paper)
hello i just got a 10900k the temps ar semi ok but not as good as i xspected i have 3x360 radsand d5 pump + ek magnatude block after testing the cpu in glass and razor blade it is perect lat yet my ek black is rather convex. the cpu will not spin on glass but as soon as i place ontop of the water block i can spin th cpu very easy
im wondrng if its worth lapping the water block
9900k die sanding was a insane benefit. Still want to know why intel did that...
nice video :) have a difrent question thou , i bought the Lian LI o11 Dynamic case you was helping designing , but i cant get my thermals down to the lvl i have with the Fractal Design R5 i using water cooler for the cpu combine with air cooled gfx card, i have som friends telling me its becouse its the 90 Degree bend the air is need to take that it less effective cooling and Air flow , got any tips to fix that?
tbh i love th design iof the o11 Dynamic case , kinda notice the mash/filters allso messing with the aif flow and termals .. im abit annoyed i cant find A water cooling block for the Gigabyte RTX 2060 Gaming OC card if you could help me out that too it would be great
thanks