Maybe the problem is reverse: Intel had to increase the thickness of the die (for a reason unknown to us) therefore they had to use stim because the regular toothpaste tim would not have been good enough to cool the 9th series. Which would explain why they went for solder. Maybe I’m too cynical but I don’t really buy that they just listened to us...
Pustul Anyway they did it ether for marketing OR technical reasons. But in any way I really like and appreciate it. Haswell was a Great leap for Micro technology in retrospective and why NOT go back to something everyone can agree it works even if it means a bit more cost for Intel. I think they thought the same thing internally and just went for it. At the end of the day they laugh about the money they lost...
@@ppsarrakis I believe a 2700X @4.2GHz 1.4V also pulls around 180-200W but I think it wastes less power therefore dumps less heat? Also the die surface area is bigger so I guess can't really compare
>Overclockers: Intel should use solder instead of their shit TIM >Intel: okay *proceeds to make the die ridiculously thick, removing the benefit of the solder*
Neiva you lost me there, who would that random kid be, as der8auer is not random and not kid. The facts in the video are solid. It is not a question whether they needed to add thickness of the die Im sure there was some reason because of soldering, but still results are the same, you need to lap the die thinner to get better thermals.
So when is caseking gonna start selling die sanding kits? Cuz I want one. BTW IIRC Splave said that for LN2 it's better to just run the chips stock and not bother with delidding. Is that correct?
The benefit is that the solder has far less expansion/retraction. Liquid metal would be inappropriate because of the sharp drop in thermal conductivity when you cool it to the point of it beginning to solidifying.
"I have no doubt that derbaur is a good delidder but it is prohibited for anyone to treat a cpu the way he did on this video :D." - It just voids warranty. Btw. the casking pre delidded cpus will have standard german warranty of business to customer 2 years i guess. Anyway you can do with your property what you want.
I was about to ask the same question, for LN2 or Cascade (-95°) do we have to delid ? I really hopes that all this soldering thing doesn’t just bring more complexity
So Intel can launch a new 9900K Refresh with a thinner layer of sillicon and keep the shareholders happy, and have new products to sell untill they have fix'd 10nm/7nm produtcion! (I am so disappointed in Intel CPU)
CADesigner.... people want more cores but they want to stay at the same mhz so their badly optimized games wont lag... since the manufacture process didnt change much from 8700k you cant just magicaly make heat dissapear even Intel has to obey the laws of physics......
Well, justified or not. 90C is NOT ok, is it now? And we're not even talking about some heavy OC'ing. I would not buy this, luckily I wasn't going to anyway. I'm waiting for 9900X, and if that has some weird issues as well, off to 2950X I go.
I mean sure, I've been an IT nut for years (my CPUs have been under water for over 19 years now) but wow, there must of been some nerves while doing it :)
Great video as always! Question is, how thick were old soldered Intel dies like 2500k/2600k or 5960X and how thick are AMD Ryzen died in comparison to 9th gen Intel?
That's what I wanna know. Everything here just seems so strange... I hope enough of a fuss is raised that Intel gives an explanation, but I doubt it we'll get one (officially).
See der8auer's delidding video on the 1800x, the solder layer seems much thinner than the 9900K's, and replacing the solder with liquid metal nets a 1-3°C temperature drop at the same testing conditions. No one has pictures of a delidded 2500k/2600k/5960x so we can't compare, but I bet it's thinner too.
@@josefinagarcia7082 yeah because delidding a ryzen cpu is unnecessarily unlike intel, who now requires someone to sand down silicon to have proper temps on their system. What an absolute joke, they have been kicked down from their pedestal after years of no competition. Next gen consoles are will alsoreportedly use ryzen 2 apu's so amd will completely dethrone intel. Exiting times for sure!
Two questions: - What would be the lowest we could grind it down to safely? - Is applying liquid metal on the ground silicon a problem? And if it is, is there some way of making the silicon less permeable (with some sort of oxydizing maybe)?
In the german video he said "I cannot tell what the long term consequences of grinding it down are, because you have to consider that (liquid metal) atoms can diffuse through the silicium and then there's there question 'is there any influence on the CPU itself because of that' because normally atoms are not supposed to penetrate through the silicium due to the diffusion barrier that is on top of the silicium to prevent any atoms from penetrating through, which then is removed by the grinding, so it is hard to tell if there are any technical influences because of that. But here at CaseKing we still decided to offer delidded 9900K CPUs..."
Thanks for confirming. I'll make a bit of research over what this barrier could be made of. Maybe with some chemical reaction we could make a new one after grinding the CPU. Edit: according to this article (overclocking.guide/the-truth-about-cpu-soldering/ ) it's made out of titanium and nickel-vanadium. I guess you would have to somehow plate titanium on the ground surface then nickel-vanadium on top of that. Except titanium seems to be exceptionnaly hard to plate. Edit2: the titanium and nickel-vanadium are apparently only for soldered CPUs, so I'm guessing the non-soldered CPUs have another type of diffusion barrier .. probably something simpler like some sort of passivation reaction. Silicon dioxide seems to be the most common silicon passivation technique except silicon dioxide is an insulating material .. I'm scourging wikipedia for high thermal conductivity silicon derivatives.
What a boring life that would be .. it's just so much more interesting to learn about all this technical stuff and try to push the limits of what we can do. And the Ryzen is maybe cheap but miles behind Intel in terms of efficiency and optimization. Anyway in the meantime I found an awesome website (www.eesemi.com , great read) which suggests that the protective layer is silicon nitride (www.eesemi.com/glassivation.htm ). But the different processes used to apply silicon nitride seem really insane, one involves plasma and vacuum, the other hellish temperatures. And in the end, silicon nitride has a pretty bad thermal conductivity, so if you deposit too much you risk loosing any benefit from the grinding. I suspect there is no real solution here apart from using conventional thermal paste that does not attack the silicon.
@@nout4269 "Ryzen might be cheap, but its miles behind intel on efficiency and optimization" - This proves that you are a intel fanboy... Firstly, i know that intel's chips are faster for gaming, just wana make that clear, but Zen is a new architecture, it's only been out for 2 years, of course AMD optimization isn't gonna compare to to Intel's now 3 generation old architecture. Secondly, do i have to remember you that Ryzen is on 12 nm, instead of 14++++++++++? Ryzen is not FX. The difference in power consumption is at best 5-10W between these chips, as could be seen in this graph : goo.gl/images/hHHmS4
Please grind down the chip as much as possible while recording the changes in temps in increasingly small steps until the metal layers. Thanks for the great work!
"We do by watching the vids" ... - rofl his yt revenue is if it comes close 200€. Nopp addrevenue ain't do the job. If anything it would be his patreons. "Caseking is his sponsor I think. They have quite a bit of CPUs. :D" - Sure they have and for public relation they will help him out. But they have a budget on how much 8auer can burn through,...
And leave AMD alone and complacent in the CPU market so we get another Bulldozer Piledriver fiasko? Sorry, no...intel may be dethroned, but we do NOT want intel to die, no matter how bad they are, it would be worse if they leave.
@@Chrinik How are they even dethroned?? They currently have the fastest CPU out there in the 9900k. In gaming they have the second the 8700k. Hardly dethroned...Infact...very far from it by the looks of it. I
I'm surprised intel doesn't thin the wafers at the factory... I was an engineer at an HP water fab in the early 2000's and all our chips underwent a thinning process back then to get the best thermal and electrical properties....I still have a load of the diamond 'paper' and compounds... Better keep hold of them for when 9900k gets cheap enough to play with
The SEMI standard for 300mm wafer thickness is approx 785 microns. So the work indicates that Intel is using the native wafer thickness, which really means that the prior nodes were backside ground to something thinner, which means it costs more. Could be as simple as competing on cost while providing sufficient cooling for their product specs..overclockers be damned...
What do you think you pay the extra for? Intel is giving you a change to start a DYI project which you can share with everyone. (Attention! Your results may vary!)
majnu100 Unlikely, as DDR5 is expected to come out in 2020 or 2021, which is part of the reason AMD says they will support AM4 until 2020 (other reason possibly being PCIE 4) even if DDR5 came out early, AMD likely wouldn't be able to use it without breaking AM4 support
Yep. I was considering getting a 9900k because some of the software I use for work doesn't run well on amd hardware. until I saw the temperature and power tests and this video.
When I was in college I was told that sanding using figure eight motion will cause the flat surface to eventually become convex. The procedure I was taught was to sand in straight lines five times in one direction then rotate ninety degrees and repeat giving equal strokes to each rotation.
On other videos from der8auer and on my experience of lapping the IHS only of my delidded 3770k few years back, It was shown that both the Naked Die silicon and the Top of the IHS are actually not 100% flat. Heres a der8auer video : ruclips.net/video/tnd2LO0IBic/видео.html at 4:25 showing that the Die is not 100% flat. at 8:06 of the same video shows that the middle part of the die is convex (its not flat from the get go). and then IHS can be Concave or Convex. In my experience, when I lapped the nickel plating on my 3770K IHS on a flat glass table and tried my best doing so. It showed that the IHS on my 3770K was concave right on the center. Whilst I saw some AMD chips had a Convex middle of the IHS right on top of the silicon die. (FX8350)
yeah figure of eight isn't good for when you really do need that perfect flatness although it doesn't matter too much in this case as long as you're getting a smooth surface
Of course not, why would anyone want a fake 8/16 CPU if intels 6/12 betas in gaming and 8/8 [9700k] beats it in everything, be it software, zip, games, HEVC, etc. 9900K Rapes it
@Wellington Campos, what do you call a 8c/16t CPU that loses to 8c/8t CPU if not fake? What do you call a 8c/16t CPU that also loses in many things to 6c/12t CPU? Do you know ZEN architecture? Each 8 core ZEN is in reality Two 4 core CPUs "glued together" [they use Interconnect], thats why RAM speed has such Great effect on Ryzen benchmarks but not on Intel and thats why AMD CPUs cost less money but Intels CPUs cost more, because they more complicated to build. Do you have Ryzen 2700x? In its control panel it has Game Mode, this game mode, disbaled One SSX [one CPU] and only leaves 4 Cores working, its done because many older games see Ryzen PC as system with 2 CPUs and they not optimized for it.
@@zg2111 I'm really happy with my 1700x, but can't wait to upgrade to the r7 3700 when it's released. Amd also bundles a free optain competitor in to their second gen and later ryzen boards, so it's gonna be so worth it.
One thing that the video is missing - is comparing the chip thickness to 2700/2700X, for example, and compare Intel 9600/9700/9900K solution's to AMD solution's.
the 9900k and the 9600k seems to have the same Die and Die Size, the 9600K is basically a 9900K with 2 disabled cores and HT Disabled while the 9700K is the 9900K with the Hyper Threading Disabled.
FX would thermal shutdown well before this. I believe max temp on the spec sheet was like 69c, but I'm pretty sure you could set thermal shutdown limits at 79c if I remember AMD Overdrive correctly. I don't have to worry about it much though, mine rarely ever peaks beyond 0c at 1.7 vcore on a small phase cooler, only a few synthetic loads can make it do that.
@@PimpMatt0 Yeah, it would show 10c when ambient was slightly higher (so impossible on air/water). FX is only accurate at higher temps (non-linear thermocouple) I think its still below 0c though, my block temp peaks at -25c after a really long load on synthetic loads (-35c in Firestrike and Timespy, it needs to use more power than that and for a much longer time, like a few minutes). Idle is -58c on the block (I didn't drill into my CPU to put a temp probe, so the only way for me to get temps is from the probe in my cooler, and from software for core temps, but software won't show below 0c, I just know it runs cold).
_Exactly_ ... Only thing is, the FX 9570 as well as the FX 9370's *EST* is precisely at *57°C* (sic!). On _everything_ above and beyond this so-called _Emergency Shutdown-Temperature_ the cpu itself will electrically emergency cut-out itself physically. Then there's a *CLICK and the system is down immediately. … and as you can figure out already, that makes all those stories about Bulldozers reaching like +100°C pure nonsense and bullshit (straight-out lies to hamper the bulldozer's reputation, FUD?) - as all the ESTs of FX Bulldozers are either 57°C, 61.10°C, 61°C, 70.50°C or 71°C respectively (depending on model).
Maybe check i7 2600k’s (or any Sandy Bridge, I guess) die thickness as a reference, since that was the last line of CPUs (mainstream, excluding other extreme CPUs) to be soldered to the IHS. That could be a starting point. Then compare that to say a i7 3770k, when they started to used thermal paste instead. Maybe that’ll give us an idea as to why it has to be so thick. Also, I love what you’re doing as far as research goes when looking into these sort of things. Some may say it’s pointless, but I find it very intriguing and informative on the manufacturing process as a whole. Hope you never stop. Great info thus far. 👍
My theory is that it is because intel made the 9th gen using 450mm wafers, and the 8th gen was on 300mm. According good old wiki, bigger wafers have to be thicker to have the mechanical strength to support its own weight, so that it does not crack during handling.
@@ellsworth1956 they're ground down and then wafer dicing separates out each individual die. The wafer needs to be mechanically stable in between these two processes. Scaling of bending stresses is nonlinear, and I can well believe that 50% greater wafer diameter requires double the *ground* wafer (and die) thickness.
It doesnt entirely make sense tho. Better thermal conductivity could mean lower cpu temps and same overall heat output, or it would allow you to push the chip harder increasing output. You want the heat out of the cpu, not trapped inside!
This is actually very funny. The intel would turn green with rage if they heard you compare the 9900k to a fx 93xx. But, they are both 8 core space heaters running at 5ghz. Personally, i think amd is killing it with ryzen, and the path ahead seems to be amd. However, jim keller works for intel now, so rip core, rip ryzen. They had a good run, but jim k will make intel billions
This only proves that AMD is better companies that Intel because Intel had tons of money at there's disposable, and had to buy AMD personal to fix their mess, first was raja koduri and now Jim kepler... So AMD has better engineers but Intel being Intel and trying to cripple AMD by buying employee from AMD.
conspiracy: chip is thick so intel can brag about improvements in the future. "Look! We cut degrees by 20*C! *picks up Novideo book* Don't delay, buy 10900k today!"
Thank you for all the good information, Der8auer. Without the expertise of reviewers like you, there would be no oversight. Your work is greatly appreciated.
Intel finally investing in solder for a single generation, just so they can half ass it and say "I told you TIM was better" and have an excuse for going back to toothpaste and therefore cutting costs next generation
but that is risky for intel because 9900K is considered a milestone chip before they shift to 10nm or 7nm architecture. plus 14nm chips are low in stock aren't they?
The thicker wafers likely cost Intel as much if not more than the soldering process since that means fewer wafers per ingot. That's even worse considering that there is an on-going ingot/wafer shortage. If the main reason for the thicker die is indeed to take up the thermal cycle strain from solder TIM, then the opportunity cost of soldering really isn't worth it from a business standpoint, better off getting ~30% more wafers per ingot with a thinner die and paste.
Lagittaja this should be discussed more. If this is exactly the same thickness as the sandy bridge die or any of the X99 CPU dies, then idk what the big fuss is. If it's thicker than the previous soldered dies, then I would start wondering why Intel did what they did.
I am still running my 2600k from april 2011 at 4.8 GHz with an voltage of 1.350 V (+0.125 V in OFFSET), and it only hits 76 C on its hottest core, no delidding just attached a normal big air cooler (Scythe Mugen). I was hoping we would get back to something like here 7½ years later...
Maybe it is because the soldering process. Thicker die so that the risk of damaging the CPU from the heat application when automating the soldering process is lower? Could be a tolerance issue or a manufacturering impact as well.
I'm thinking the added thickness might have to do with the soldering. Previously the problem was thermal cycling causing spalling of the silicon, which is why Ivy Bridge's die shrink didn't allow for soldering, at least according to what I've read from researchers. Maybe the added thickness prevents this?
Or it could be another con by Intel that they will magically bring out a new version of the CPU with better performance as a "tock" version of the 9900K and the only thing the will have done is made it a bit thinner?
I never got that AMD meme. I have an OC 8 core FX and it's one of the coolest CPUs I have ever owned since my Celeron 766Mhz. I run it on passive water cooling with temps never getting over mid 45s on full load. My GTX 980 pumps out far more heat with a full water block, making up about 2/3rd of the total system heat output. The hottest CPU I personally ever worked with is i7 975 XE, custom loop at 4GHz it ran around 75c water temp on dual 120 x 50mm rad. with 2 x 3500 or so rpm delta fans. That thing was hot!
You should also consider making a "jig" to hold your cpu or ihs perfectly flat and perpendicular to your sandpaper while your grinding. I saw that you hand hold it and rub it on the sand paper. Even if your paper is on a 100% flat surface, it's still impossible to hold it 100% flat with your fingers. The grind could be at a slight angle, even if that angle is in fact perfectly flat. I highly recommend this.
Prettt much exactly what I thought would happen as soon as I heard the 9900k was going to be soldered. Indium based solder does not conduct as well as gallium based LM and these were never going to do as well as a delidded 8th gen so no surprises there. Intel were between a rock and a hard place with this one. Probably needed to solder it to keep it cool enough for general usage by non-enthusiasts. Unfortunately that means delidding for LM is a lot more tricky for the small number of customers comparatively Ho wish to delid. However it’s very interesting to see the results after grinding! I hadn’t considered that the chip would be thicker.
This is from Indium Corp on their website: Compensate for a mismatch of the thermal expansion between two mating surfaces. Different materials expand and contract at different rates, creating a need for some compensation in any intermediate layer (like a thermal interface material). The other candidates are not as forgiving as indium. "Indium is really, really soft. Its flow stress is is about 150psi (1MPa). This allows indium to conform to the mating surfaces and to absorb deformation due to differential thermal expansion without transferring the shear forces to the delicate semiconductor materials," says Bob. It will also optimize thermal conductivity by its increased contact even on imperfect surfaces. Bob adds, "Indium conforms to an asperity, bump, or dip in the interface surface where a harder material would simply lose contact around such obstacles. Indium maximizes the contact surface area."
thermal conductivity of conductonaught is about the same as indium solder. The issue is layer thickness. When you apply liquid metal, you essentially paint it with a negligibly thin layer. The indium solder layer is 0.5 mm.
HOW DID RYZEN GET IT RIGHT 2 YEARS AGO? Intel engineers are getting so lazy or the business environment is so bad that they have no chance to be competitive in the future
I'd be curious to see a comparison of die thicknesses between Ryzen and 9th gen chips, and thermal/overclocking results with both as the silicon gets sanded/lapped thinner and thinner until the point of failure
Their room for error is much larger because they aren't really trying to push high clock speeds. Their chips can't keep up with Intel's for high-refresh gaming, though.
@@seanmcc09 I would actually say that AMD are trying to push high clock speeds, but just don't have the hardware. The overclocking headroom on their X sku chips is practically non-existent; single cores in my 1700X spike to 1.45V stock.
The funny thing is yes the Fx Series were space heaters but they maxed out at 80 degrees and they would do that all day long I want to know how long it will take until they degrade the silicon at measured 95 degrees
Will the direct die bracket work if you sand down the die? That would seem to yield the coolest temps(except for exotic cooling). I'm running an out of the box 9900k with an AIO and it gets pretty warm. I've considered de-lidding but it seems the real problem with this chip is the die thickness...
It's funny because intel are the ones who came up with this soldering technique and have the patent on it. AMD has to pay intel a licencing fee to use it :P
10:00 WHAT?????? No! Transistors spikes are sunk toward top side of silicone! Close to the side where heat-sink is. Closer to the PCB the bigger metal layers, closer to heat-sink smaller details. Pattern looks like arrowhead one next to another with spikes pointed toward heat-sink. Regardless of how thick initially wafer was process of etching silicone remains the same, and thickness of active layer remains the same for both Coffee lake and 9 series. Digging deeper and deeper from PCB side the smaller patterns (holes) are made then admixtures applied, metal layers, oxidization etc.. To be able soldering heat-sink and to to prevent top of die crack destroying transistors stretched up by iridium, also cause die is longer initially they cut thicker wafers using more material, so in result top side is little bit thicker than usual and there is also additional metal layer on the top that makes it even thicker. In result more distance between transistor heat-sink. Im curious how 9 series dimentions compares to Ryzen's silicon die cause AMD got even larger dies, also soldered.
Damn it... So to get the best performance now we have to add removing solder and lapping to the delid, a lot more can go wrong now 😡 7:57 Maybe i missed it somehow but whats the cooling solution used in these benchmark results?
Just buy used 7700/8700K and you won't be needing anything better for games, nor will you get any performance hit on a desktop environment (like browsing, watching videos) from Ryzen.
@@janchovanec8624 honestly you dont notice a difference with 4k between ryzen and Intel. I'm guesaing if you can afford a 9900k your playing at 2k or 4k. So ryzen still is the better bet.
@@corwinblack4072 if memory serves the P3 was the first one to be that hot. And while the P4 topped it it was still the P3 heat shock that stuck with me.
I will be performing my first delidding and lapping of my 9900K die next week. What I would like to know is does the bottom of the IHS (the bottom that actually touches and glues to the PCB) need to be sanded down equal to the amount that is taken off of the die so that the lid touches the die? I have not seen this done nor discussed. Thx for your time, knowledge & help! You are greatly appreciated!
Joey I’m doing this right now... just lowered die 0.53mm mirror finish. Now I have to lower ihs same amount. It’s coming out great can’t wait to see where temps end up...
@@hermanquinones Awesome! Did you cut the die first and then sand it down? Is .53mm safe? I was going to do .40mm. Please let me know how it turns out! Very interested! Much success! :) Thx!
Joey No cutting at all... bought the supplies for delid and solder removal from RockItCool. Sanded starting at 240 grit and worked up to 7k for mirror finish. Spent majority of time at lower grit 240 & 400 to remove initial .30mm. Flitz metal polish was finishing touch. But total amount removed was 0.53 mm.
@Joey Safe? To be honest we are in uncharted territory. That said I believe the height being close to stock 8700k should be fine. Also, don’t be surprised if surface cracks are exposed on initial sanding they go away when proceeding.
@@hermanquinones I'm looking forward to seeing your before and after temps... and Cinebench R15 scores. Right now my best is 2274 at 5.2 GHz and 1.470V. I should be able to delid next week and hoping to see some great improvements :-) I can Boot It 5.3 but it is not stable when i running cinebench most of the time. It is stable running my Photoshop speed test.
+SaintQ perhaps you've misunderstood my intent? Unfortunately I find people like you tedious so I won't waste my precious time trying to explain it to you. I do suggest you invest your own time attempting to figure it out though. Good luck.
@@1pcfred What intent? The kind of intent that undermines logic? Wyatt was not right or wrong by imagining future scenarios, that is not a bad thing as if said scenario does happen he can make that point... that is how humanity gets smarter by thinking. :D Fetcher.
It's common knowledge that CPU manufacturers intentionally hobble chips in various ways to differentiate pricing, for example, disabling cores and lanes in microcode or whatever. It's not too far-fetched to imagine they might also intentionally create a heat issue that limits performance, such as inadequate TIM, too-thick silicon, etc., which is what Der8auer's video made me think of.
Thermal conductivity is a material dependent property. It does not depend on the thickness of the material. I think that you meant to say that as the thickness of the silicon increases the heat transfer rate increases.
Yep, even with so small of an AIO you can see my CPU sticking out all around AIO, my Threadripper 1950X runs about 65c at ~4 Ghz (cooled by a 120mm AIO, gen 1 Asetech pump made for DELL XPS before Alienware existed, so its a very, very old cooler). I'd say 3.9 Ghz is about right for 65c (just so the fan is quiet since I sleep close to it) running 4x Handbrake encodes at once in a batch run.
Question, please, to anyone that knows about this stuff. I'm half way through the process. I've removed the IHS and scraped the solder off of the inside of the IHS and the CPU die. That part was way easier than I thought it would be. So now I'm about to sand down the die, but I can't find the sandpaper he's using in this video. Where can I find it? Or can I just use regular 3M sandpaper? Also, what grit/micron is the 2nd sandpaper that he's using?
Maybe by the time you can achieve that much tangible benefit Intel will have their shit figured out. Or they won't, i'll be even happier to be honest :D
I was thinking of replacing my 6700k running at 4.9Ghz for Ryzen 2700x, however more cores wouldn't give me any benefits for games and desktop activities because of its low clocks, so I'll in stead wait for another generation.
why is it so thick? a lot of people are reporting cracked dies when they delid. a few of them have sanded them down to find that the cracks disappear very near the surface. so it could be that the tension from soldering and then the different materials cooling at different rates is causing forces to crack the die and the top layer is there to strengthen it and maybe to act as a sacrificial layer.
This ain't good, so will a NH-15 Noctua cool this thing even at stock? Cus that's what I ordered lol. I'm wondering if they needed the extra thickness for the solder process, to dissipate the heat so it don't degrade the chip.
I feel Intel kinda boned the pooch on this one. By refusing to price it competitively they are fighting a two front battle. On one side the 2700x which is cheaper and slightly loses out to it for gaming. But on the other side with the price of the chips and mobos puts it in thread ripper land... Would be neat to measure die thickness of and verse Intel parts.
Now that reviews finally came out, looks like stock you're right, shouldnt' be a problem around 80C (that's what my 3570k Runs currently when playing GTA).
Really nice video der8auer. Informative and entertaining lol! that Risky STIM Delid + die lapping ! And if that thickness is really affecting CPU Thermals. I guess Its okay to say as an example that the 8600K is a way better choice than that 9600K (8 core with 2 disabled cores) on the intel side atleast I feel like, not only the 9600K doesn't have much of an edge against an Overclocked 8600K. The 8600K easily goes to 5GHz. Even better chances of hitting that 5GHz if its delidded. And its cheaper brand new vs a 9600K, and even cheaper if bought 2nd hand. And has a more petite Die that does not have any Overheating issue. But Damn, really nice video man! proper entertainment :D First the PT's meme benchmarks, and now this Thicc Silicon. The good +++ The Bad #1 Intel Finally made an 8 core 16 thread to go against AMD's 8 core 16 Threads +++ But its only marginally fast and its VERY expensive #2 Intel Finally Used S-STIM +++ But it still runs hot because of the Thiccer Die and quoted from someone saying, perhaps the die was made thiccer to have better yields from the manufacture process. This is why 3rd party reviewers and people like you der8auer are the Good guys. It is not fun buying a Product at pre-order not knowing what it entails. Specially if the product cost a lot i.e. 9900K for 719 Euro. I am happy with my used and delidded 8600K and 8700K XD Seriously though man, You and this videos of yours are Epic! You deserve more! Most tech tubers only focus on benchmarks to be honest. Not many goes in and Delid a soldered CPU and resolder them again then re-delid them. But also freaking Scrape the S-Tim on the Die with a blade and freaking Sand/lap it down! That is crazy! but it is very entertaining to watch and to see how much it affected temperatures. Seriously, Good Stuff! :D
I was expecting that the 9900k with solder would be able to do 5.0GHz on all cores with a big air cooler and stay under 80 C on the hottest core, but to reach that is seems you need buy a pretested golden delidded lapped CPU with liquid metal applied, WTF...... My 2600k from April 2011 is still running at 4.8 GHz with an voltage of 1.350 V (+0.125 V in OFFSET), and it only hits 76 C on its hottest core, no delidding just attached a normal big air cooler (50$ Scythe Mugen). It ran 5.0 Ghz with an voltage of 1.425 V (+0.200 V in OFFSET), but didn't like that the hottest core would reach 85 C, so turned it down to 4.8GHz around 5 years ago. Looks like my 2600k is going to last yet another generation, but to be honest at 1440p ultra it still seems fine with a GTX 1080 Ti, matches my friends 1800X in games anyway.
@Martin - I'm still running a 2500k @ 4.6GHz as well because CPU progress has stagnated and replacing half of my system for a small performance boost hasn't been worth it yet.
@Salt Maker That's £350 for the CPU and motherboard for a roughly 35% performance boost. Not worth it still when my 2500k can still handle 9/10 games and £350 on a GPU gets you a 100% performance boost.
@Salt Maker 24GB of 1833MHz. And at 1440p the 2500k can keep up with a 1070 which is what you'd get for £350. But I'm just gonna block you know because you're a prick :)
When sanding the die, are you still able to use the direct-die frame / oc-frame? Caseking states that the frame itself is only 0.1mm lower than the actual die. So if you sand the die more than 0.1mm the cooler / coldplate won't touch the die. In this video the die is sanded 0.20mm and then the oc-frame shouldn't be compatible any longer. But the video doesn't say if the frame was being used or not after this. Or if the IHS was mounted back and the standard socket was used. If using the IHS after sanding down the die, you will probably have to sand the outer edges of the IHS 0.2mm also, to make contact between the underside of the IHS and the newly sanded die. If using the die without any frame at all, and simply mounting the cooler directly (as in actual direct-die cooling) there might be contact issues on some of the pins along the edge of the socket since you are only pressing down directly above the die.
@@sirius4k No, but it might be funny to hear whatever cocksure reply Intel gives this time since they've been caught bsing for the 3rd time in a month.
They won't say officially shit, but you can be sure, he has contacts to them - he has also mentioned that he had reached out to his contacts, but nothing specific was replied
I think that the die lapping might be even better than you describe, since you didn't mention (unless I wasn't paying close enough attention) that lapping is actually removing a lot of silicon nitride, which has a WAY worse thermal conductivity than pure silicon (4-5 times worse, to be more specific). It doesn't change the end result from the actual tests, but I suppose it helps to better explain WHY they went down as much as they did.
There are research papers out there regarding the silicon fracturing under thermal cycling. Your hypothesis about end-user pressure to solder and then Intel following up with the solution requiring a thicker die due to mechanical stress is right one point.
As a chemical engineer I think I would look at a better cooling solution, as multi-layered conduction heat exchange is usually dependent on the external temperature of the final layer i.e. air or water in pc cooling. Fresh air would be cooler than recycled water, giving higher delta T's. But the internal design of pc's and noise could limit what is possible. The area that heat transfer occurs across is also a potential problem as many people exclude it from their calculations. If one layer is smaller than the others it could restrict the performance of the entire system.
Keep in mind that no testing was done with the poor intel thermal compound. This could be a case where with the stock compound, intel simply couldn't get under 100C with any of today's popular coolers.
I think it's great that you took the time and did this It is a very interesting result and needs to be investigated further for modders Thank you and well played Doing the PC lord's work der8auer XD
With solder there is no phase change within operational range, and the solder layer is aready _much_ thicker than LM, Cryonaut or KPx would be, so its mass should accomodate the solder to take on the stresses of thermal contraction, rather than passing (most of) these on to the silicon or the solder-silicon bond. As long as there's no phasechange, no fractures, the solder should be fine all the way down.
Maybe the problem is reverse: Intel had to increase the thickness of the die (for a reason unknown to us) therefore they had to use stim because the regular toothpaste tim would not have been good enough to cool the 9th series. Which would explain why they went for solder. Maybe I’m too cynical but I don’t really buy that they just listened to us...
Or vice versa.... Thinner die wouldn't survive thermal cycle.
Pustul Anyway they did it ether for marketing OR technical reasons. But in any way I really like and appreciate it. Haswell was a Great leap for Micro technology in retrospective and why NOT go back to something everyone can agree it works even if it means a bit more cost for Intel. I think they thought the same thing internally and just went for it. At the end of the day they laugh about the money they lost...
do we know of any other cpus with the same TDP as 9900k and their die thicness?
they went with solder to get the higher clocks. thats it. to be more competitive.
@@ppsarrakis I believe a 2700X @4.2GHz 1.4V also pulls around 180-200W but I think it wastes less power therefore dumps less heat? Also the die surface area is bigger so I guess can't really compare
>Overclockers: Intel should use solder instead of their shit TIM
>Intel: okay
*proceeds to make the die ridiculously thick, removing the benefit of the solder*
its like this guy who is too lazy to work and fucks up his work order on purpose so he must not do it again.
Bingo!
>intel: 1-800-VERGE. How much TIM did you want on the 9900k again? Put more on top? No problem!
So is this like the classic genie in the bottle kind of thing? What you want really bites you in the ass when you get it.
Neiva you lost me there, who would that random kid be, as der8auer is not random and not kid. The facts in the video are solid. It is not a question whether they needed to add thickness of the die Im sure there was some reason because of soldering, but still results are the same, you need to lap the die thinner to get better thermals.
Pentium 4 sends its regards!
HAHAHA, made me laugh!
Yes, it seems a repetition of the "Presshot" era (and people will still claim AMD chips are heaters).
10/10, it's all about the Pentiums.
Pentium 4 has a low IPC
@Zero11s You don't say? Someone completely missed the point :P
So when is caseking gonna start selling die sanding kits? Cuz I want one.
BTW IIRC Splave said that for LN2 it's better to just run the chips stock and not bother with delidding. Is that correct?
You need a kit for that? Why not just do it?
The benefit is that the solder has far less expansion/retraction. Liquid metal would be inappropriate because of the sharp drop in thermal conductivity when you cool it to the point of it beginning to solidifying.
"I have no doubt that derbaur is a good delidder but it is prohibited for anyone to treat a cpu the way he did on this video :D." - It just voids warranty. Btw. the casking pre delidded cpus will have standard german warranty of business to customer 2 years i guess. Anyway you can do with your property what you want.
I was about to ask the same question, for LN2 or Cascade (-95°) do we have to delid ? I really hopes that all this soldering thing doesn’t just bring more complexity
Normally on LN2 we run Kryonaut/KPx. I was wondering if those could beat solder.
"why the hell is this chip so damn THICC ?"
Vladek16 true Story 😏
Lol. Of course you would say that Vladek16.
So Intel can launch a new 9900K Refresh with a thinner layer of sillicon and keep the shareholders happy, and have new products to sell untill they have fix'd 10nm/7nm produtcion! (I am so disappointed in Intel CPU)
I asked my ex the same thing. She also blamed me.
Intel is probably saving their 10nm arcitecture for something bigger like graphics card production?
sTIM -> Shitty Thermal Intel Memes
Ayy
Oh I can't wait this is gonna be funny!!
MD
AyyMD
😂
M should stand for material xD
This is absurd, Intel is manufacturing worst CPUs each year in terms of temperatures, It doesn't make any sense.
What are you complaining about? Heating season is upon us! Crank up that CPU and enjoy that Intel warmth in those cold autumn/winter days and nights.
They are following in AMDs footsteps, get ready for their comeback in ~4 years.
CADesigner.... people want more cores but they want to stay at the same mhz so their badly optimized games wont lag... since the manufacture process didnt change much from 8700k you cant just magicaly make heat dissapear even Intel has to obey the laws of physics......
Well, justified or not. 90C is NOT ok, is it now? And we're not even talking about some heavy OC'ing. I would not buy this, luckily I wasn't going to anyway. I'm waiting for 9900X, and if that has some weird issues as well, off to 2950X I go.
whats in your current pc?
Omg lapping a die, the risk level is insane!
whatever works :D
I mean sure, I've been an IT nut for years (my CPUs have been under water for over 19 years now) but wow, there must of been some nerves while doing it :)
It seems less risky than delidding to me. Looks a lot more controlled with a lot less force :D
Can we get a der8auer tool that will micro lap die's ?
I know of soldered dies that literally stay on the IHS after delidding. Lapping is a much more controlled process, believe me.
Der8auer always pushing the limits!! Love it!! Great content.
yaa bro
gold content😊
Great video as always! Question is, how thick were old soldered Intel dies like 2500k/2600k or 5960X and how thick are AMD Ryzen died in comparison to 9th gen Intel?
That's what I wanna know. Everything here just seems so strange... I hope enough of a fuss is raised that Intel gives an explanation, but I doubt it we'll get one (officially).
See der8auer's delidding video on the 1800x, the solder layer seems much thinner than the 9900K's, and replacing the solder with liquid metal nets a 1-3°C temperature drop at the same testing conditions.
No one has pictures of a delidded 2500k/2600k/5960x so we can't compare, but I bet it's thinner too.
@@josefinagarcia7082 yeah because delidding a ryzen cpu is unnecessarily unlike intel, who now requires someone to sand down silicon to have proper temps on their system.
What an absolute joke, they have been kicked down from their pedestal after years of no competition.
Next gen consoles are will alsoreportedly use ryzen 2 apu's so amd will completely dethrone intel. Exiting times for sure!
very good question. those 2nd gen i-cores were incredible overclockers in their time
@@fiqirr intel never had anything in the console market, they've always been using AMD so there's no real difference
Two questions:
- What would be the lowest we could grind it down to safely?
- Is applying liquid metal on the ground silicon a problem? And if it is, is there some way of making the silicon less permeable (with some sort of oxydizing maybe)?
In the german video he said "I cannot tell what the long term consequences of grinding it down are, because you have to consider that (liquid metal) atoms can diffuse through the silicium and then there's there question 'is there any influence on the CPU itself because of that' because normally atoms are not supposed to penetrate through the silicium due to the diffusion barrier that is on top of the silicium to prevent any atoms from penetrating through, which then is removed by the grinding, so it is hard to tell if there are any technical influences because of that. But here at CaseKing we still decided to offer delidded 9900K CPUs..."
Thanks for confirming. I'll make a bit of research over what this barrier could be made of. Maybe with some chemical reaction we could make a new one after grinding the CPU.
Edit: according to this article (overclocking.guide/the-truth-about-cpu-soldering/ ) it's made out of titanium and nickel-vanadium. I guess you would have to somehow plate titanium on the ground surface then nickel-vanadium on top of that. Except titanium seems to be exceptionnaly hard to plate.
Edit2: the titanium and nickel-vanadium are apparently only for soldered CPUs, so I'm guessing the non-soldered CPUs have another type of diffusion barrier .. probably something simpler like some sort of passivation reaction. Silicon dioxide seems to be the most common silicon passivation technique except silicon dioxide is an insulating material .. I'm scourging wikipedia for high thermal conductivity silicon derivatives.
What a boring life that would be .. it's just so much more interesting to learn about all this technical stuff and try to push the limits of what we can do. And the Ryzen is maybe cheap but miles behind Intel in terms of efficiency and optimization.
Anyway in the meantime I found an awesome website (www.eesemi.com , great read) which suggests that the protective layer is silicon nitride (www.eesemi.com/glassivation.htm ). But the different processes used to apply silicon nitride seem really insane, one involves plasma and vacuum, the other hellish temperatures. And in the end, silicon nitride has a pretty bad thermal conductivity, so if you deposit too much you risk loosing any benefit from the grinding. I suspect there is no real solution here apart from using conventional thermal paste that does not attack the silicon.
I thought that was epoxy, not silicon...
@@nout4269 "Ryzen might be cheap, but its miles behind intel on efficiency and optimization" - This proves that you are a intel fanboy... Firstly, i know that intel's chips are faster for gaming, just wana make that clear, but Zen is a new architecture, it's only been out for 2 years, of course AMD optimization isn't gonna compare to to Intel's now 3 generation old architecture. Secondly, do i have to remember you that Ryzen is on 12 nm, instead of 14++++++++++? Ryzen is not FX. The difference in power consumption is at best 5-10W between these chips, as could be seen in this graph : goo.gl/images/hHHmS4
Please grind down the chip as much as possible while recording the changes in temps in increasingly small steps until the metal layers. Thanks for the great work!
and you give him the money for that experiment?
We do by watching the vids...
Caseking is his sponsor I think. They have quite a bit of CPUs. :D
"We do by watching the vids" ... - rofl his yt revenue is if it comes close 200€. Nopp addrevenue ain't do the job. If anything it would be his patreons.
"Caseking is his sponsor I think. They have quite a bit of CPUs. :D" - Sure they have and for public relation they will help him out. But they have a budget on how much 8auer can burn through,...
@@Airwave2k2 yeah but if no one watched the vids then they would have no sponsors
*Intel is going to catch SO MUCH HELL this weekend.*
*And rightfully so.*
well the new chips are running hotter than hell
@@henrylau1643 lmao
king jeng lau Burn burn it till it's nothing but glass ahahahhahaaa
And leave AMD alone and complacent in the CPU market so we get another Bulldozer Piledriver fiasko?
Sorry, no...intel may be dethroned, but we do NOT want intel to die, no matter how bad they are, it would be worse if they leave.
@@Chrinik How are they even dethroned?? They currently have the fastest CPU out there in the 9900k.
In gaming they have the second the 8700k.
Hardly dethroned...Infact...very far from it by the looks of it.
I
I'm surprised intel doesn't thin the wafers at the factory... I was an engineer at an HP water fab in the early 2000's and all our chips underwent a thinning process back then to get the best thermal and electrical properties....I still have a load of the diamond 'paper' and compounds... Better keep hold of them for when 9900k gets cheap enough to play with
Thats quite interesting, good idea!
The SEMI standard for 300mm wafer thickness is approx 785 microns. So the work indicates that Intel is using the native wafer thickness, which really means that the prior nodes were backside ground to something thinner, which means it costs more. Could be as simple as competing on cost while providing sufficient cooling for their product specs..overclockers be damned...
So the wccf and videocardz rumors that 9900k runs at 5.3 with an air cooler was totally
scam?
Arous instruction for overcloking - please use water cooling and you can get 5.0 ghz with 1.35v ( in 80% cases if you don't have luck well )
I dont expect the regular chip to out perform the review samples
@wwwelkam I was thinking the exact same thing.
When you send out samples for review, you send the best of product lines, not the average samples or bad one.
Reviewers get binned chips.
Great, now you have delid soldered CPUs, and lap the silicon down, just to get acceptable temperatures?
What do you think you pay the extra for? Intel is giving you a change to start a DYI project which you can share with everyone. (Attention! Your results may vary!)
so i have to buy brand new CPU cost 800€ to delid Before using it
Happy halloween from Intel
If you consider 84 degrees "acceptable"…
sirius4k intel selling Pride and Accomplishment like EA
Intel goes full EA (never go full EA!) : The intent is to provide buyers with a sense of pride and accomplishment
Waiting for 3700x :)
If AMD doesn't screw up in some stupid way, they will completely destroy Intel with zen 2 :)
Diablo 4 confirmed
majnu100
Unlikely, as DDR5 is expected to come out in 2020 or 2021, which is part of the reason AMD says they will support AM4 until 2020 (other reason possibly being PCIE 4)
even if DDR5 came out early, AMD likely wouldn't be able to use it without breaking AM4 support
Yep. I was considering getting a 9900k because some of the software I use for work doesn't run well on amd hardware. until I saw the temperature and power tests and this video.
Zen++ or zen2
When I was in college I was told that sanding using figure eight motion will cause the flat surface to eventually become convex. The procedure I was taught was to sand in straight lines five times in one direction then rotate ninety degrees and repeat giving equal strokes to each rotation.
On other videos from der8auer and on my experience of lapping the IHS only of my delidded 3770k few years back, It was shown that both the Naked Die silicon and the Top of the IHS are actually not 100% flat.
Heres a der8auer video : ruclips.net/video/tnd2LO0IBic/видео.html
at 4:25 showing that the Die is not 100% flat.
at 8:06 of the same video shows that the middle part of the die is convex (its not flat from the get go).
and then IHS can be Concave or Convex.
In my experience, when I lapped the nickel plating on my 3770K IHS on a flat glass table and tried my best doing so. It showed that the IHS on my 3770K was concave right on the center.
Whilst I saw some AMD chips had a Convex middle of the IHS right on top of the silicon die. (FX8350)
yeah figure of eight isn't good for when you really do need that perfect flatness although it doesn't matter too much in this case as long as you're getting a smooth surface
figure 8 might give parallelism, whereas your method probably gives excellent flatness?
So the moral is if you want a good 8 core cpu with good thermal, buy a Ryzen 2700
yes
Of course not, why would anyone want a fake 8/16 CPU if intels 6/12 betas in gaming and 8/8 [9700k] beats it in everything, be it software, zip, games, HEVC, etc.
9900K Rapes it
@@NoBodysGamer why fake 8/16 ?
@Wellington Campos, what do you call a 8c/16t CPU that loses to 8c/8t CPU if not fake?
What do you call a 8c/16t CPU that also loses in many things to 6c/12t CPU?
Do you know ZEN architecture? Each 8 core ZEN is in reality Two 4 core CPUs "glued together" [they use Interconnect], thats why RAM speed has such Great effect on Ryzen benchmarks but not on Intel and thats why AMD CPUs cost less money but Intels CPUs cost more, because they more complicated to build.
Do you have Ryzen 2700x? In its control panel it has Game Mode, this game mode, disbaled One SSX [one CPU] and only leaves 4 Cores working, its done because many older games see Ryzen PC as system with 2 CPUs and they not optimized for it.
Wow your level of stupidity reaches new low levels.
I'm tired of this shit... I'm switching to Zen 2 next year and it will be my first AMD bought product 😒
why is that sad? AMD is a great company...
good choice
Welcome to the fold :-)
I've actually been looking for a good AMD cpu and I hope zen 2 has really good performance
@@zg2111 I'm really happy with my 1700x, but can't wait to upgrade to the r7 3700 when it's released. Amd also bundles a free optain competitor in to their second gen and later ryzen boards, so it's gonna be so worth it.
One thing that the video is missing - is comparing the chip thickness to 2700/2700X, for example, and compare Intel 9600/9700/9900K solution's to AMD solution's.
the 9900k and the 9600k seems to have the same Die and Die Size, the 9600K is basically a 9900K with 2 disabled cores and HT Disabled while the 9700K is the 9900K with the Hyper Threading Disabled.
Man love the attention to detail, would have been heaps of work to do all this testing, appreciate it! 👍
So 9900k is basically an FX 9590 from a temperature standpoint.
FX would thermal shutdown well before this. I believe max temp on the spec sheet was like 69c, but I'm pretty sure you could set thermal shutdown limits at 79c if I remember AMD Overdrive correctly. I don't have to worry about it much though, mine rarely ever peaks beyond 0c at 1.7 vcore on a small phase cooler, only a few synthetic loads can make it do that.
@@jakegarrett8109 The temperature sensors were also off by a few degrees.
@@PimpMatt0 Yeah, it would show 10c when ambient was slightly higher (so impossible on air/water). FX is only accurate at higher temps (non-linear thermocouple)
I think its still below 0c though, my block temp peaks at -25c after a really long load on synthetic loads (-35c in Firestrike and Timespy, it needs to use more power than that and for a much longer time, like a few minutes). Idle is -58c on the block (I didn't drill into my CPU to put a temp probe, so the only way for me to get temps is from the probe in my cooler, and from software for core temps, but software won't show below 0c, I just know it runs cold).
At least FX was honest about it’s thermals with 220w TDP in the specs
_Exactly_ ...
Only thing is, the FX 9570 as well as the FX 9370's *EST* is precisely at *57°C* (sic!). On _everything_ above and beyond this so-called _Emergency Shutdown-Temperature_ the cpu itself will electrically emergency cut-out itself physically. Then there's a *CLICK and the system is down immediately.
… and as you can figure out already, that makes all those stories about Bulldozers reaching like +100°C pure nonsense and bullshit (straight-out lies to hamper the bulldozer's reputation, FUD?) - as all the ESTs of FX Bulldozers are either 57°C, 61.10°C, 61°C, 70.50°C or 71°C respectively (depending on model).
Maybe check i7 2600k’s (or any Sandy Bridge, I guess) die thickness as a reference, since that was the last line of CPUs (mainstream, excluding other extreme CPUs) to be soldered to the IHS. That could be a starting point. Then compare that to say a i7 3770k, when they started to used thermal paste instead. Maybe that’ll give us an idea as to why it has to be so thick.
Also, I love what you’re doing as far as research goes when looking into these sort of things. Some may say it’s pointless, but I find it very intriguing and informative on the manufacturing process as a whole. Hope you never stop. Great info thus far. 👍
This is exactly where I thought of too. Please do this!
My theory is that it is because intel made the 9th gen using 450mm wafers, and the 8th gen was on 300mm.
According good old wiki, bigger wafers have to be thicker to have the mechanical strength to support its own weight, so that it does not crack during handling.
Handle Different.
at the end of the fabrication process the wafers are ground down .
Aren't we all ground down at the end? I know just how those poor wafers feel.
I’m pretty sure 450mm steppers don’t exist. They were cancelled a few years back.
@@ellsworth1956 they're ground down and then wafer dicing separates out each individual die. The wafer needs to be mechanically stable in between these two processes. Scaling of bending stresses is nonlinear, and I can well believe that 50% greater wafer diameter requires double the *ground* wafer (and die) thickness.
I can't believe that Intel even managed to mess up the soldering ... honestly :(
"WINTER IS COMING".. oh wait i have a 9900K cold worries no more :)
bing jr I saw this comment on HUB’s video today too... heh.. you really want those likes, don’t you? x)
That argument is not fair!
As it was on AMD in the past...
Lol
It doesnt entirely make sense tho. Better thermal conductivity could mean lower cpu temps and same overall heat output, or it would allow you to push the chip harder increasing output. You want the heat out of the cpu, not trapped inside!
@Simon T - Exactly... makes one wonder what kind of TDP measurements they have at Intel... For cooling the IHS or the cores...
Oh, if you really want to not worry about winter, build a dual Pentium Pro or some Pentium 4 Prescott systems.
Intels fx 9xxx series! :D hahaha
This is actually very funny. The intel would turn green with rage if they heard you compare the 9900k to a fx 93xx. But, they are both 8 core space heaters running at 5ghz. Personally, i think amd is killing it with ryzen, and the path ahead seems to be amd. However, jim keller works for intel now, so rip core, rip ryzen. They had a good run, but jim k will make intel billions
@@Sageofthe16 You do realize Jim Keller was behind the Zen architecture as well right?
The new Intel 9590! When you really need an oven and CPU in one!
@@TheRickity yes, exactly. Whatever Jim learned from ryzen, he can use to make chipzilla for intel
This only proves that AMD is better companies that Intel because Intel had tons of money at there's disposable, and had to buy AMD personal to fix their mess, first was raja koduri and now Jim kepler... So AMD has better engineers but Intel being Intel and trying to cripple AMD by buying employee from AMD.
conspiracy: chip is thick so intel can brag about improvements in the future. "Look! We cut degrees by 20*C! *picks up Novideo book* Don't delay, buy 10900k today!"
Half of these fanboys would pay ott money for them too.
Thank you for all the good information, Der8auer. Without the expertise of reviewers like you, there would be no oversight. Your work is greatly appreciated.
Intel is the new furnace
It's ok. Winter is coming
Lol, everyone will be calling for toothpaste TIM again so it'll be easier to apply liquid metal on their delid.
I'll be asking for delidded CPU's out of the factory.
Intel finally investing in solder for a single generation, just so they can half ass it and say "I told you TIM was better" and have an excuse for going back to toothpaste and therefore cutting costs next generation
but that is risky for intel because 9900K is considered a milestone chip before they shift to 10nm or 7nm architecture. plus 14nm chips are low in stock aren't they?
The thicker wafers likely cost Intel as much if not more than the soldering process since that means fewer wafers per ingot. That's even worse considering that there is an on-going ingot/wafer shortage. If the main reason for the thicker die is indeed to take up the thermal cycle strain from solder TIM, then the opportunity cost of soldering really isn't worth it from a business standpoint, better off getting ~30% more wafers per ingot with a thinner die and paste.
intel used to solder stuff, they started using paste in ivy bridge, My i7 2600 is still a marvelous and cold beast
Der8auer keeps getting to delid! Winning!
Der8auer had a mission, and he pretty much killed it!
awesome video man. i'm really diggin this channel.
I wonder what's the Sandy Bridge 2600K die thickness?
Lagittaja this should be discussed more. If this is exactly the same thickness as the sandy bridge die or any of the X99 CPU dies, then idk what the big fuss is. If it's thicker than the previous soldered dies, then I would start wondering why Intel did what they did.
I am still running my 2600k from april 2011 at 4.8 GHz with an voltage of 1.350 V (+0.125 V in OFFSET), and it only hits 76 C on its hottest core, no delidding just attached a normal big air cooler (Scythe Mugen). I was hoping we would get back to something like here 7½ years later...
It ran 5.0 Ghz with an voltage of 1.425 V (+0.200 V in OFFSET), but turned it down to 4.8GHz around 5 years ago.
Running my 2600k at 4.4 Ghz 1.33 V and it peaks at 60 C
Running my i5-4670 @ 899.597 MHz and it peaks at +29.0°C
Your English and narration have gotten much better, I would say it is now excellent. Thank you for the hard work and effort.
This is an old marketing trick. They sell you more material for more, the size is linear, the profit is exponential.
Maybe it is because the soldering process. Thicker die so that the risk of damaging the CPU from the heat application when automating the soldering process is lower? Could be a tolerance issue or a manufacturering impact as well.
12:18 "Why the hell is this chip so damn THICC"
CatheteriZedEYE because intel don’t want you to overclock too much
Sabotaging thier own chip
I'm thinking the added thickness might have to do with the soldering. Previously the problem was thermal cycling causing spalling of the silicon, which is why Ivy Bridge's die shrink didn't allow for soldering, at least according to what I've read from researchers. Maybe the added thickness prevents this?
So I guess we're at 14nm+1mm+++++
Or it could be another con by Intel that they will magically bring out a new version of the CPU with better performance as a "tock" version of the 9900K and the only thing the will have done is made it a bit thinner?
wait this was basically a 10th gen prediction thats how they cooled 10900k
Great questions der8auer. Learned a lot here. Thank you for spending so much time testing and modding this CPU.
so intel is trying to replace AMD as spaceheaters?
they already did
After the appearance of the 7980XE, there are no heating competitors left.
gonna have to turn the radiators next winter when i buy amd perhaps lol
I never got that AMD meme. I have an OC 8 core FX and it's one of the coolest CPUs I have ever owned since my Celeron 766Mhz.
I run it on passive water cooling with temps never getting over mid 45s on full load. My GTX 980 pumps out far more heat with a full water block, making up about 2/3rd of the total system heat output.
The hottest CPU I personally ever worked with is i7 975 XE, custom loop at 4GHz it ran around 75c water temp on dual 120 x 50mm rad. with 2 x 3500 or so rpm delta fans. That thing was hot!
Just in time for Winter :D
You should also consider making a "jig" to hold your cpu or ihs perfectly flat and perpendicular to your sandpaper while your grinding. I saw that you hand hold it and rub it on the sand paper. Even if your paper is on a 100% flat surface, it's still impossible to hold it 100% flat with your fingers. The grind could be at a slight angle, even if that angle is in fact perfectly flat.
I highly recommend this.
how thick is the ryzen cpu's compared to the 9900K? seeing as they are both soldered
Indeed! I really want to know this.
Prettt much exactly what I thought would happen as soon as I heard the 9900k was going to be soldered. Indium based solder does not conduct as well as gallium based LM and these were never going to do as well as a delidded 8th gen so no surprises there.
Intel were between a rock and a hard place with this one. Probably needed to solder it to keep it cool enough for general usage by non-enthusiasts. Unfortunately that means delidding for LM is a lot more tricky for the small number of customers comparatively Ho wish to delid.
However it’s very interesting to see the results after grinding! I hadn’t considered that the chip would be thicker.
This is from Indium Corp on their website:
Compensate for a mismatch of the thermal expansion between two mating surfaces. Different materials expand and contract at different rates, creating a need for some compensation in any intermediate layer (like a thermal interface material). The other candidates are not as forgiving as indium. "Indium is really, really soft. Its flow stress is is about 150psi (1MPa). This allows indium to conform to the mating surfaces and to absorb deformation due to differential thermal expansion without transferring the shear forces to the delicate semiconductor materials," says Bob.
It will also optimize thermal conductivity by its increased contact even on imperfect surfaces. Bob adds, "Indium conforms to an asperity, bump, or dip in the interface surface where a harder material would simply lose contact around such obstacles. Indium maximizes the contact surface area."
thermal conductivity of conductonaught is about the same as indium solder. The issue is layer thickness. When you apply liquid metal, you essentially paint it with a negligibly thin layer. The indium solder layer is 0.5 mm.
HOW DID RYZEN GET IT RIGHT 2 YEARS AGO? Intel engineers are getting so lazy or the business environment is so bad that they have no chance to be competitive in the future
I'd be curious to see a comparison of die thicknesses between Ryzen and 9th gen chips, and thermal/overclocking results with both as the silicon gets sanded/lapped thinner and thinner until the point of failure
Their room for error is much larger because they aren't really trying to push high clock speeds. Their chips can't keep up with Intel's for high-refresh gaming, though.
AMD got it right even 7 years ago. The FX Series (That are bad) for example.
@@seanmcc09 I would actually say that AMD are trying to push high clock speeds, but just don't have the hardware. The overclocking headroom on their X sku chips is practically non-existent; single cores in my 1700X spike to 1.45V stock.
Jim Keller.
I must really say that you took it a step further then the other reviews which is much appreciated because it explains a lot.
that one dislike is from intel ceo
The funny thing is yes the Fx Series were space heaters but they maxed out at 80 degrees and they would do that all day long I want to know how long it will take until they degrade the silicon at measured 95 degrees
Someone isn't gettin a Christmas card from Intel this year.
Will the direct die bracket work if you sand down the die? That would seem to yield the coolest temps(except for exotic cooling).
I'm running an out of the box 9900k with an AIO and it gets pretty warm. I've considered de-lidding but it seems the real problem with this chip is the die thickness...
Are any of the soldered AMD chips this thick as well or is this just Intel being solder n00bs?
It's funny because intel are the ones who came up with this soldering technique and have the patent on it. AMD has to pay intel a licencing fee to use it :P
actually i think AMD hold a patent for soldered ihs , but im not sure
maybe amd has different soldering technique that actually works good
der8auer said in another vid that it's an intel invention.
10:00 WHAT?????? No! Transistors spikes are sunk toward top side of silicone! Close to the side where heat-sink is. Closer to the PCB the bigger metal layers, closer to heat-sink smaller details. Pattern looks like arrowhead one next to another with spikes pointed toward heat-sink. Regardless of how thick initially wafer was process of etching silicone remains the same, and thickness of active layer remains the same for both Coffee lake and 9 series. Digging deeper and deeper from PCB side the smaller patterns (holes) are made then admixtures applied, metal layers, oxidization etc.. To be able soldering heat-sink and to to prevent top of die crack destroying transistors stretched up by iridium, also cause die is longer initially they cut thicker wafers using more material, so in result top side is little bit thicker than usual and there is also additional metal layer on the top that makes it even thicker. In result more distance between transistor heat-sink.
Im curious how 9 series dimentions compares to Ryzen's silicon die cause AMD got even larger dies, also soldered.
Awesome video, when i saw what was inside the cpu it looked like the worst soldering job i have ever seen.
that's pretty much what any solder chip looks like delided
Damn it... So to get the best performance now we have to add removing solder and lapping to the delid, a lot more can go wrong now 😡
7:57 Maybe i missed it somehow but whats the cooling solution used in these benchmark results?
NH-D15
I'll be returning my 9900K 10 minutes after delivery on Monday. No thanks.
Yea you should wait for reviews. Hopefully the return goes well for you.
Forgot I ordered from Amazon and already canceled it.
Just buy used 7700/8700K and you won't be needing anything better for games, nor will you get any performance hit on a desktop environment (like browsing, watching videos) from Ryzen.
@@n.i.5381 just wait for zen 2.
@@janchovanec8624 honestly you dont notice a difference with 4k between ryzen and Intel. I'm guesaing if you can afford a 9900k your playing at 2k or 4k. So ryzen still is the better bet.
wow! now that’s a really comprehensive analysis! well done!
It's like the P3 all over again.
Think P4 had it worst (actually still have one :D).
Pentium D?
@@corwinblack4072 if memory serves the P3 was the first one to be that hot. And while the P4 topped it it was still the P3 heat shock that stuck with me.
Nah the P3 was fine , it was the P4 that was a joke. There is a reason they based Core off P3 not P4 .
You are thinking of northwood pentium 4 that was bad and Prescott p4 that was the original nuclear reactor
I will be performing my first delidding and lapping of my 9900K die next
week. What I would like to know is does the bottom of the IHS (the
bottom that actually touches and glues to the PCB) need to be sanded
down equal to the amount that is taken off of the die so that the lid
touches the die? I have not seen this done nor discussed. Thx for your
time, knowledge & help! You are greatly appreciated!
Joey
I’m doing this right now... just lowered die 0.53mm mirror finish. Now I have to lower ihs same amount. It’s coming out great can’t wait to see where temps end up...
@@hermanquinones Awesome! Did you cut the die first and then sand it down? Is .53mm safe? I was going to do .40mm. Please let me know how it turns out! Very interested! Much success! :)
Thx!
Joey
No cutting at all... bought the supplies for delid and solder removal from RockItCool.
Sanded starting at 240 grit and worked up to 7k for mirror finish. Spent majority of time at lower grit 240 & 400 to remove initial .30mm. Flitz metal polish was finishing touch. But total amount removed was 0.53 mm.
@Joey
Safe? To be honest we are in uncharted territory. That said I believe the height being close to stock 8700k should be fine. Also, don’t be surprised if surface cracks are exposed on initial sanding they go away when proceeding.
@@hermanquinones
I'm looking forward to seeing your before and after temps... and Cinebench R15 scores.
Right now my best is 2274 at 5.2 GHz and 1.470V. I should be able to delid next week and hoping to see some great improvements :-)
I can Boot It 5.3 but it is not stable when i running cinebench most of the time. It is stable running my Photoshop speed test.
oh my god this fail CPU cost 750 € in Poland today
great video @derBauer, i learned a whole lot there. very comprehensible given the complex subject matter. subbing for sure!
"Why the hell is this chip so damn thiccc?"
So much Detail! You Clearly deserve more than that 120k subs.Thanks Bro! love your content!
Really happy to hear this :) Thanks
It's almost like Intel intentionally adds defects then later "corrects" them to suggest performance improvements.
Why don't you make CPUs if you think you can do so much better.
@Paul...such a childish response. Grow up.
+SaintQ perhaps you've misunderstood my intent? Unfortunately I find people like you tedious so I won't waste my precious time trying to explain it to you. I do suggest you invest your own time attempting to figure it out though. Good luck.
@@1pcfred What intent? The kind of intent that undermines logic? Wyatt was not right or wrong by imagining future scenarios, that is not a bad thing as if said scenario does happen he can make that point... that is how humanity gets smarter by thinking. :D
Fetcher.
It's common knowledge that CPU manufacturers intentionally hobble chips in various ways to differentiate pricing, for example, disabling cores and lanes in microcode or whatever. It's not too far-fetched to imagine they might also intentionally create a heat issue that limits performance, such as inadequate TIM, too-thick silicon, etc., which is what Der8auer's video made me think of.
Thermal conductivity is a material dependent property. It does not depend on the thickness of the material. I think that you meant to say that as the thickness of the silicon increases the heat transfer rate increases.
I'm a happy R7 1700 user running at 60C :D
Yep, even with so small of an AIO you can see my CPU sticking out all around AIO, my Threadripper 1950X runs about 65c at ~4 Ghz (cooled by a 120mm AIO, gen 1 Asetech pump made for DELL XPS before Alienware existed, so its a very, very old cooler). I'd say 3.9 Ghz is about right for 65c (just so the fan is quiet since I sleep close to it) running 4x Handbrake encodes at once in a batch run.
I'm a happy i5-4670 user running at +24.0°C
My Pentium 2 core G4560 running max at 64: tootpaste and stock cooler forever lol
Why not buy 2 hand Dell Optiplex 7010? you get i5 3470 4gb ram and a monitor for like 150$ refurbished.
Bcause i just need a 7700 later/soon (Asus z270 board, 16GB 3200MHz, RX 570 right now)
Question, please, to anyone that knows about this stuff. I'm half way through the process. I've removed the IHS and scraped the solder off of the inside of the IHS and the CPU die. That part was way easier than I thought it would be. So now I'm about to sand down the die, but I can't find the sandpaper he's using in this video. Where can I find it? Or can I just use regular 3M sandpaper? Also, what grit/micron is the 2nd sandpaper that he's using?
Ryan Steffer you found what grits were used? What sand paper?
no reason to upgrade my delidded 7700k for that. Im going AMD next time
Maybe by the time you can achieve that much tangible benefit Intel will have their shit figured out.
Or they won't, i'll be even happier to be honest :D
I was thinking of replacing my 6700k running at 4.9Ghz for Ryzen 2700x, however more cores wouldn't give me any benefits for games and desktop activities because of its low clocks, so I'll in stead wait for another generation.
im on a delidded 7700k, 5ghz, looking to upgrade but theres nothing worth upgrading to
you can try hard modding your motherboard to support 8th gen cpus
bad idea for hard modding. power delivery system wont cope with it.
Interesting points der8auer. Thanks for measuring the thicknesses.
Another scam, nvida scam, intel scam. 5.0ghz 300 W 1.5 v very good and long life for sure for that kind of overcloking.
they calculated that so that u have to buy their next Generation + another board 😂
why is it so thick? a lot of people are reporting cracked dies when they delid. a few of them have sanded them down to find that the cracks disappear very near the surface. so it could be that the tension from soldering and then the different materials cooling at different rates is causing forces to crack the die and the top layer is there to strengthen it and maybe to act as a sacrificial layer.
This ain't good, so will a NH-15 Noctua cool this thing even at stock? Cus that's what I ordered lol. I'm wondering if they needed the extra thickness for the solder process, to dissipate the heat so it don't degrade the chip.
Get a 2700x
I feel Intel kinda boned the pooch on this one. By refusing to price it competitively they are fighting a two front battle. On one side the 2700x which is cheaper and slightly loses out to it for gaming. But on the other side with the price of the chips and mobos puts it in thread ripper land... Would be neat to measure die thickness of and verse Intel parts.
@Jesus McBeth no problem 80 /90 Celsius o more all time,
That Noctua is a beast, outperforming or on par with most AIO water coolers on the market. You shouldn't have any issues.
Now that reviews finally came out, looks like stock you're right, shouldnt' be a problem around 80C (that's what my 3570k Runs currently when playing GTA).
Awesome explanation. Keep on delivering these informative vids!
You look like jimmy neutron
😎
10:55 your making circles in an 8 just like in your name
*you're
@@entswaggles *kid
I'd rather take the risk and not put the lid back on - it's what we did back in the days and used shims to prevent cracking. Worked great
The community pushed for solder but their implementation could be flawed since they haven't done it for so long?
They solder Xeons, its just the regular customers that get the garbage... Servers must not be made for longevity, am I right?
Really nice video der8auer. Informative and entertaining lol! that Risky STIM Delid + die lapping !
And if that thickness is really affecting CPU Thermals.
I guess Its okay to say as an example that the 8600K is a way better choice than that 9600K (8 core with 2 disabled cores) on the intel side atleast
I feel like, not only the 9600K doesn't have much of an edge against an Overclocked 8600K.
The 8600K easily goes to 5GHz. Even better chances of hitting that 5GHz if its delidded. And its cheaper brand new vs a 9600K, and even cheaper if bought 2nd hand. And has a more petite Die that does not have any Overheating issue.
But Damn, really nice video man! proper entertainment :D
First the PT's meme benchmarks, and now this Thicc Silicon.
The good +++ The Bad
#1 Intel Finally made an 8 core 16 thread to go against AMD's 8 core 16 Threads +++ But its only marginally fast and its VERY expensive
#2 Intel Finally Used S-STIM +++ But it still runs hot because of the Thiccer Die
and quoted from someone saying, perhaps the die was made thiccer to have better yields from the manufacture process.
This is why 3rd party reviewers and people like you der8auer are the Good guys. It is not fun buying a Product at pre-order not knowing what it entails. Specially if the product cost a lot i.e. 9900K for 719 Euro. I am happy with my used and delidded 8600K and 8700K XD
Seriously though man, You and this videos of yours are Epic! You deserve more!
Most tech tubers only focus on benchmarks to be honest.
Not many goes in and Delid a soldered CPU and resolder them again then re-delid them. But also freaking Scrape the S-Tim on the Die with a blade and freaking Sand/lap it down! That is crazy! but it is very entertaining to watch and to see how much it affected temperatures.
Seriously, Good Stuff! :D
I was expecting that the 9900k with solder would be able to do 5.0GHz on all cores with a big air cooler and stay under 80 C on the hottest core, but to reach that is seems you need buy a pretested golden delidded lapped CPU with liquid metal applied, WTF......
My 2600k from April 2011 is still running at 4.8 GHz with an voltage of 1.350 V (+0.125 V in OFFSET), and it only hits 76 C on its hottest core, no delidding just attached a normal big air cooler (50$ Scythe Mugen). It ran 5.0 Ghz with an voltage of 1.425 V (+0.200 V in OFFSET), but didn't like that the hottest core would reach 85 C, so turned it down to 4.8GHz around 5 years ago.
Looks like my 2600k is going to last yet another generation, but to be honest at 1440p ultra it still seems fine with a GTX 1080 Ti, matches my friends 1800X in games anyway.
@Martin - I'm still running a 2500k @ 4.6GHz as well because CPU progress has stagnated and replacing half of my system for a small performance boost hasn't been worth it yet.
@Salt Maker That's £350 for the CPU and motherboard for a roughly 35% performance boost. Not worth it still when my 2500k can still handle 9/10 games and £350 on a GPU gets you a 100% performance boost.
@Salt Maker 24GB of 1833MHz. And at 1440p the 2500k can keep up with a 1070 which is what you'd get for £350. But I'm just gonna block you know because you're a prick :)
When sanding the die, are you still able to use the direct-die frame / oc-frame? Caseking states that the frame itself is only 0.1mm lower than the actual die. So if you sand the die more than 0.1mm the cooler / coldplate won't touch the die. In this video the die is sanded 0.20mm and then the oc-frame shouldn't be compatible any longer. But the video doesn't say if the frame was being used or not after this. Or if the IHS was mounted back and the standard socket was used. If using the IHS after sanding down the die, you will probably have to sand the outer edges of the IHS 0.2mm also, to make contact between the underside of the IHS and the newly sanded die.
If using the die without any frame at all, and simply mounting the cooler directly (as in actual direct-die cooling) there might be contact issues on some of the pins along the edge of the socket since you are only pressing down directly above the die.
k so get a 9900k, delid it, sand it down by .3mm and then run at 5.5ghz. thanks man
0.3mm - u got bOOls i wanna 2 see it... 0.32mm is the cover layer so .... be carful....
better -0.2 and dont use ihs just nickelplated colling on it
Actually Hardcore Overclocking & Der8auer - LOVE you
Have you contacted Intel about this issue?
What do you think they're going to do about a product that's about to release any moment now? Cancel it? Recall? lol
@@sirius4k No, but it might be funny to hear whatever cocksure reply Intel gives this time since they've been caught bsing for the 3rd time in a month.
They won't say officially shit, but you can be sure, he has contacts to them - he has also mentioned that he had reached out to his contacts, but nothing specific was replied
Well.. one answer from many possible would be: "Listen here fuckers, you wanted it, stop complaining."
Intel: "What issue? This is a feature."
I think that the die lapping might be even better than you describe, since you didn't mention (unless I wasn't paying close enough attention) that lapping is actually removing a lot of silicon nitride, which has a WAY worse thermal conductivity than pure silicon (4-5 times worse, to be more specific). It doesn't change the end result from the actual tests, but I suppose it helps to better explain WHY they went down as much as they did.
Well 8 degrees delta is worse than AMD.
There are research papers out there regarding the silicon fracturing under thermal cycling. Your hypothesis about end-user pressure to solder and then Intel following up with the solution requiring a thicker die due to mechanical stress is right one point.
Ryzen is soldered and it doesn't have these kinds of problems.
It doesn't reach 5GHz on air or water either.
Which does not matter at all. The "5 GHz" argument is pointless as frequency is only one of many factors that influence single-core performance.
Thank you for putting up English videos. It's a lot of work for you to put up two languages but i really appreciate it.
so i have to buy brand new CPU cost 800€ to delid Before using it
Happy halloween from Intel
They gave you a trick instead of a treat.
Realy good work from u guys , the content that u provide is realy unique and proffesional. just subbed! keep up the good work.
Here's my theory : Intel sucks and overcharges their crappy cpu's, buy AMD.
I would if their single core speeds didn't suck. That's what you need for gaming.
As a chemical engineer I think I would look at a better cooling solution, as multi-layered conduction heat exchange is usually dependent on the external temperature of the final layer i.e. air or water in pc cooling. Fresh air would be cooler than recycled water, giving higher delta T's. But the internal design of pc's and noise could limit what is possible. The area that heat transfer occurs across is also a potential problem as many people exclude it from their calculations. If one layer is smaller than the others it could restrict the performance of the entire system.
SORRY INTEL WE MESSED UP! PLEASE TAKE THE SOLDER BACK 🤣🤣
Keep in mind that no testing was done with the poor intel thermal compound. This could be a case where with the stock compound, intel simply couldn't get under 100C with any of today's popular coolers.
I think it's great that you took the time and did this
It is a very interesting result and needs to be investigated further for modders
Thank you and well played
Doing the PC lord's work
der8auer XD
That's one Ｔ Ｈ Ｉ Ｃ Ｃ Ｂ Ｏ Ｙ Ｅ．
Very interesting. Thank you for all of this hard work!
With solder there is no phase change within operational range, and the solder layer is aready _much_ thicker than LM, Cryonaut or KPx would be, so its mass should accomodate the solder to take on the stresses of thermal contraction, rather than passing (most of) these on to the silicon or the solder-silicon bond. As long as there's no phasechange, no fractures, the solder should be fine all the way down.
I love your in depth testing!
Holy shit! Outstanding work, as always, from Roman.