Lapping a CPU, the proper way [AMD 3950x]
Vložit
- čas přidán 21. 07. 2020
- This video explains the process by which I created the flattest CPU in the world, bringing a Ryzen 3950x to within 0.3 microns of flatness error, two orders of magnitude more accurate than what is possible with sandpaper.
The impact on cooling performance from lapping a CPU this flat will be investigated by an independent computer hardware reviewer and those results will be linked here when available.
Many thanks to Tom Lipton for the inspiration and advice he provided. If precision machining is something that interests you, check him out: • Precision Lapping 101
For an in depth look at the history of precision, I recommend watching Machine Thinking: • Origins of Precision
The music used is Quirky Bones by Borislav Slalov
• Divinity Original Sin ...
If you want to encourage me to waste more time on this nonsense, you can do so here:
/ penrowe
If you think what I publish is awful and you'd like to pay someone to flatten me in a hydraulic press, your deposit will be accepted here:
BTC: bc1qmm49ze4qrt2vmg54t43tpdc6y6ttw4fxaaz5nf
XMR: 4ABGGhG4kMyG863A2w4FcUSU9pRuHMNCoawSsUCDBkWv97vsP9xTQriHyaXRckjKrtB2LB8p5mt6SNc7gS6qfuL9Krkvir3
Came here after seeing the article on PC Gamer. Very excited to see what comes out of this.
samesies
Were you disappointed?
I'd like to thank PC Gamer and its readership for all the positive attention. I understand there's also a bit of frustration that my video ends on a cliffhanger, I assure you that I share your desire for BAR GRAPHS but while I could go out and spend two days holed up in my workshop and publish those juicy numbers for you I did after all sell my CPU and its matching water block to a hardware news and review publication with the implicit understanding that the story was part of the sale. I'm not going to go out and spoil the scoop for a friend.
was a bit confused by the intro (or lack thereof). It felt like someone had just ripped der8auer and reposted. But then the "real video" starts. I look forward to learning more!
Same
Can't wait to see temps without thermal paste. It's going to be a good day.
For the love of god...
...please start making more videos.
Your attention to detail and narration voice would make you an instant sensation on CZcams IMO...
PC related, machining related, doesn't matter.
I still have my regular job to keep up with and spending time with my family often takes priority when I'm home so my writing/recording/editing pipeline is always full.
That's a freaking thing a beauty right there! Bout time someone did a true lapping job on a CPU! Can't wait to see the results. Even if the temp difference isn't massive, I'd still be happy knowing that incredible lapping job has been done and that the CPU and cooling block can be rung together without paste!
Right?
Liquid metal might even be overthrown...
For all we know...
If this works totally gunna pay him to do a proper lapping job for me.
@@alanfidurski6860Honeywell PTC already beats in most practical applications. This is neat to see, but you don't actually want it perfectly flat. You want a slight curve, concave IIRC, due to thermal expansion, so it's flatter at load.
dude you have a great voice, very nice to listen to.
@X Oh well, can't please everyone.
Paul over at PC Gamer seems to agree:
"He partially chronicles the process in a 10-minute video, detailing various steps in an extremely soothing voice. Seriously, I could listen to him read a phone book (remember those?)."
Source: www.pcgamer.com/a-machinist-lapped-an-amd-ryzen-16-core-cpu-within-a-nanometer-of-its-life/
Great voice. Here, have my subscription.
I mean...damn...I mean dudes voice got me feeling like a baby ready to sleep out of comfort. Yeah he can ready me a couple bedtime stories! 👍🏿
@@Penrowe More suited for a murder story than tricks for overclocking a PC.
Was that "Beve Sturke" I heard, Steve Burke's alter ego? Gamers Nexus!
I am not at liberty to disclose such information.
Yes, thank you, I knew I was right!!! That's definitely Gamers Nexus!
@@XeoNIvan When?! When do we get to see part 2?
9:59 I thought I heard that too.
As soon as he mentioned bar graphs....
Wow! So smooth that the Casimir effect is even observed :O
I've worked in Mc shops for almost a decade and worked as a QC tech. It always killed me when people would "lap" with sandpaper. It's like milling with a lathe. Sure you can get a flat surface, but it's not the same thing! I've never seen an IHS rung with a coldplate and I have to admit I'm pretty excited to see what Beve can show with that. It's so counter intuitive that the lap plate needs to be softer than your work piece, but this has inspired me to learn more about this and makes sense when you think of the charging process, but it's still a little "wait, wut" if you are used to traditional machining where the cutting tool is harder and the cause of material removal. But if you think of the lap plate as a tool holder, and the tool removing material is the abrasive, it becomes a little more intuitive. Great video, thank you for making this!
Biggest lie you told: "At the end of the day I'm JUST a machinist..." To the contrary, this is great work and your presentation very understated. I'm so glad you're investigating this phenomenon. I've asked other PC-Performance CZcamsrs about the effect of wringing a CPU IHS and a cooling block for heat transfer performance, but all I've ever gotten was 'crickets'.
I just subscribed. Can't wait to check out your other videos.
Very interesting! I really enjoyed the backround, never thought about precision flatness this way. Looking forward to the next part.
Tuned in for a brief but interesting cpu modification, got info on the history and science of flatness. Loved it, awesome video
It's been 8 months. I'm still waiting for the part 2 of this video
Sorry dude something happened, and we won't hear why.
Eagerly awaiting those graphs, very fascinating work here! 👍🏻
Great video, concept, quality and audio. wish you more success with such content. await the next one in this series
Absolutely fascinating. Eager to see what the results come back as.
I was silly - I used sandpaper on glass when I should have used my milling machine and cast iron lapping plates.
Oh well... next time, maybe :)
I can't be the only one who got a braingasm from the Three Plate Method
The basis of all precision.
You wouldn't believe the rabbit hole I had to go into to find that out before youtube machinists/engineering channels started talking about it.
best comment ever...I had the exact same experience
definitely not! one of those 20/20 hindsight situations.
Oh man! Can't wait for the follow-up! EPIC video! GOOD JOB!
That was weirdly fascinating.
Please tell me you're making a follow-up to this.
I can't promise when but there will absolutely be a follow up to this video.
@@Penrowe 7 months later...very curious to see the results considering I had great success years ago but seems like all of CZcams says it doesn't work
Dude, love video, love quality, love your editing, love your voice, and love how you, like others mechanics or electronic engineer can explain stuff that in school it would probably take month to learn xD
Fun project, looking forward to see their results.
Great stuff! Very interested in the follow-up video :)
I knew about gauge blocks, but damn, that 3 plate method blew my mind.
I saw this video like a week after it was posted and have been anxiously awaiting the results the entire time. The wait is killing me
I too am feeling a little restless, been sitting on results from my own testing for a while that I'm looking forward to publish.
@@Penrowe so when should we expect anything? I am eating more than normal just because of waiting in anxiety from the result if that makes any sense. lool
This was in my recommended. Thank you, youtube.
This is some true expert and engineering stuff bruhh keep it up 🔥👍
Never heard of this before in my life, tanks for the informative video!
I did this to my P4 and Heatsink back in the day.
I learnt the hard way that they would become inseparable, ripping the CPU from the socket. When I went to remove the heatsink.
I ripped my A8-3800 out of the socket just from old paste being in the system (6 years not being turned on or changed). Good thing I only was taking it out to try lapping
I'm still waiting.. the hanger broke, but I fell on the floor safely, but still waiting.
Getting serious Ahoy vibes from this video. Great video. Subbed
This is absolutely incredible and very interesting. Great video!
Thanks, glad you liked it
I had to watch a second time because your voice is too soothing it almost put me asleep.
Finally, a true lapping vid. It has always bothered me watching people sand dies with obvious edge rounding. I'm looking forward to seeing a follow up. I've always wanted to see somebody toss theirs on a surface grinder too.
"Hello, this is Beve Sturke." Sounds like we're about to get some killer content on this!
@Nolo Kobo It's Steve Burke's alter ego he used to use to make orders without being recognized by the PC manufacturers. If you like this kind of content, you should check out the GamersNexus CZcams Channel!
yeeeeeeeeessssss
@Sugarz maybe next time answer the guy's question in the first reply instead of being snarky about someone who doesn't know what you know.
ok this was great. I said lapping and sanding are different but u know big channels and answering viewers is not a big thing unless it's a mass of ppl. Thank you for this, now I'm off to sand the IHS on my 5900x...Good luck everybody!
Right, where is part 2! You've left me hanging bro! 😁😂
Amazing video!
Gosh, this makes me really want to go the extra mile and try no TIM, so I’m a machinist and have done lapping for 15+ yrs, also XOC overclocker, with my water blocks and cpu dies, Direct die 9900k/10900k, I ground the die flat to 2 light bands then lapped it lightly from there, along with water block cold plate, but with water block assembled to avoid the mistake and make everything square to itself, die to PCB, ect. On 9900k chips, took my best 6 I also tested different material removal to get closer to the transistors which did make a significant improvement, and let’s just say the results with both surfaces at 1 light band flat they will wring together, but getting them to wring while assembling cpu/waterblock in motherboard would be an extremely challenging process. Non the less on my golden 9900k I have insane core to core deltas of +/- 1c controlled coolant temp (chiller) and on my golden 10900k +/- 2c using liquid metal. And my temps compared to stock saw significant decrease of around 10c to 25c depending on chips and silicon quality. My 10900k was just going to have IHS lapped but I had core to core delta issues even after lapping IHS, so a delid was needed.
Only thing stopping me is time, look forward to seeing another video.
I'm excited for the next part
Penrowe, this is incredible work and I'm so excited to see you make follow up videos about lapping processors and heatsinks. About 10 years ago I had a CPU cooler and when I lapped it with sandpaper and polishing compound I lowered the temperature of my CPU by 3-5* C (my testing methodology was extremely primitive so take this with a pinch of salt), since then I've always had a weird fascination with it. I've read quite a bit about lapping over the past 10 years but I think this is the first time that someone with your background and experience is tackling this subject. Please, please, please continue making videos and thank you for your time and effort
Can you give any advice for people who are trying to lap their CPU, heatsink, or waterblock? Most people will not go to the lengths that you have gone to in order to get a perfectly flat surface, but are there any tips you can give to get closer to a flat surface with a relatively small cost? Most of the advice I have read in the past focuses on using a flat piece of glass, using different grades of wet/dry sandpaper, and sometimes it suggests using polishing compound to get a mirror finish. Is this correct, or how should it be changed?
So, there are no even absolute results for this experiment after all of this? Man, you can't have us hanging like this :'(
Man, you've got one hell of a good voice-over voice 👍🏻
So? What are the results?
This is uncanny.
I was talking with someone about launching this exact business the other day.
I too believe in this process, I believe it to be the best and most effective way.
PC Gamer sent me here. But I subscribed because I must know more about this mystery man, Beve Sturke. I wonder if he means old Steve Burke? I don't know anyone named Beve, but old Steve lives out beyond the Dune Sea.
Can't wait to see the follow up from Gamers Nexus! I better get a detailed notification! Lol. Great video!
what's the update on this? Is there an ETA on any future experiments?
Hey man, I hope I haven't missed anything here and I don't sound rude or offend you. Just your channel is really small, but this is a damn good video. I didn't know any of the stuff you talked about in the video but you made it so interesting and I've definitely learned a few things. If you made more content like this I think your channel could really grow!
Klockren! vilken underskattad kanal. Vi behöver en revanchvideo
Wow! I learned a whole bunch of stuff like the counterintuitive bits about how hard the lapping plates should be.
@Penrowe
Hi i am back, any progress?
Tech Ingredients and Linus Tech Tips did a paste video review.
So i am here again to ask if you had the new tvideo ready for us?
Can't wait for the next video... hopefully?
This is the real deal. 2020, and too few people actually uses the resources available to improve stuff.
Great!
Any updates on that last Beve Sturke comment?
Dear @penrowe, where to find one of those lapping plates for copper!?
I didn't know what CPU lapping was 15 minutes ago and now I'm all for it.
I AM SOOOO FUCKING EXCITED! You blew my mind when you discussed the history of lapping and how it was invented. I couldn't believe how simple and elegant the solution to creating a perfectly flat surface was. Now I await the second video and I believe Gamers Nexus will do this test justice.
Är det van der waal-effekten som får de plana ytorna att klibba ihop?
Cool concept, awesome video, then.. nothing.. Are there any updates to this? The result of these efforts deserves to be shared!
great content tbh, such a interesting and widely misunderstood topic
So if I wanted to get my cpu flat relative to just my heatsink, I could just lap them together? What abrasive(s) would I need to do this?
You know you play to much Divinity: Original Sin 2 when you can recognize its music right away in a video.
Also knew I heard this immediately
Does anyone know what is the status on this one, I'm still curious.
Awesome stuff.
I am however curious in regard to increasing the surface area, by using a rough paper on the flat surface. (much like the rifling on the bottom of cooking pans helps.) Any ideas of how that could be accomplished?
if you want to rough up your surfaces, get a fine abrasive surface of your choice, and create your minor grooves in a specific direction. should take just a few seconds to make the grooves you're thinking of. make the grooves in the same orientation on the surface it will be mated to.
then fill in the imperfections you just added with Paste and wonder if it was worth bothering.
Mmmm, dat voice dough. Lap me up, sweet siren. Professor Rick would approve of your attempt at true flatness.
So by how many degrees of core temperature has this process lowered?
still waiting for the update, come on man this is super interesting.
So is there any new news about this? Wish to know the results but found nothing searching on google, reddit, youtube or any other searching plataform
Finally a proper lapping video. It's very annoying when people just sand their cpu and say they have lapped it.
I have used a set a grade 0 blocks. They wring together very well and I have always wondered what would happen if you wring the cpu and cooler together.
Personally I think once the cpu heats up the thermal expansion will completely ruin all the efforts, they will separate as the ihs and cooler expand at different rates.
I'm not sure the deformation you speak of would be that bad. A temperature difference of 50°c will cause a 1mm thick copper sheet to deform by 1 micron. For one that's still within the scale of wringability and actual real world temperature deltas between the edges of the IHS/cold plate are unlikely to be anywhere near that bad. Either way it'll become obvious if it's a problem.
@@Penrowe it would be across in x and y that might be the problem. The amd ihs is 37mm square so in x and y that's 37 microns. The base of the cooler is a different thickness to the ihs as well as a different width. The water temp will be 40c max, and the cpu will be running at 80c
Will be interesting to see though. I look forward to the results.
Content of this quality but 400 subs? Very very nice, can’t wait for the follow up. Though I’m sure we will be lol, takes a lot of time and effort
I only can find silicon carbide powder can you link where I can get the silicon carbide you used
I use autosol but I'm sure other brands work fine too (mother's for example). You could also look for valve lapping paste from any auto parts store since those almost always include silicon carbide as the abrasive.
it would be interesting to see the effect of mounting pressure on thermal performance under wringing
probably minimal given how hard it is to pull apart gauge blocks that have been wrung together. You have to slide them off of each other.
did you ever use the plates you made without the aluminum foil, or did you only use the plate with aluminum foil for all of the passes, because you did say that the material for the lapping plate has to be softer than the material your lapping with which is copper, and what was the grit of the silicon carbide because I what to do it myself like what did you start with?
The silicon carbide was ~3 microns, I used regular chrome polish from the hardware store, same stuff I use for my knives. I did try at first to use the bare plates but quickly discovered what a poor idea it was, could never get a test cpu done with a better result than maybe 15 microns of error across the surface after an hour or two of trying. Lots of fall-off and high risk of fouling.
Mounting the heat sink directly on the silicon is the most thermally efficient way to cool it. Direct die cooling. The inside of the IHS doesn't even make contact with the top of the silicon! Manufacturers use either solder or thermal paste to transfer the heat from the silicon to the IHS, and then you use thermal paste again to transfer the heat from the IHS to the heat sink. The trick is making sure you don't crack the silicon!
How flat is the die? The etching process is so ridiculously precise that I expect the die to be super flat too. And BTW, many CPUs in the early 2000s didn't have an IHS. You put the heat sink directly onto the die. And yes, lots of chipped dies, shops has signs explicitly telling they don't accept returns for chipped dies. Probably the reason why modern CPUs have IHS.
I do have some older CPU's I could attempt to delid to measure their flatness and check for wringability but even if they're not flat already it's something I should be able to solve. The only issue is time.
3D stacked chiplets from AMD will eventually kill direct die cooling as the height of the individual chips will be of varying height.
this remember me when i was doing my own telescope mirror !
good job man !@
Instant subscribe! Just have to know what happens next...
Was the video supposed to be cut off abruptly?
Does this mean that the "home lapping" technique can be easily improved using the 3 surface method? And would that still work best if used an abrasive or would it work with sandpaper as a shortcut using glass to lap CPU & Heatsink, then the CPU on the Heatsink and the with the opposite direction of Heatsink on the CPU?
Theoretically you could lap just the CPU and heatsink to each other and achieve two perfectly matched surfaces (neither will be flat though). There are several challenges in this, first both are of equal hardness so you'll never have a perfectly smooth result, you also need a relatively soft abrasive to prevent grit from embedding and causing deep scratches, without slotting there's also roll-off that will occur to the edges.
In short, it can be done but it's hardly worth the effort.
This video is so underrated!
Sooo, 6 months later, where is that follow up? 🤔
is this still on its way to gamers nexus? or did it go some where else
Awesome 🔥
But what if there are no V O I D S ?
Man this got deep.
Or did it?
This is amazing!! Would you please consider making a video showing how you made the lapping plates using the 3 plate method and cutting the channels into them etc? Also if you lap the CPU and thermal conductor together how does that work with the channels and such?
Check the links in the description, oxtoolco's videos on hand lapping are several hours long in total and give you all the information you'd want. The only thing I did different was milling the groove pattern.
Härlig video :D Ser fram emot testerna från Steve
*Beve
Would the wrung surfaces not separate with the changes in temperature?
Maybe but the heat output of a CPU die is not enough to meaningfully deform its IHS.
Great video! Looking forward to the follow up, take my sub! Såg nog PostNord, antar att du är från Norden? 👍
Was there ever a follow-up to this? Would definitely still be good content even if the results were disappointing
DUDE. Can you please tell us where to find the follow up. I can't find it anywhere and the hardware review publications only mention this video but no testing or follow up. Can you please direct us to where to find the follow up?
I can't speak for GN but I've been busy working on projects that put food on the table. Sorry I haven't been able to finish this thing but it just doesn't pay a regular salary. I hope to close this out once I have the time and money to spare on my less serious projects.
@@Penrowe Thanks man. I understand your situation and respect that your life is not lapping CPU's! Thanks for the highly informative video, I found it super interesting. Your style is really good.
thank you
The question of running a wrung processor/heatsink pair without paste has been kicking around in my mind for a bit.
Perhaps this would be easier with an air cooler, even if it's only a single sample with a sample before and after test.
Interested on seeing how this works out with the flattened IHS and paste though.
I suspect it would be difficult, as the two may unwring when the processor heats up.
So... Does perfect level really screw with your head?
This video makes me wanna become a machinist
also, dat voice tho
this is quality content.
I have done my fair share of metallographic cross-sections for electron microscopy. Final polishing of cross sections is done with on soft polishing clothes with a colloidal polishing compound. This cloth is on a rotating table. A small amount of compound is used charge the cloth. A small stream of water goes on the table as you polish. It works very well. It's not really for flattening however. The surface with flat be to itself, but not to a particular plane.
But I do not think a mirror finish is desirable. It will never prefect and there will be slight air gap at the interface. If the air is removed, heat transfer across such voids will be EM radiation (radiant heat transfer), which is not every efficient at low temperatures. You need something to fill this gap.
However having the CPU and heatsink flat as possible with maximize the contact area.
@Penrowe
You seen the Hardware Canucks video?
"Can this MYSTERY Metal Replace Thermal Paste?"
May 7th 2021
Indium Sheets are about the same as top paste in filling the gap and transferring heat.
Indium is often used in solder and is the primary alloying compound used alongside gallium in liquid metal products. It's hardly a mystery metal, certainly not to anyone familiar with the manufacturing of modern computer technology.
I NEED MORE.
Great video. Next time please just lower the music by 5db or so, you have a great voice. Let us listen to it even more 😁
Fair point
Any updates?
The suspense. Will you be releasing the video of the results or will I need to subscribe to another Chanel to see them?
Gamers nexus channel is were the results will be when testing had concluded.
@@DizzTheDoc thanks, already subscribed.
What a fucking good video. I'm locked on it.