• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

13900KS Delid

Solaris17

Super Dainty Moderator
Staff member
Joined
Aug 16, 2005
Messages
27,892 (3.84/day)
Location
Alabama
System Name RogueOne
Processor Xeon W9-3495x
Motherboard ASUS w790E Sage SE
Cooling SilverStone XE360-4677
Memory 128gb Gskill Zeta R5 DDR5 RDIMMs
Video Card(s) MSI SUPRIM Liquid 5090
Storage 1x 2TB WD SN850X | 2x 8TB GAMMIX S70
Display(s) 49" Philips Evnia OLED (49M2C8900)
Case Thermaltake Core P3 Pro Snow
Audio Device(s) Moondrop S8's on Schitt Gunnr
Power Supply Seasonic Prime TX-1600
Mouse Razer Viper mini signature edition (mercury white)
Keyboard Wooting 80 HE White, Gateron Jades
VR HMD Quest 3
Software Windows 11 Pro Workstation
Benchmark Scores I dont have time for that.
I wasn't happy with my 13900K temps and bought a delid kit w/ IHS copper upgrade. However knowing the KS was going to come out I figured I would just wait and held onto it.

My 13900KS came in today, and as this was going to be a CPU swap I decided that instead of installing it twice, we were just going to roll the dice and open the box and just proceed immediately. Total time of ownership before beginning? about 15min.

I do things.

Anyway lets get to it.

Rockitcool kit
Thermal Grizzley Conductonaut
Starbond
Heatgun

Thats about it.

Diving right in, the processor itself pulled out of its box.
IMG_2861.JPG

We can't get too attached so to the garage we go. We will need to set the CPU down and prep our workstation. I took my super charger out yesterday, so we don't have a lot of room. That's ok though prep is most of the work and the lack of it is why most people fail.

IMG_2862.JPG

First we need to get our heatgun plugged in and ready, so I grabbed the old wagner, mine was covered with hydraulic fluid, but yours doesnt have to be.

IMG_2865.JPG

Next we need to find a place to hold the chip. I dont have any surfaces that aren't covered in engine parts right now, so bench vice it is. We will do it "arrow" to the bottom left. This way I have easy tool access to the adjustment bolt and because we need to make sure the IHS slides off the correct side or we will destroy the caps :)

IMG_E2863.JPG

Now lets test fit the CPU so we can logically double check and visually confirm we are pushing the IHS off in the correct direction. We will also mock setup the top cover to make sure the push mechanism is aligned properly so we dont need to waste time on bolt turns. We dont need to go super fast, but the clock will be running.

IMG_2866.JPG

/shrug seems legit. Now that we have test fit everything we will put our wrenches and bolts to the side in preparation. Then its time to get started. We are gonna keep this only a few inches away, and only for a few min. I set my gun to "high" but I dont have a temp range, I just warm things up a lot.

Now the tool is a polymer, so it can take more of a beating then those melty KB keys you have because you let your cig fall into it. We do still need to be careful though.

IMG_E2867.JPG

Now that the chip has third degree burns its time to get some of our own, remember Upham you gotta be johnny on the spot with the screws, take your time but be DILIGENT the clock is against you, but the thermal soak will keep it liquid for a little.

Once the top is on, Get to cranking, this is the important part, this is a CPU not your GF she might be satisfied with 20seconds but with this we need to break the epoxy. Our enemy is the solder if it hardens, but right now its like mayonnaise, the thing that is fighting us is the epoxy/rtv type material they use. This does NOT melt, it just gets soft but it is like cement. So when we are cranking down we will encounter HEAVY resistance, but keep CONSTANT SLOW pressure. Thats the trick, dont treat it like a ratchet and turn it 1/4 turn in .03 nano seconds. We want to start and slowly apply one smooth slow turn with even pressure.

You will hear some cracks and pings. If it sounds like gravel congrats you turned your core back into sand. If it sounds like your grandmas mercury sable cooling down in the walmart parking lot keep going. Eventually you will hear a plop type sound and the IHS will freely move. If you took a swig to stop the shakes before you did this, it will look like this.

IMG_2868.JPG

Remove it for a surprise

IMG_2869.JPG


Yay and the IHS


IMG_2870.JPG

Nice! So now that you probably heave a dead CPU its time to take the half a chopstick they gave you and use physics to rub it off, I'm sure you neckbeards will nail this part.

Once thats done we need to surround the core with tape. It doesnt matter what you use, we will clean everything after. The kit comes with what rockitcool calls "quick silver" now since "quick silver" is technically mercury and that isnt what they are giving you I can only assume this some kind of gallium compound. Now I know judging by the thermal threads here most of you may have failed high school earth science, but basically we are taking a higher concentration than what you would find in say conductonaut and using it to dissolve the indium solder used on the stock IHS. Some of this has cooled onto the silicon core. So we will use their included abrasive Q-tips and smear this all over the die and rub it off.

IMG_2875.JPG

Now this has some nasty effects, so like above we are going to protect the rest of the CPU by taking it off. Use the tip of the chopstick they gave you to apply some gentle pressure and a crease around the core. This will help prevent any leaks under it. We do NOT want this to touch anything else on the CPU.

Now we cant and wont use this for thermal paste, so use the while damn syringe. I chose not to do my IHS because well, I have a replacement and I dont care. Once it is on, you need to agitate it, come on I wont do your homework for you. Agitating it will increase the speed at which this dissolves, and that will be apparent by the film that soon forms on the top of the gallium we add. In this pic you can see all of this. You can see the film, you can see I covered it in the stuff, and you can see me smearing some away, showing you that it is dissolved.

Now the applicator is abrasive, so if you have a DEX stat higher than 5 this will feel like a light grit sand paper to you, but you need to differentiate between solder still on the core and the feeling of the applicator itself. We need ALL OF IT OFF or you are going to destroy it :), but its alright. Its probably already broken.

IMG_2879.JPG

Alright so you have managed to get all the indium off. Sweet. Now wipe it all away. At this point we are going to clean and re-tape the CPU because now we must polish the die with the included cream. You ever waxed a car or done body work? Then you know. Dollop that shit on a towel and start rubbing. Apply some pressure, but this is not a sand. The chemical reacts to the residue on the core and you will see black tarnish immediately. Keep going. I did it twice. Half the pack rubbed it until it was dry black goo then wiped it clean and did it again with the last of it.

IMG_2890.JPG

Now that your done polishing, the die should look like a mirror. Congrats your ugly. Now we need to apply the conductonaut on the new IHS and the die.

IMG_2895.JPG


Wow that looks like a lot! Yeah it isnt, but its more than you would have put on, which is too little.

IMG_2899.JPG

After you get the LM applied to the IHS prepare your tools once again and grab the pressure plate included. We need to have all of this ready because as seen in the photo we will be applying some glue. You can use any kind of super glue you want. I use CA glue. This is used for model kits mostly and delicate work. Its very runny by nature but does come in grades and phases like Gel, or heavy med light. Anyway, this shit sets QUICK so be ready.

Just dont use elmers glue, you only eat that. Have some after you done, you did a good job. Dont use RTV either. That worked on some of my older CPU mods but the clearences are too small now. It doesnt matter where you apply it. I did the winglets but you can do the corners. Just dont go wild.

Anyway, once you put a little on the IHS its time to set it and align it in the tool. This will let the IHS sit how it would from factory geometry wise. Just be careful about the SMDs! Remember, just like the heating part. patient, but brisk. Do not rush.

IMG_2900.JPG

Once its in place bolt on your clamp assembly. Remember this is all one go. You are aligning and clamping all in sequence. Do not stop.

IMG_2901.JPG

Now the big bolt looks mighty scary but its plastic. You dont even use the tool. You just hand tighten it until its firm and let it sit. I would do it for 5-15min. It was cold so I decided 15 even though its a quick bond glue and went and got my rig ready for the transplant.

Anyway now that 15min have gone by your ready to find out you killed it. Remember dont RMA, that makes you a scumbag and raises costs. You did this.

IMG_2906.JPG



Woot it booted! Congrats your marriage is still in shambles. Watch your temps, but if you took your time everything should be fine.


Q/A

Q: Why did you do this?? AMD and intel and the KS :froth-at-mouth:
A: Because I wanted to. TPU is not about fortnight FPS or arguing over AMD vs Intel. TPU was and is first and foremost a tech enthusiast forum and I did this because I wanted too.

Q: Why LM? I heard from %member% in %thread% that LM is for non-religious people and will hurt your computer.
A: wack all of my PCs and delids still work and run cool.

Q: You replaced a K with a KS why?
A: You already have a few threads where people eat crayons and bitch about just this. This thread isnt that.



My temps would hit 90 all the time in game on my K. I would reach 100ºC during heavy workloads causing throttle. I imagine by no stretch my KS would do the same.

Under the same work load I now max at 77ºC.

It seems counter intuitive, but this was actually written this way on purpose. The forums over time have shifted and dont include projects like this as much anymore from what I have seen. In many threads people that do this are generally ripped to shreds by die hards that think this kind of thing is the end of the world. Writing it in a devil may care fashion was part of the whole presentation. My other guides are alot more astute and formal.

Thanks for reading!
 

Attachments

  • IMG_E2867.JPG
    IMG_E2867.JPG
    3.1 MB · Views: 542
Last edited:
Congrats mate nice work! I would love to do this to a dirty KS!

How did you get the indium solder off? Just work at it with some light sanding paper?

I've done a few delids but none with solder YET..

Cheers.
 
Congrats mate nice work! I would love to do this to a dirty KS!

How did you get the indium solder off? Just work at it with some light sanding paper?

I've done a few delids but none with solder YET..

Cheers.

Yeah I mentioned it a bit. They give you a LM compound that you put on the die after you get the IHS off. They give you whats basically a cottom swap but its more gritty. You start rubbing and the indium disintegrates.

Always fun to delid stuff. What oc are you going for with the KS?

Hm not sure yet. I am modest in my expectations or try to be, and will start at around 5.6-5.7 all core and try from there. This thermal density is wack though so who knows what I can maintain ya know?
 
Great stuff man!
 
Post your Cinebench R15 and R20 scores!

I dont know what they would have been before I havent ran cinebench in years. It also wouldn't be comparable. This is my main machine not a score runner. I drive 3 4k displays and have all my shit set to autostart. I dont install any third party apps and tear apart the core OS.

I stopped doing all of that garbage awhile ago. I run 3dmark to baseline myself and tests for stability, but I choose not to compete with anyone.
 
Great write-up! Congrats on not screwing this up :D
 
I have a 13900KS sample that came in today as well, ill be more focused on investigating silicon quality compared to the 13900K I have from when that launched.
I wanna see how well these power-optimize in a restricted TDP
 
this is fucking epic


liquid metal needs to become standardized at the factory level, like PS5 did. those are some insane gains on temps.
 
My temps would hit 90 all the time in game on my K. I would reach 100ºC during heavy workloads causing throttle. I imagine by no stretch my KS would do the same.

Under the same work load I now max at 77ºC.
Was the K model a delid also?
 
Just wondering why you went with copper ihs instead of direct die?
 
Seems like 5.9 GHz all core with around 1.35 V is doable, 6 might need 1.4+

 
Seems like 5.9 GHz all core with around 1.35 V is doable, 6 might need 1.4+


That's only if you do static OC w/ all core ratio and no power limit at 1.4V - which is a suboptimal way to OC these chips. This guy is great at ram OC but hasn't played with dynamic OC's enough IMO.

At gaming load you can probably get away with 6.2Ghz all core fairly easily on his sample.
1674064315618.png


I have a really crap binned 13700KF that can't hit 5.7Ghz without dumb volts, but I can keep the chip and these clocks/volts since games don't pull much wattage even at max boost --- I just have it throttle down to stay at 230W for cinebenchy type stuff where it stays in at a reasonable temp.

With his KS sample he could be well above 6 ghz as long as he didn't insisnt on a static all core with a locked 24/7 voltage and no power limit. He would get higher FPS in all of his games at the cost of a few % in synthetic CPU loads, which, for a gaming rig is a great trade.

Either way at about 5.7/5.8 ghz you're mostly memory bottlenecked anyways, so it's kind of a moot point. They need to stack some HBM on these to really unlock these cores.
 
Also, what would the effect of not cleaning all of the indium off of the die?
You melted it and slid the IHS off. It’s no longer a flat smooth surface like it was when originally manufactured. You have little mountains now. If you attempted to mount anything at this point you would crack the die.
 
You melted it and slid the IHS off. It’s no longer a flat smooth surface like it was when originally manufactured. You have little mountains now. If you attempted to mount anything at this point you would crack the die.
Ah - I didn't realize the clearance was so low. I had figured that the LM layer would be thicker.
 
Great write-up! Congrats on not screwing this up :D
I was relieved to not have destroyed it hahah

Was the K model a delid also?
It was not

Just wondering why you went with copper ihs instead of direct die?
Ripping my water cooler out is literally more work than delidding. I'm not breaking work records and I didnt want to modify my mounts or my block for direct die. This is the most this chip will see.
 
Last edited:
Isn't IHS made of copper already?

It is. Its copper w/ nickel coating. The after market one has more surface area and is machined more true. This is what helps with the temps. While surface area makes the most benefit, the stock one would see some gains being lapped.

TBH the rockitcool copper IHS could also use a lapping, but lapping isnt the geometry that is getting the gains here.
 
Delid always useful. (Unless of course you kill the processor :D)
Especially on high-end processors. I've never tried it, but I would like to if I could.
 
Fun read, nice work, enjoy!
 
Any concern about keeping the die surface level when you are polishing it? If the question is ignorant, forgive me. Saw a recent Gamer'sNexus video where the EVGA 4090 bare die is lapped with extreme care to ensure level surface.
 
Would've been interested for you to run this chip before doing the delid to know what the temps were beforehand, so you could know how much you'd dropped with the delid. But that would've required you to swap the CPU out twice instead of once, and as a lazy person myself I absolutely get not caring to do that.

In many threads people that do this are generally ripped to shreds by die hards that think this kind of thing is the end of the world.
I only rip apart the people who (a) delid/apply LM without understanding or proper preparation and f**k it up (b) expect that delidding/LM will magically make a power-hungry component consume less power (c) expect that delidding/LM will magically fix the fact that the power-hungry component in question fundamentally lacks enough cooling and no amount of improving said component's ability to dissipate heat will fix that (this is especially true for the idiots who buy laptops crammed to the gills with high-end hardware then try to fix the throttling with LM). Not to mention those who do this who f**k it up and then try to RMA - as you say, scumbags.

On the completely opposite end of the spectrum, you're doing this for the sole reason that you want to and that you can, you've taken full responsibility for what happens if it goes wrong, you've had fun doing it, you've had fun sharing the experience with us, and I've enjoyed reading it. So thank you for posting a thread that's interesting and entertaining, as opposed to yet another "liquid metal doesn't work" or "liquid metal destroyed my GPU" or similar sob story of stupidity.
 
Last edited:
Any concern about keeping the die surface level when you are polishing it? If the question is ignorant, forgive me. Saw a recent Gamer'sNexus video where the EVGA 4090 bare die is lapped with extreme care to ensure level surface.
None; because polishing and lapping in this context are not the same.

I am not removing material from the die, the polish is used to lift the residue of LM and dissolved solder. It’s just abrasive enough to get small particles off that the cotton swab missed.
 
Last edited:
Back
Top