Overclock.net banner

1 - 20 of 29 Posts

·
Registered
Joined
·
95 Posts
Discussion Starter #1
So I decided to get adventurous and take some sandpaper to my CPU and AIO cold plate because I've never been happy with my temps. Started with 240 grit wetordry, finished with 2000 grit, on a piece of glass for flatness. The cold plate was remarkably even (check picture, no hotspots) and didn't need lapping. As you can see if the picture, the CPU IHS was convex and I took some material off. At first I thought the overheating was because water got into the small hole in the IHS and hadn't dried completely and was shorting something, but now I'm not sure. Here is the order of events so far:

- Lap CPU and cold plate

- Top off AIO with distilled water and cleaned out the gunk build-up in the coldplate's fins.

- Resocket and repaste with IC Diamond

- Replace GTX 1080 with RTX 3060 Ti

- Turn on computer and update Nvidia drivers first thing (was 2 updates behind)

- Fans start ramping to max towards end of update

- Computer starts shuddering and crashes within 15 seconds of stuttering starting

- Try turning back on, system shuts off within 2 seconds of pressing power button.

- Wait a few min, try again, POST displays CPU overtemp warning. Shut down and wait 20 min to cool down more.

- Turn on and go into BIOS. CPU temp starts at 70⁰C and steadily climbs to 85⁰C in 2 min. Shut down for the night.

- Next day, same behavior

- Take CPU out, clean off and put distilled water into the IHS hole. Used hair dryer to dry it out.

- Next I put isopropyl alcohol into the hole, and used the hair dryer again.

- Resocketed, repasted and changed back to my GTX 1080.

- Temps in BIOS are now stable at 33-35⁰C, figured the problem was solved at this point because before this it would overheat even in BIOS.

- Boot into windows, CPU hits 90⁰C within 3 minutes, except it doesn't crash. Downclocks to 400MHz and everything is slow as hell, but doesn't crash. I shut down after being on for 10 min.

So this is where I'm stuck. I don't know why it was overheating in OS and BIOS at first, but now after trying to displace any water that may have been in the hole, it only overheats in the OS. I've of course reset the BIOS to defaults as well.

Anyone have any ideas or suggestions? already have another 5820k order from ebay but I'd like to explore what is going on with this
 

Attachments

·
Registered
Joined
·
378 Posts
CPU lapping this is unneeded in our days, flat surfaces are now warranted due the use of latest technology.
Another mistake that we did this is CPU lapping when CPU this is cold (room temperature), because in the end of the day, you expect flat surface when the CPU this is at 45C or higher.
Metals they do expand when be heated.
AS5 thermal paste was capable to fill any gap and any mistake at lapping, the modern thermal pastes are not that capable.
 

·
Registered
Joined
·
233 Posts
As a mechanical engineer/master fabricator, if you didn't lap it using a mill/cnc or calibrated surface grinder you will have issues. Or even use something like a calibrated surface plate and some super fine emery paper. If you didn't do one of those things, you will easily develop leaks in the mating surfaces. Kiraikos is correct in the thoughts that in today's tech, lapping really isn't needed unless there is a serious flaw in the mating surfaces. Fluid/thermal dynamic properties hold as well in the expansion/contraction of metals. Especially something as heat absorbent as copper. Even a micron sized flaw could mean a potential leak depending on location/temp/fluid pressure.
 

·
Graphics Junkie
Joined
·
2,814 Posts
Clearly did not do a good job on the CPU. You should not lap the heatsink.

Need a flat edge, glass works well. Also need very fine grain. Sand paper.
 

·
Registered
Joined
·
233 Posts
Use some layout die to coat the ENTIRE surface you intend to lap. Then use UltraMega's idea of a piece of glass as the main surface. Get some fine emery paper and of course water. If you are doing it by hand, it's the best option you'll have. The layout die will show you how consistent the surface is as you go.
 

·
Registered
Joined
·
455 Posts
So in reading it sounds like you refilled the AIO which means you took it apart. My bet is you have an air bubble and need to bleed the loop and add more coolant. Try and put on the stock air cooler or put an aftermarket air cooler on to isolate the problem. AFAIK you should never open an AIO. To me it 100% sounds like a failed AIO and id bet putting a stock heatsink on it will get you going.

Also, you need to keep sanding, that looks way too rough still but that shouldn't cause over heating like you describe.
 

·
AMD OC'ing Enthusiast
Joined
·
3,053 Posts
You dont need anything special. Have a sharpie marker? Make an X corner to corner, then do a border about 1/4 inch from edge around cpu. Now do a cross in the center going to edges.


Now lap the cpu with 1500 grit for a few seconds, and look at the pattern. See if it attacked only one section, or if everything has a scuff now.


I've lapped cpu's and heatsinks, I find just resting your hand the wrong way upsets the pressure, so just be careful
 

·
Iconoclast
Joined
·
30,654 Posts
At first I thought the overheating was because water got into the small hole in the IHS and hadn't dried completely and was shorting something, but now I'm not sure.
Don't cover/plug the IHS vent, just wash the CPU thoroughly after lapping it and let it drain and dry completely.

My money is on something being wrong with the cooler, not the CPU.

CPU lapping this is unneeded in our days, flat surfaces are now warranted due the use of latest technology.
I have never owned a CPU that I'd consider flat out-of-the box or that didn't benefit from lapping. Some are better than others, but I can always improve on the stock flatness.

Even objective measurements modern CPUs reveal variations several times that of recommended TIM bondlines. Igor's Lab looked at some recent CPU IHS not long ago.

As a mechanical engineer/master fabricator, if you didn't lap it using a mill/cnc or calibrated surface grinder you will have issues.
Don't need anything fancier than a flat surface (a dollar sheet of float glass) and some decent sandpaper to significantly improve upon surface flatness of a CPU.

We don't need micron tolerances when the IHS is likely to deviate by ~100 microns in flatness out of the box.

The points about thermal expansion are relevant, as is the fact that Intel's retention mechanism warps the whole IHS and package when clamped down. These can be hard to account for; I actually use a salvaged socket from a dead board as a CPU holder/lapping aid and heat the parts to check flatness at different temperatures.
 

·
Premium Member
Joined
·
5,971 Posts
I second using a piece of glass. I just try to apply even pressure when I am sanding, use water and a little dish soap to help with friction, and make sure to rotate the CPU in your appliance to reduce the chance of grinding it unevenly, as well as use different motions like side-to-side, circles, ect. You don't need your lapping to be perfect, because all you really need is pretty even contact between the IHS and the cooler. With that said using a fine grit sand paper at the end can help to flatten out any uneven surfaces. Sometimes I'll draw on the IHS with a sharpie and sand until all the sharpie is gone with 2000 grit.
 

·
Registered
Joined
·
233 Posts
I have never owned a CPU that I'd consider flat out-of-the box or that didn't benefit from lapping. Some are better than others, but I can always improve on the stock flatness.

Even objective measurements modern CPUs reveal variations several times that of recommended TIM bondlines. Igor's Lab looked at some recent CPU IHS not long ago.



Don't need anything fancier than a flat surface (a dollar sheet of float glass) and some decent sandpaper to significantly improve upon surface flatness of a CPU.

We don't need micron tolerances when the IHS is likely to deviate by ~100 microns in flatness out of the box.

The points about thermal expansion are relevant, as is the fact that Intel's retention mechanism warps the whole IHS and package when clamped down. These can be hard to account for; I actually use a salvaged socket from a dead board as a CPU holder/lapping aid and heat the parts to check flatness at different temperatures.
[/QUOTE]

Didn't mean the surface out of the box is measure in microns. Just that enough heat and depending upon the materials that even at a that low of a level could be affected. As for glass, remember that's not even a "true" surface. I've seen warped pieces of glass as well. And even hand sanding using emery paper can cause a bevel if the pressure isn't even. That's why I recommended the layout die...but yes, you can "paint" the surface with whatever you have. Sharpie will definitely do the trick because it's removable easily enough. I'm just used to what's on hand and forget not everybody has the tools/materials I do.
Good idea on using the salvaged socket to check flatness and use as a tool for lapping. Something that can help you keep the surface from warping by bracing it will help quite a bit.
 

·
Registered
Joined
·
378 Posts
I have never owned a CPU that I'd consider flat out-of-the box or that didn't benefit from lapping. Some are better than others, but I can always improve on the stock flatness.
Even objective measurements modern CPUs reveal variations several times that of recommended TIM bondlines. Igor's Lab looked at some recent CPU IHS not long ago.
Igor's he is one of the few that worth my respect, but he is not a machinist, when a CPU on top cooling plate becomes flat when CPU operating at 50C then this is perfect.
Copper this is a soft metal, it shape changes.
Your best bet, this is getting the best tower air-cooler in the market, something which can remove double amount of thermal load than the produced by the CPU.
 

·
Laptop Enthusiast
Joined
·
9,348 Posts
As a mechanical engineer/master fabricator, if you didn't lap it using a mill/cnc or calibrated surface grinder you will have issues. Or even use something like a calibrated surface plate and some super fine emery paper. If you didn't do one of those things, you will easily develop leaks in the mating surfaces. Kiraikos is correct in the thoughts that in today's tech, lapping really isn't needed unless there is a serious flaw in the mating surfaces. Fluid/thermal dynamic properties hold as well in the expansion/contraction of metals. Especially something as heat absorbent as copper. Even a micron sized flaw could mean a potential leak depending on location/temp/fluid pressure.
Yup. Admittedly, I ended up using the glass stovetop as a flat area and lapped an old Phenom II x2 chip about ten years ago and that turned out phenomenal. But yeah, if you're not using a proper flat surface, you will screw this up. I probably got lucky.
 
  • Rep+
Reactions: cloppy007

·
Registered
Joined
·
2,028 Posts
I can't understand why you lapped that 5820k, I'd never do that on mine unless the contact patch between the IHS and the waterblock is not good (in my case, it is). I did it in my old Phenom, though.

Anyway: how much time went by between the water ingress in the IHS and powering the system on? That hole is for preasure equalisation (Intel is concerned about that on its soldered IHS). If water found its way to the surface mounted resistors, that's potentially bad news.
 

·
Registered
Joined
·
1,886 Posts
I can't understand why you lapped that 5820k, I'd never do that on mine unless the contact patch between the IHS and the waterblock is not good (in my case, it is). I did it in my old Phenom, though.

Anyway: how much time went by between the water ingress in the IHS and powering the system on? That hole is for preasure equalisation (Intel is concerned about that on its soldered IHS). If water found its way to the surface mounted resistors, that's potentially bad news.
If the smds on a cpu were shorted he would know. Missing ram channels/dead cpu etc etc.
 

·
Registered
Joined
·
119 Posts
So in reading it sounds like you refilled the AIO which means you took it apart. My bet is you have an air bubble and need to bleed the loop and add more coolant. Try and put on the stock air cooler or put an aftermarket air cooler on to isolate the problem. AFAIK you should never open an AIO. To me it 100% sounds like a failed AIO and id bet putting a stock heatsink on it will get you going.
Agreed, the symptoms you are describing makes it think it is the AIO. Do you have the stock air cooler to test with to rule out the lapping as being the cause?
 

·
Registered
Joined
·
233 Posts
Yup. Admittedly, I ended up using the glass stovetop as a flat area and lapped an old Phenom II x2 chip about ten years ago and that turned out phenomenal. But yeah, if you're not using a proper flat surface, you will screw this up. I probably got lucky.
Actually that was probably the best glass surface you could have chosen. They aren't typical window glass and have pretty strict specs when made. Likely the flattest/smoothest surface anybody could choose in their house.
 

·
Registered
Joined
·
95 Posts
Discussion Starter #19
Hey guys, sorry for late reply, busy weekend.
Agreed, the symptoms you are describing makes it think it is the AIO. Do you have the stock air cooler to test with to rule out the lapping as being the cause?
Don't cover/plug the IHS vent, just wash the CPU thoroughly after lapping it and let it drain and dry completely.

My money is on something being wrong with the cooler, not the CPU.
So in reading it sounds like you refilled the AIO which means you took it apart. My bet is you have an air bubble and need to bleed the loop and add more coolant. Try and put on the stock air cooler or put an aftermarket air cooler on to isolate the problem. AFAIK you should never open an AIO. To me it 100% sounds like a failed AIO and id bet putting a stock heatsink on it will get you going.

Also, you need to keep sanding, that looks way too rough still but that shouldn't cause over heating like you describe.
AIO seems good. Outlet is quite warm (trying to cool a 90+℃ CPU) and inlet is much cooler. Pump and fans are all functional. Also, since it's X99 platform/HEDT, there is no stock cooler to test with 😂

First get a straight edge and see if your lapping is flat.
The pink stuff looks high based on shadows
What is the dark area at the bottom, looks like some TIM that hasn't been removed?
Need a flat edge, glass works well. Also need very fine grain. Sand paper.
Use some layout die to coat the ENTIRE surface you intend to lap. Then use UltraMega's idea of a piece of glass as the main surface.
You dont need anything special. Have a sharpie marker? Make an X corner to corner, then do a border about 1/4 inch from edge around cpu. Now do a cross in the center going to edges.
Don't need anything fancier than a flat surface (a dollar sheet of float glass) and some decent sandpaper to significantly improve upon surface flatness of a CPU.
I second using a piece of glass. I just try to apply even pressure when I am sanding, use water and a little dish soap to help with friction, and make sure to rotate the CPU in your appliance to reduce the chance of grinding it unevenly, as well as use different motions like side-to-side, circles, ect.
But yeah, if you're not using a proper flat surface, you will screw this up. I probably got lucky.
Used a piece of glass for a flat surface, used sandpaper made to be used with water and used water. Went from 240 grit all the way up to 2,000. The pink thing was me digitally highlighting the vent hole. Dark area on the bottom is just a dark reflection. Overall pretty sure the lapping is fairly even, and it has certainly leveled out the IHS more. I could sharpie it and do a few more passes but I'm not sure that is the problem. The paste pattern pre-lap showed a uneven patch and the paste patter post-lap shows a much better evenness.
If the smds on a cpu were shorted he would know. Missing ram channels/dead cpu etc etc.
This is really where I think things might be wonky. No dead RAM channels and the computer boots. Like I said in the first post, at first it was overheating in both BIOS and in Windows. After trying to clean out the vent hole, it fixed the overheating in BIOS but it still overheats in Windows. I'm wondering if Windows is activating a part of the CPU that uses a specific SMD that isn't in use when I'm in BIOS. That's the only conclusion I can come to. Wondering if I should get the CPU de-lidded by a service so that everything under the IHS can be cleaned.
 

·
Registered
Joined
·
1,886 Posts
Hey guys, sorry for late reply, busy weekend.



AIO seems good. Outlet is quite warm (trying to cool a 90+℃ CPU) and inlet is much cooler. Pump and fans are all functional. Also, since it's X99 platform/HEDT, there is no stock cooler to test with 😂








Used a piece of glass for a flat surface, used sandpaper made to be used with water and used water. Went from 240 grit all the way up to 2,000. The pink thing was me digitally highlighting the vent hole. Dark area on the bottom is just a dark reflection. Overall pretty sure the lapping is fairly even, and it has certainly leveled out the IHS more. I could sharpie it and do a few more passes but I'm not sure that is the problem. The paste pattern pre-lap showed a uneven patch and the paste patter post-lap shows a much better evenness.

This is really where I think things might be wonky. No dead RAM channels and the computer boots. Like I said in the first post, at first it was overheating in both BIOS and in Windows. After trying to clean out the vent hole, it fixed the overheating in BIOS but it still overheats in Windows. I'm wondering if Windows is activating a part of the CPU that uses a specific SMD that isn't in use when I'm in BIOS. That's the only conclusion I can come to. Wondering if I should get the CPU de-lidded by a service so that everything under the IHS can be cleaned.
DELID A soldered cpu??? leave it and fix your cooler
 
1 - 20 of 29 Posts
Top