Lapping CPU's...

[DavidBoren]

Now this is the exact sort of nonsense that interests me.

There is a quantifNow this is the exact sort of nonsense that interests me.

There is a quantifiable benefit, but most of all it's something a tinkerer can do with little more than time (and sandpaper).

Anyone lap their CPU and/or heatsink?

[Siress]

Nope. The only way there's a meaningful benefit is if you're OCing. My current rig doesn't need to be OCed. If I start needing more from it before a logical upgrade path exists, then I may OC it. In which case, I may lap it if time permits.

Lapping is really easy, but I see a lot of people getting it wrong. Sandpaper works decently well, but only if it's placed on a flat surface. A wooden desktop isn't it. An easy trick is to use a thick sheet of glass, like the reflective side of a mirror. These typically have a great flatness to them and they will not deform under pressure if supported by something like a bench top. Ceramic countertops are not recommended since they have waviness to them due to the way they are made. The sandpaper should be a quality brand, like 3M, and grit shouldn't be higher than ~1000 - due to the effects of the adhesive backed paper between the abrasive and the glass. I typically move to a 5micron alumina or SiC slurry as a next step, but it takes a long time to get better results at this point. I've polished down to 0.2 micron, and it took about a full day on automated equipment if memory serves. The goal with this interface is to minimize the distance that heat has to travel through the thermal compound, which has a relatively high thermal resistance as compared to the two metals interfacing. So, the flatter the surface - the better the result. I'm not sure if anyone has done the work to show how marginal those results get in the micron range.

[punkncat]

I think that by and large the qualities of most modern coolers and pastes make it a null point. If you can't see that something is amiss, which you shouldn't and it pass QC, I haven't spent the time doing this since long ago.
With engineering samples and such, or delidded, yes.

[gabe]

Hard pass on lapping. Your heat conductive paste makes CPU lapping pretty much useless.

Delidding on the other hand...that has huge gains

[DavidBoren]

And now I know about delidding. Lol.

I see now reason that delidding and lapping couldn't be combined, even done at the same time.

[gabe]

My temps dropped from 75-80* C now down to 68*C max under stress testing. Delidding is the real magic

[Falcon16]

Lapping the physical die itself is asking for major trouble. Some people have done it but it's people that are highly experienced

[Grendel]

Even with overclocking most people will not know how to make proper use of lapping the heat sink surface and most likely will make it worse. Even when my operation (with another company) made high end servers we did not lap CPUs but instead just insured they were inspected and cleaned prior to installing.

[Siress]

There were some Intel chips notorious for having low spots of >.005" depth. That's where lapping could make a very noticeable difference. It probably isn't relevant for other CPUs, as Grendel said. You can easily check this by scraping off old thermal compound using a straight edge and seeing if any thick spots remain.

[Falcon16]

lapping used to do something way back when in the early 00's now it does next to squat, de lidding is still viable. That all said with the power of modern CPUs overclocking doesn't produce a ton of extra performance, when most things like games are GPU bound

[Tracker]

i wouldnt recommend lapping, the only thing i have ever done extra is that i tend to rub the thermal paste into the mating surfaces and wipe it clean before applying the correct amount and mounting

i remember an old brand of thermal paste that was part of the instructions, the idea was to try and get as much of the compound "pushed into the pores of the metal" as possible.. im sure it does nothing. but you use it kinda like a rubbing compound, and it always comes off black from the heatsink side, so i can at least feel like it did something

[Siress]

According to renowned OCer, Kingpin, delidding doesn't really help these days (as of series 9 core i CPUs) while "lapping" is still performed. Note that what they did I'd just call sanding and polishing. I tend to reserve the term lapping for when there's precise control of flatness.

Exhibit a: https://www.youtube.com/watch?v=iShcG91eLoc

[Nish]

...

I read the title and came here expecting pictures of your dog licking parts on a computer you were working on.

[DavidBoren]

Some people say lapping is snake oil, others same the same thing of delidding... clearly the only thing to do is either both or neither.

You are either looking to get rid of heat because heat is bad, or you are bored, or you are seeking maximum performance... doesn't matter, if you're going to do it at all, it looks like you should probably do it right... or not at all.

[gabe]

1. Lapping makes no sense because the IHS and cooler plates will both have uneven surfaces which is why you use thermal paste. Even if you true them up perfectly you still can't run the metal heat conductive plates together without some form of thermal paste. So even if you get both surfaces perfectly true they still won't directly contact one another without some sort of paste between which would fill in any uneven spots and give a good contact.

2. You're not actually "lapping" unless you are mechanically truing the faces so they are flat and true to one another. What people do is polishing which just makes the faces smooth but not true to one another so regardless you will have some gaps.

3. The point is 100% to lower the temps. It's not about maximum performance, it's about longevity of your hardware. Electrical components (especially the micro transistor switching in processor dies) wear out with heat. Lower temps will make them last longer before the circuit pathways break down.

4. You don't need to do any of this with stock clock speeds but it doesn't hurt. Really it's not necessary unless for some reason you are seeing temps upwards of 70+ degrees C on any cores under stress testing.

5. Overclocking is still commonplace and necessary on some chips. The lower end Intel series benefit quite a lot from overclocking. My I5-8600k was able to go from 3.2mhz to 4.6mhz stable and runs right along with a friend's I7-8700k for a fraction of the cost. All that it needed was a good overclock tune and delidding to keep temps within range. Voltage didn't even need to be boosted past stock levels.

6. Delidding is still the way to get the best results with lowering temps. Stock thermal paste that is applied is applied sloppily and isn't of a high quality. Intel is notorious for this especially on the I9's since they run a higher wattage and are fairly inefficient. On my I5 a simple delidding and using grizzly liquid metal lowered temps 20-30* C. On the I7 it was the same story. These are 8th gen chips so not exactly old. Even the newest chips suffer from sloppy production techniques. They're building them by the millions, not hand building them for maximum quality.

[JohnSatclaire]

The issue often is a mismatch between the IHS and waterblock's respective contours, rather than whether there are minor surface imperfections that could be polished out. Thermal paste ensures good contact regardless, but because it's relatively resistant to heat transfer (compared to something like LM or a soldered TIM), you want the paste layer to be as thin and uniform as possible. If your IHS is unusually concave and your block is flat, lapping the IHS down will cause that even, thin paste spread and lower temps.

Something like Aquacomputer's vario blocks is a different approach to solving the same problem. Personally, I'm looking forward to finally upgrading from X99 so I can get a bare-die compatible chip. Though there are some pretty crazy potential gains to be had from lapping the dies...

[DavidBoren]

Do you use the same liquid metal on both sides of the lid?

[gabe]

No, you use liquid metal between the IHS (integrated heat sink) and the die and you use thermal paste which is thicker between the IHS and heat plate (water cooled plate or heat sink if air cooled).

The IHS and die can be made very tightly fitted because they are semi-permanently bonded which is why you use a thinner but more efficient liquid metal between them. The IHS outer side to the cooling plate (in the case of a liquid cooler) is a much looser fit as it is just held in place by 4 thumb screws which are never perfectly torqued on a large area of the motherboard which flexes when you tighten things down. This means there is always some gaps between the IHS and cooler plate. This relatively larger gap necessitates the thicker and less efficient thermal paste.

In any case the bottleneck is always the CPU die to IHS medium (liquid metal) first because the factory never uses the best ($$$) compound, and secondly the efficiency of the cooling unit whether liquid or air cooling. Improving these two areas has freed up a bunch of thermal headroom in builds I've done.

As a side note if you get a nice air cooling unit like a Noctua double or triple vapor chamber unit they cool just as well as a 240 or 260mm radiator AIO cooler. They just suck a lot more to try and bolt down the big ass air cooler but the air cooling is FAR quieter than listening to the water pump whirr in a liquid cooled system. The noise doesn't bother me but if it was an everyday work computer instead of a gaming computer I'd go air cooling.

[Seajay]

Good luck if you want to use Liquid Metal. That stuff can be really nasty and it is not nearly as user friendly as just a good paste. If you are not seriously overclocking it won't do much either. The difference from going from 60'c to 54'c doesn't really matter, sure it is cooler, it does work, but it's still producing heat and energy output.
If the LM leaks it will ruin your board and anything else it drips on.

The only time I've ever used LM is on a horizontal setup, like a test bench, specifically for doing silly amounts of overclocking. Bascially a temporary setup to do a particular task.

If you will be using it for long term on a vertical setup, like most cases, grab some thermal sealant (not epoxy) to re-seal the IHS to the CPU. That way any leakage is contained within the CPU.
And no... hot glue does not work :P

For lapping grab a lapping kit. Lap both the IHS and the bottom of the cooler so you can try to get both surfaces the same flatness. The more you take off the IHS (thinner) the easier it will be to transfer heat to the Cooler, but also the less heat it can manage, so the cooler will need to be sufficient to pull that heat quickly (not usually an issue).