Lapping Sandy Bridge

[ShadeZeRO]

Has anyone lapped their Sandy Bridge processors yet?

I get my 2600K monday, and plan on lapping it to 2000 grit.

Just wondering what everyone's experience is.

 

[Diogenes2]

Wouldn't it be a little early to void your warranty before seeing what you can do without lapping ?

 

[Kenmitch]

Before lapping it just check it with a little compound and a piece of glass. I doubt it would need lapped....But it's your chip and your warranty that will get voided in the end.

 

[BTRY B 529th FA BN]

I think you are making a mistake lapping that.

 

[SickBeast]

You should never mess with the heat spreader IMO. I've done it before and I lived to regret it.

 

[Anomaly1964]

YIKES...

Risk - HIGH
Gain - Seems MINIMAL

 

[BTRY B 529th FA BN]

Risk - HIGH
Gain - Seems MINIMAL

^this...

its 32nm anyways and people are talking about 5Ghz on these chips... i mean that coherently includes and implies that they run cooler anyways

 

[VirtualLarry]

Why would you even NEED to lap it? Temps are supposedly low for this chip as it is.

Now a Q6600, I could see lapping. (But I got mine to 3.6 without lapping, so it's definitely not necessary.)

 

[Edrick]

Why would you even NEED to lap it? Temps are supposedly low for this chip as it is.

Now a Q6600, I could see lapping. (But I got mine to 3.6 without lapping, so it's definitely not necessary.)

This +1

 

[Seero]

Lapping isn't necessary, it was never necessary. People do it not because they think it is necessary. I think, they simply do it for fun, and for the OC blood within them.

I lapped my Q6600 the first day. I lapped my I7-920 the first day, and I will lap my 2600k the first day. I am still waiting for the right mobo before purchase.

 

[mnewsham]

Lapping isn't necessary, it was never necessary. People do it not because they think it is necessary. I think, they simply do it for fun, and for the OC blood within them.

I lapped my Q6600 the first day. I lapped my I7-920 the first day, and I will lap my 2600k the first day. I am still waiting for the right mobo before purchase.

And how did the temps work out? significantly lower? Low enough to risk being SOL if your processor craps out and you need to return it? Sorry i will keep my warranty.

 

[ShadeZeRO]

And how did the temps work out? significantly lower? Low enough to risk being SOL if your processor craps out and you need to return it? Sorry i will keep my warranty.

My Q9650 and TRUE are both lapped, and it did help lower temps by a pretty decent margin.

 

[Zap]

And how did the temps work out? significantly lower?

If it was lapped on the first day, then there is probably no basis for comparison.

 

[Seero]

If it was lapped on the first day, then there is probably no basis for comparison.

Wow, at least one person who can read. Yes, i boot it up, to make sure that it runs, and test is a bit, then start lapping it for several hours. Once the lapping is done, installed and tested to make sure all other related stuffs are done right. In short, I really can't have basis for comparison.

And how did the temps work out? significantly lower? Low enough to risk being SOL if your processor craps out and you need to return it? Sorry i will keep my warranty.

I can tell you how good the lapping job is, but I really can't tell you how good the CPU performs after the lapping. It works, and technically lapping doesn't change the performance of the chip, and that applies to heat generation of the chip.

As to better contact, I will say a poor application of the thermal paste can have a bigger impact on contact, and therefore temp. I tried many different methods and stayed on the ones that works best, yet I forgot which method was that. Hey, I wasn't going to write a book about it. The fact is, since the actual spreading happens when you put the HS on the CPU and won't be able to see how it spread until you remove it, you will actually have no idea how it spread. Now does thermal paste spread better on smooth surface than rough? I will say no, but that itself can be a thread.

So regardless of how good the lapping is, you still needs to put Thermal paste on it. However, if it is lapped well, then the amount of thermal paste needed is, in theory, reduced. Because of that, applying thermal paste on lapped surface is far harder than lapping the surface.

After you have done all those and question all those, you would have run enough tests that others probably will never do in their entire life. In fact, this is what Overclocking is about IMO. Instead of working with a lot of variables at once, you focus on resolving a variable at a time. To me, make sure that the heat generated by CPU as well as all other component radiates properly is the third thing to do before OC (the second thing is component selection and make sure that they work under specification, some may take one more step in modding components to secure stability related issues, and the first thing is to read instructions, and specifications).

In comparison, the time and work to lap CPU is minimal as reading and part searching can take up to weeks individually. Lol test.

Does it worth it? It varies upon people. Again, the time and work of lapping is really small in comparison, and relatively safe. I will say changing resistors are more dangerous, and pencil mod is 1000x more dangerous than lapping. Some of my friends actually brought silver ink pen for the job and we ended up with arguments because I argued that changing the resistor is easier and safer blah blah blah.

Is it worth it for me? Of course, the fact that I lapped it right makes it worth it. Yes, after spending hours lapping your newly brought CPU against different sandpapers with water, it is actually rewarding to see the CPU still works.

Is it risky? Nope, not in comparison. Never even heard of someone breaking their CPU by lapping it.

Return when broken due to lapping?
http://www.youtube.com/watch?v=lrF5x1OSJuQ

You will know the answer after the first pass of lapping.

 

[BonzaiDuck]

Happened across this thread as I contemplate lapping my own (retail) i7-2600K.

Suddenly, all this "new refinement" crops up on self-congratulatory guides and demonstrations for lapping an i7-2600K CPU.

I notice someone is STARTING the lapping with 800-grit wet-or-dry sandpaper. [And I must ask -- "Why not 320, then 400?" The subject article notes a progression to 2000-grit?!

The subject article warns "Do NOT use water(!)"

Here's what I'd found and posted per LGA775 cores (and I have lapped exactly every single such CPU that came into my possession.)

First: I conducted some experiments that deployed the CoreTemp or RealTemp logging feature -- to record perioidic temperatures under either PRIME95 or LinPack stress. The ascii-text file logs were then loaded into Excel -- to generate arithmetic means, standard deviations, and bar-graph frequency distributions.

For a processor with between 65W and 95W TDP, you should expect a drop in temperature at load up to 5C degrees -- IF -- you lap off all the nickel-plating from the IHS, exposing the copper.

Second: If your heatsink-cooler base is nickel-plated copper, you can expect an identical temperature improvement if you similarly lap off all the nickel.

Third: I don't see what good it does to polish the IHS with finer and finer grades of sandpaper until the reflected image is equivalent to HDTV 1080p quality. A surface that has the polish of the original IHS (which doesn't reflect an image) should be fine.

And fourth: choice of thermal paste may get you between 2C and 3C improvement. I don't just "recommend" IC-Diamond; I don't just "think" or "believe" that nano-diamond paste is "better." IT IS better. However, don't waste your money on a thermal grease purporting to include nano-diamond particles comprising only 10% of the ingredients. You can MAKE your OWN paste if you're willing to pay for nano-diamond powder by the gram -- using a cheap silicone thermal grease that comes with most heatpipe coolers. Go ahead and spend the $7 (or whatever) on the IC Diamond product, and suffer through the experience of spreading such a thick, highly loaded paste. In fact -- recent posts decry the use of a razor-blade, noting that the "pea" method of just putting a "drop" of paste on the IHS works better. Instead -- I recommend spreading an imperfect coating of diamond-paste with a razor-blade on both the IHS and heatsink, then adding a half grain of rice amount to the center of the IHS before placing the heatsink.

Now -- the improvements will scale down as the TDP of the processor increases. What might be a 10C improvement for a 65W core may be proportionately less for 95W core. And certainly -- for a 135W core, even less.

Maybe someone can enlighten me as to why a drop of water on the sandpaper would be different for an i7-2600K as opposed to an LGA-775 core. And you could further explain to me why the DIY-project with pictures I saw had the author handling the gold-contacts on the CPU bottom with his fingers. Or why he didn't use the plastic protector the chip is shipped with to hold it while lapping.

But the main thing: I have to agree with those who say "Why bother?" with the i7-2600K. With that sort of scalability, you might just as well temper your over-clocking efforts to keep the temperature and voltage at near-stock levels. You'd take less risk by lapping the heatsink bottom (even if you void the warranty on a $70 item).

And either way -- there's a day of soreness in the wrists -- especially if you're flippin' OLD like me -- or the worry that you gave yourself "carpal-tunnel."

That being said . . . . I'll probably do it anyway. I just won't do it today . . . Manana!!

 

[yottabit]

Might as well just take off the heat spreader and go straight on the die. Don't tighten the clamps too much!...'

That's what I'd do if I wanted to void my warranty rather than waste time lapping. No point unless you're doing extreme overclocking though. On a prescott it might have made sense, but not nowadays...

 

[khon]

Bad idea.

The chip already runs very cool, and I doubt you would gain much of anything at all.

 

[Bartman39]

Grinding metal is fine if your doing body work or fitting for a weld but on a modern cpu...? Sure I lapped Celerons & some heatsinks to make then perfectly flat but with modern cooling and die sizes we have now there is no real need unless your going for a record or just have money falling out your @zz...?

I really see no need with a 2600K @ 4.6Ghz idle temps 28-31C and 1 hour of Prime95 62-64C (stock H70)...? (fans on lowest settings too...) Spend more time and money on your case & HS/fan to get the kind of cooling you want...

 

[quakefiend420]

i lapped my i7 920 and saw a pretty decent improvement in temps, don't remember how much exactly now, but it was somewhere in the neighborhood of 8-10C load drop IIRC

i'm with the OP, i'll keep lapping my CPUs until there's nothing to be gained from it

 

[Suspicious-Teach8788]

so how does warranty matter if you're going to overclock the hell out of your processor anyway?

 

[Ben90]

so how does warranty matter if you're going to overclock the hell out of your processor anyway?

A lot of people are okay with warranty fraud.

 

[PlasmaBomb]

Now -- the improvements will scale down as the TDP of the processor increases. What might be a 10C improvement for a 65W core may be proportionately less for 95W core. And certainly -- for a 135W core, even less.

What?

You should see larger improvements with high TDP processors.

If you see an improvement in the thermal conductivity of the TIM by 0.02 C/W, then the difference between the two solutions would be 1.3 degrees C @65W, 1.9 degrees C @95W and 2.7 degrees C @135W.

so how does warranty matter if you're going to overclock the hell out of your processor anyway?

Yeah I'm surprised that no one pointed out that the warranty is void as soon as your OC your chip. So you aren't loosing anything by lapping after you have overclocked...

 

[BonzaiDuck]

What?

You should see larger improvements with high TDP processors.

If you see an improvement in the thermal conductivity of the TIM by 0.02 C/W, then the difference between the two solutions would be 1.3 degrees C @65W, 1.9 degrees C @95W and 2.7 degrees C @135W.



Yeah I'm surprised that no one pointed out that the warranty is void as soon as your OC your chip. So you aren't loosing anything by lapping after you have overclocked...

My mistake. I think you're quite correct about that. I believe I must have been thinking of a temperature improvement as a fraction of a base temperature.

We should also keep in mind that OC'ing the chips also raises their thermal power.

In any event, a one-degree improvement at a load value of 55C would be < 2%; the same improvement at a load value of 70C would seem smaller.

My rigorous experiments with lapping and diamond-paste were conducted with an E6600 and then a Q6600.

I've seen posts on other forums that either the lapping would be deleterious, or of little value. Some seem to think it is only a matter of flattening the IHS and heatsink base, but that's not true. Nickel has a higher thermal resistance than Copper, so eliminating the nickel where the IHS meets the heatsink-base is important. Decreasing the thermal resistance of the thermal paste (conversely increasing thermal conductivity) is also "important" given the stunning difference shown by diamond in those factors as compared to the various metals, compounds and formulas found in commercial pastes.

I don't disagree with Bartman39 that all this trouble may seem less useful at the margin.

On the other hand, unless the very basic construction of CPUs and choice of IHS materials has changed, the thickness of the IHS pretty much eliminates the risk of doing any damage by lapping off only the nickel plate. What I'd heard which makes the most sense: you run some risk of getting metal dust into parts of the core where they would do damage. This was also the precaution about using a drop or two of water. But with a little care and some sponging during the lapping process, I don't count this risk as very big.

Some say it takes about 40 minutes. I wonder if it isn't at least double that time, and then double again for lapping the heatsink base.

 

[Ben90]

Generally lapping high end heatsinks lowers performance. They are bowed for a reason.

 

[SonicIce]

you should unlid the cpu. there are no exceptions