Lapping Sandy Bridge

[Markfw]

you should unlid the cpu. there are no exceptions

Wrong.... Just wrong

 

[BonzaiDuck]

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

You may be in the embrace of another myth with that. My 2600K IHS was perfectly flat before I ever desecrated it with sandpaper. There WAS a slight bow in the NH-D14 heatsink base, just like that of the old TRUE -- but not nearly so pronounced as the TRUE.

Wrong.... Just wrong

You got that right. Some surfers may find it thrilling to try the waves at Ruby Beach, Washington, or push their luck in favorite hunting grounds for the Great White. That doesn't make such decisions prudent . . .

On my own project, I finally broke down and lapped the heatsink base. Then pondered whether to do the 2600K IHS for several hours. Finally -- I just did both surfaces. And no -- I didn't make them to a mirror shine -- they're nice and scratchy. Like a bronze-age attempt at a mirror. I finally decided to give them a little time on 1000-grit wet-or-dry, but not much . . . .

It most likely wasn't necessary; I voided two warranties; but it's just a habit now, and I can't help myself . . . .

 

[ShadowVVL]

huh you know I was just about to lap the stock heat sink on my amd 5600+ since it runs hot,like 52c and 60% load with a room temp of 70f.

after reading this the question is should I?

 

[996GT2]

huh you know I was just about to lap the stock heat sink on my amd 5600+ since it runs hot,like 52c and 60% load with a room temp of 70f.

after reading this the question is should I?

I've lapped the IHS on an Athlon X2 and it did result in some gains...but honestly the few C reduction in temps is not worth the effort. Get a better heatsink if you really want to reduce your temps by any meaningful amount.

 

[Ben90]

huh you know I was just about to lap the stock heat sink on my amd 5600+ since it runs hot,like 52c and 60% load with a room temp of 70f.

after reading this the question is should I?

Since you had to ask, the answer is a definite no. You don't lap solely for improved temperatures. It is for the people that have a love for the hardware and want to maximize its potential in whatever bounds are set.

Whether there is an appreciable difference in temperatures or not is irrelevant. Its the peace of mind that you have done everything you can do to minimize the temperature delta between the IHS and HS.

 

[ShadowVVL]

ahhh and I already paid 89cents for all the sand paper lol.

thx 996gt2, you probably saved me from 1-2 hours of labor ill get a $15-$20 H.

 

[Rifter]

I've lapped every CPU i have ever owned, my current i7 930 had a large depression in one corner that took hours to get flat and im 100% sure would have had a large negative effect on cooling. While i didnt do before and after comparisons of my i7 i have in the past and it usually yields a 5-10c load temp drop. Obvioulsy i lap the heatsink base as well.

 

[john3850]

I lapped my 930 and a few c2d.
I only use a HSF to run off a new mb out of case.
Does the extra weight and flex of a large hsf on a vertical mounted mb add to tempeture problems.
Would the use of a oc station were the mb sets flat be better then a case mout when using heavy hsf.

 

[BonzaiDuck]

I lapped my 930 and a few c2d.
I only use a HSF to run off a new mb out of case.
Does the extra weight and flex of a large hsf on a vertical mounted mb add to tempeture problems.
Would the use of a oc station were the mb sets flat be better then a case mout when using heavy hsf.

Never had that problem. Using LGA-775 as a reference, Intel advises -- for its motherboards -- that the HSF fan weight be less than about 400 grams. But since before the ThermalRight Ultra-Extreme 120, many of these heatsinks, like the massive products from Zalman of that period, exceeded 800 grams. some HS/Fan combinations would exceed a kilogram.

But the retention mechanisms and brackets shipped with these newer units, like the Prolimatech Megahalem or the Noctua NH-D14, are quite sturdy. Noctua seems very confident that their D14 -- with two fans -- will do no damage. Noctua is also very explicit in mentioning the INtel standard. Also, keep in mind that many of the newer motherboards (like the Gigabyte Ultra-Durables) have a sturdier design.

I was quite stunned with the NH-D14. I'm using a CoolerMaster HAF 922 case. The motherboard just slid right in to fit the 1/4" standoffs -- after I'd installed the NH-D14. Later, investigating with a piece of paper to test clearance with the motherboard pan, the tolerance for the backplate is so small that you can't fit a piece of paper between one corner of the backplate and the motherboard pan. But there is absolutely no tension on the board arising from the close fit. The remaining corners seemed to just allow sliding a piece of paper in there . . .

In theory, a motherboard sitting horizontally for a test-bed has been standard procedure. But the best coolers behave as promised. Never had a problem.

 

[BonzaiDuck]

Here's a quote from the ThermalRight web-site for the MUX-120 cooler (reminiscent of the old TRUE):

"Convex copper base design, to ensure the Highest thermal conducting thermal efficiency between the cpu and the heatsink."

It would seem that heat dissipates from the IHS spreader across its entire area. The hottest points on the spreader would likely be where the cores are located. Perhaps for a quad-core, the orientation of the cooler itself would be indifferent to those hot-spots.

But I fail to see how a convex base would draw heat from the entire area of the spreader. It would seem that the more thorough the contact with the IHS, the more "efficient" would be the heat transfer. Areal contact is a prime factor -- that's always been known.

The difference between the contact of nickel-plated surfaces and copper-surfaces seems equally significant.

Let's put it another way. The benchmark results for an OC'd i7-870 have been posted in a review, and other reviews and testbeds are available. If "my way" gives inferior results considering thermal power and room ambient, I will amend my ways. However, I already ran those "experiments" on a convex TRUE lapped flat to compare before and after scenarios without the nickel plating. I'll be surprised if -- as another poster insinuates -- there is a "degradation" in cooling efficiency.

The purpose of the nickel-plating -- is established generally and touted in the sample ThermalRight promotion. It inhibits corrosion. That's why nickel is chosen: it is adjacent to copper in a galvanic table. But I'm skeptical as to how removal of the nickel from two surfaces sealed in thermal paste will be vulnerable to corrosion.

 

[john3850]

thanks BonzaiDuck

 

[yottabit]

Wrong.... Just wrong

AFAIK, you can get a few extra degrees more by unlidding the CPU. It's not something I'd recommend for your average plug and play tinker-tot, but it's a proven tool in OC'ing

 

[BonzaiDuck]

AFAIK, you can get a few extra degrees more by unlidding the CPU. It's not something I'd recommend for your average plug and play tinker-tot, but it's a proven tool in OC'ing

I've seen forum posts with photos -- circles and arrows on each one (to quote Arlo) -- per "unlidding." One such project just ground down the copper IHS to expose the cores.

But that's a territory I never wanted to explore. It's the "Extremo di tutti Extremi" if I'm even close to the italian words.

Junking the nickel plate has a significant effect, but not so significant that you can't live without it. I depend on the INtel retail-box 3-year warranty -- not because I expect to use it -- but because it signifies that they don't ship processors any way likely to fail. You could OC the heck out of a core and actually get it replaced as you exercise a lawyer's honesty (don't tell 'em what you did). But once you strip off that nickel and the identifying stamp, you're driving without any insurance. Which . . . is something I choose to do . . . with computer-parts -- that is.

On the "convexity" angle. They may think -- may even be able to prove -- that extra pressure and near-metal-to-metal contact at the "center of things" pulls a maximum of heat away from the core and the area nearest the core before that heat can spread over the entire processor cap. But all I ever noticed by flattening a TRUE was a difference between nickel and copper. . . .

ALSO -- MY PERSONAL DISCLAIMER (since I didn't start this thread). This thread probably belongs in the "cases and cooling" forum. However . . .

If I recall, increased temperature requires slightly greater voltage to hold a clock setting. But greater voltage means more heat . . . . If you don't water-cool but you over-clock, you may want to mitigate that synergy. Maybe -- lapping would be worth a few Mhz, a few dozen Mhz -- heck! -- maybe even a hundred or more.

Howsoever that works, I try and find the maximum speed and minimum voltage, and then reduce it a tad. So you're free to speculate what the final setting might be, with or without lapping the IHS and HSF-base.

On the "waste of time" criticism, it may depend on whether you work for a living, have the distractions of family -- all kinds of factors. I can say this last time around it may have cost me an hour's time. But you can do it while watching Daffy Duck on Looney-Tunes TV . . . . so how could it be a waste of time for that sort of effective multi-tasking?

 

[TridenT]

You only lap when you've got a shit ton of money, you're pushing everything to the limit already... (cooling wise and overclocking wise) and you're like, "How can I squeeze 50 mhz more out of this?"

 

[BonzaiDuck]

You only lap when you've got a shit ton of money, you're pushing everything to the limit already... (cooling wise and overclocking wise) and you're like, "How can I squeeze 50 mhz more out of this?"

But what's the worst that will happen? I may invalidate a warranty on a $300 CPU, but the chances that same CPU will either prove defective or get "injured" in the lapping process are pretty small. Suppose you spent $70 on the heatsink. So? You voided your warranty on the heatsink, but you didn't destroy it -- didn't "damage" it.

Point being -- if you can squeeze the load temperature down by 10+C, then it should similarly affect temperatures at less-than-load stress. That's better for the system. So -- you voided the warranty, but reduced operating temperatures.

Do I disagree with an observation that it's "unnecessary?" No. But we constantly go day-to-day making decisions that either eliminate risk -- at a cost -- or embrace and accept some level of it -- also at a cost. We just spent $230 to repair a 20-year-old clothes-dryer. It works better than it ever did. We still face the risk that something else will go wrong with it. We might have bought a new dryer with a 5-year warranty -- for $900 clams.

No matter how you make those choices, life is a stroll through the Blackjack tables at the Indian Casino. Mitigating risk also costs money -- often more. And would I lap a $1,000 retail-box Gulftown? I don't know. That's a lot of money. But then -- a Gulftown has a 135W TDP!! Maybe . . . . I would!

 

[Ratman6161]

I've lapped the IHS on an Athlon X2 and it did result in some gains...but honestly the few C reduction in temps is not worth the effort. Get a better heatsink if you really want to reduce your temps by any meaningful amount.

I think your statement "not worth the effort" pretty much sums it up. Personally I tend to be going for moderate overclocks where I don't have to do anything more than a minor voltage increase and I normally have a decent but not top of the line cooler. I have always found that heat is never the thing that ends up being the limitation for me. I have always reached the point where I was backing off to maintain stability before I got to the point where heat was an issue. So saving a small percentage by lapping certainly wouldn't make any sense to me.

Example: with my Old Q6600 (which for some reason people seem to think ran hot - mine didn't) I got it stable at 3.6 Ghz but at a voltage I wasn't comfortable with for the long haul. Even at the increased voltage it never went above about 68c under load (I had it set to shut itself down at 70 which it never did). After lowering the voltage, I ran it completely stable for three years at 3.2 Ghz and stock voltage and under full load it might occasionally hit 60c. Experimenting with my 2600K at 4.6 it doesn't seem to go above the low 60's either (with my $24.99 cooler). So unless you care more about setting some sort of record rather than actually use the system - I just can't see lapping.

Actually I used to hear about people doing this all the time but I thought (hoped) the idea had died out. That said, if you own it, do as you like with it.

 

[Phunk0ne]

not into some crazy overclocking and such, leave your IHS the way it is.

But my experience with lapping the IHS has done nothing but miracles.

my trusty C2D 6750 saw a 10C difference between stock and lapped. Yes, I thouroughly tested this using the stock set-up.

I do have to mention that I had my TRUE-120 pre-lapped, but after recieving the cooler, I was not really satisfied and spent a few more hours in getting a very nice shine.

all in all it did exactly what I expected it to do, having a minimum of 5C difference between lapped and stock. The same still applies to my C2Q 9650.

but please beware: I am more of the lucky ones I suppose to achieve cooler temperatures at same speeds between stock and lapped IHS & cooler. I have read and seen cases where shit hit the fan after a lapping session.


you think it's worth it, of just like the idea of a mirror finish CPU IHS... well you know the deal

 

[BonzaiDuck]

. . . . . .
but please beware: I am more of the lucky ones I suppose to achieve cooler temperatures at same speeds between stock and lapped IHS & cooler. I have read and seen cases where shit hit the fan after a lapping session.


you think it's worth it, of just like the idea of a mirror finish CPU IHS... well you know the deal . . . .

That's a new one for me -- never heard anything like that. I've lapped six heat-spreaders and four HSF-bases since '07, and they all came out running cooler.

I also tried some dingy experiments, like grinding down a silver dollar to fit between IHS and HSF (which was bound to yield poor results, when you think it through. Silver plating? Sure. But adding another pair of surfaces? No cigar . . . )

BTW: I just don't think getting a high polish on the surfaces is guaranteed to improve things. Perhaps the result would also depend on what type of TIM being used. I use IC Diamond, which might as well be a mix of industrial abrasive diamond particles and a thermal grease, so you'd imagine that enough of the worthy particulate would get trapped between IHS and HSF. But even so, I'd think a slightly rougher finish would work fine -- even optimally.

If I'm right about that, all the extra effort to achieve a high polish IS what we've been discussing -- a waste of time. As I may have mentioned, I saw someone's "project" on the web where they started with 800-grit and finished with 2000-grit. Saving time, I'd start at 220 or 320, and finish with either either a progression of 400 and 600, or -- heck -- just 400.

EDIT: and finally . . . Speaking of risk, folklore and news about results "being worse" -- I can still imagine such. These elaborate heatpipe coolers can be damaged if you grab them the wrong way and stress the welds while lapping. So? You daintily grasp the heatsink-base.

Or -- you could get careless using too much water for lubricating the wet-or-dry, to allow a slurry of metal particles and water to infect the interior under the IHS. There's risk . . . sure . . . . For what others have said about this, our discussion itself would seem to waste time over wasting time actually lapping the surfaces. But like I said -- it should take less than an hour, and you can do it while watching TV.

 

[Seero]

You only lap when you've got a shit ton of money, you're pushing everything to the limit already... (cooling wise and overclocking wise) and you're like, "How can I squeeze 50 mhz more out of this?"

Why do you need to have a shit ton of money to lap a CPU? If you spend more then ten bucks, then either you have been seriously robbed or you did something seriously wrong.

Look, you can spend 50 bucks on a hs over stock for a better temp, or you can lap the stock and the CPU to squire the samething. You can also lap a 50 bucks hs and not drop 50 bucks more on a super fanny hs. Get the picture?

There is a difference between dangerous and may cause damage. Lapping a CPU may damage it, but the chance is lower then one to damage the socket while placing the CPU. Yes, one do need some crafting skills to do it, but that is it. Compare to blindly overvolt, lapping is by far safer.

Note that I didn't say it is necessary, in fact oc is unnecessary, and so is an aftermarket hs. These kinda thing is what oc is about. Instead of unloading cash on overpriced, handpicked units, one will utilizes what they have, craft with their skills, and tune with their knowledge, to maximize the unit at hand.

I won't claim I am an expert, but I can tell you how does it feel to beat others setup with a third of the cost. It is something that those who believe oc kills PC and/or lapping is not worth it will even have a chance to experience. And this feeling is what drives overclocker to oc. It isn't about 50mhz, it is the definition of what is possible and how to get there.

It is about passion, not necessity. It is about pushing the boundary, not Hz.

 

[Dark Shroud]

Seero, there are things in your post that I disagree with and things that are flat out wrong.

You don't need to drop a lot of cash to get a better than stock cooler. XIGMATEK has several $30 or less options that are top notch.

Where I live an over-clock without a good cooler will burn your CPU out in the summer. Not to mention having a decent cooler prolongs the life of the CPU when you push them hard. I can understand lapping the bottom of the CPU cooler. I see no reason to lap a CPU & void the warranty for a very slight bump in performance that could also be had with an extra $10-$15 invested in a cooler.

 

[Rifter]

Seero, there are things in your post that I disagree with and things that are flat out wrong.

You don't need to drop a lot of cash to get a better than stock cooler. XIGMATEK has several $30 or less options that are top notch.

Where I live an over-clock without a good cooler will burn your CPU out in the summer. Not to mention having a decent cooler prolongs the life of the CPU when you push them hard. I can understand lapping the bottom of the CPU cooler. I see no reason to lap a CPU & void the warranty for a very slight bump in performance that could also be had with an extra $10-$15 invested in a cooler.

Lapping just the cooler is only half of it, if the heatspreader is not perfectly flat like the heatsink you just lapped its not going to transfer heat as well as it could be. As seero mentioned lapping and the more extream side of overclocking is about getting as much performance as possible out of a chip without spending $1000 on EE or handpicked CPU's. Its nice to have a midrange CPU pushed to the limit and able to compete with or even beat units costing 2-3 times more.

The warranty doesnt even matter as overclocking will void your warranty anyways, so who cares if you lap or not you still have no warranty if you overclock.

 

[BonzaiDuck]

The warranty doesnt even matter as overclocking will void your warranty anyways, so who cares if you lap or not you still have no warranty if you overclock.

[Heh-heh . . .] Looking at some other posts, gotta throw in that line from "Cool Hand Luke." "What we have heah . . . . 's a failuh to comoonicate . . . a failuh to comoonicate . . . "

So the "shitpile of money" remark. That's somebody's perception of risk (versus mine or someone else's) in a reverence for warranties.

But I'll beg to differentiate . . . not entirely disagreeing . . . with your statement. You can "get away with" an RMA return after OC'ing a processor, if there's no cost-effective (emphasis) way for INtel (we're talking about Sandy Bridge) to conclude that you "abused" the processor.

Lap off the nickel-plating and identifying information from the IHS, you won't even bother to try or spend the postage.

I just fired up my Z68 system for the first time, and had to do a little research to find that the new chipsets report higher temperatures in BIOS than the older technology. So I really have to wait until I install CoreTemp or RealTemp with a final OS installation. It would seem that 44C idle with room-ambient 78F is on the "high" side. But I simply know that the NH-D14 cooler is properly installed; copper-to-diamond-to-copper will yield a lower temperature; and all thermal sensors including tCase have a potential to be defective.

Given that, and seeing that the OS installed in what seemed less than 20 minutes, I'm not worried at all.

I was looking at the default voltage reading on this puppy. These are LOW voltages -- it's stunning. And anyway, the proof of the pudding is in the load temperatures. It looks like there will be a "lot of headroom" for OC'ing on the voltage angle. For the temperature -- like I said -- I have to wait for the final OS installation Friday (when a VelociRaptor arrives) and the installation of the monitoring software. . . .

But there will be NO RMA's on any of these parts, and no "loss" for destroying an opportunity to RMA my CPU or heatsink. At this point, I have only "good things" to look forward to here . . .

 

[taltamir]

OP: if you want to lap, then lap away.
But its recommended to run it for a day or two first to make sure it isn't defective before voiding your warranty.

Personally I don't lap

 

[Skiprudder]

Odd but interesting thread. What I'm curious about is lapping in terms of an industrial and financial sense. Take the nickel plating on cpu spreaders. Putting this coating on obviously costs Intel and AMD money, and if it is detrimental to performance, why would they do it? I would guess there are longevity reasons at play here, perhaps the oxidation of copper? Added to this is the point that if AMD or Intel could bump chip speeds up by a few hundred Mhz simply by mirror polishing their chips--and bump up the price they sell them for (and still maintain the same TDP, etc) wouldn't it make financial sense for them to do so? I'm sure they've looked at this, and the fact that they don't lap to this degree makes me rather suspicious of significant performance advantages. Or at least ones that don't have significant ancillary drawbacks.

 

[taltamir]

Odd but interesting thread. What I'm curious about is lapping in terms of an industrial and financial sense. Take the nickel plating on cpu spreaders. Putting this coating on obviously costs Intel and AMD money, and if it is detrimental to performance, why would they do it?

Lapping does not involve the removal of the heat-spreader. instead it is made to be perfectly flat and polished. Quality heatsinks are factory lapped. CPUs aren't because it is an added cost that intel and AMD does not feel to be justified.