Lapping Sandy Bridge

[Joseph F]

I think that you should definitely go for it if you have had any prior experience with lapping. If this is your first time, don't try it unless you have another $300 ready to buy a new CPU if you somehow irreparably screw up your IHS. I have heard that Intel's IHSes are terribly concave when unmodified. (enough to make a 5-10C temperature drop after lapping)

 

[RobDickinson]

Because even if it made 2 or 3 degrees difference no one should be running a CPU that close to the edge.

 

[Seero]

Because even if it made 2 or 3 degrees difference no one should be running a CPU that close to the edge.

Why not?

 

[Rifter]

Because even if it made 2 or 3 degrees difference no one should be running a CPU that close to the edge.

why not, my 930 in the summer hits 80-90c under load when stress testing(prime95, intel burn, etc) which is within 10c of the limit. If i can lower than even a few degree's its worth it IMO.

 

[BonzaiDuck]

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.

As I said, the nickel -- close to copper on the galvanic table -- is indeed there to prevent corrosion. But it adds a significant increase to thermal resistance. they also imprint the CPU information on the nickel. But they're not going to worry about 5C improvement in load temperatures, because they make the chips to run at a stock setting. Even over-clocked, the 5C isn't that significant until you push the voltage and the clock-speed to a point where the adjustments give less in increments of speed. All these things -- printed in an Anandtech article on OC'ing the Yorkfield C2Q back in '07 or '08 -- are explained as exponential relationships. The OC'er is looking for the sharp bend in the curve.

When you throw in the additional margin for lapping the cooler, and then again a difference between TIM formulas, you have less heat stress on the processor whether you OC or not.

But the warranties are voided in the process. I've been building machines since the mid-'90s, and never had to return a processor. I'm quite sure that people had been returning them because they didn't get the "lucky VID" they were looking for. And that's the drawback on voiding the warranty -- lapping -- before you find out what that VID is, or whether or not you want to keep the chip.

As to how flat they are. I've seen IHS's that are uneven -- definitely not perfectly flat. And those CPUs are going to show some irregularities in their thermal range -- at least among the cores. This last puppy I lapped -- the i7-2600K -- my particular CPU was near perfectly flat to begin with. So lapping it took a little less effort to remove the nickel plate.

As to "whether you should do it" or "whether it's a good idea if 'it's your first time'" -- I can't say. It's not a lot of effort, and with a little care, not a lot of risk. This time, because these CPUs don't ship with the same protective plastic cover that can be used while lapping, I used a latex glove. You can either lap it dry, or use a few drops of water. If you use a few drops of water, you have to be mindful of the slurry of metal particles sticking to the IHS when you feel inclined to examine your progress -- you want to dab it with a cloth before you flip it right-side up . . . .

If you're not good with your hands, not comfortable with things like shop-tools and sandpaper, if you're "all thumbs" or crippled -- either get someone else to do it or don't do it.

It's not "that big a thing" -- really . . . .

 

[PlasmaBomb]

^ if it's your first time you could always look for a very cheap used cpu on an auction site to practice with...

 

[BonzaiDuck]

^ if it's your first time you could always look for a very cheap used cpu on an auction site to practice with...

Certainly -- that is a possibility. Some of us who build periodically and pass older machines on as hand-me-downs to extended-families build up a parts-locker of surplus "stuff" -- or "junk" when one goes through the occasional sorting and judgments over "worth-keeping." And putting that "junk" up for sale has costs, so weighing the costs and returns or even the chances that someone will actually buy, the best option may be a trip to the computer-recycler.

I've got an E2140, just decommissioned two weeks ago. I've got an old Northwood 3.0. Sitting on my desk -- a Conroe E6600, slated for an attempt to resurrect an old motherboard. In my case, they've already been lapped (except the Northwood . . . )

But for the life of me, it's hard to imagine how somebody would need to make a "trial run" with this sort of thing. You find a sturdy glass surface -- like a 5/8"-thick coffee-table top. You get some wet-or-dry sandpaper, and either tape it or us a temporary adhesive to secure it flatly to the glass surface. You optionally put a milliliter or two of tap water in a shot glass to drip a little occasionally on the sandpaper, or use an eye-dropper to put two drops on the sandpaper. You avoid touching the gold-contacts with your fingers or otherwise contaminating it. You keep a bottle of (preferably 99%, but 70% will do) isopropyl alcohol handy for clean-up. You gingerly put the processor cap/IHS against the sandpaper, and slide it back and forth. You check your progress by lifting the CPU off the sandpaper, holding it with the IHS face-down, dabbing it with a cotton rag, and flipping it over for inspection.

And you switch sandpaper to progress from one grit to another. I even do my best to "wash" the sandpaper of excess particulate by adding a little extra water and dabbing it with a (different) rag . . or paper-towel . . .

. . . And don't forget to turn on Cartoon Channel or the McLaughlin-and-Company-fool-pool to further your education while you do the lapping. You should be able to get this done within an episode or two of Dexter's Laboratory (or an episode of Dexter and the Ice-Truck-Killer), or by the time the Fool-Pool winds up with inane conclusions and weekly political pronouncements.

You don't need to remove every last trace of nickel. You just need to expose most of the copper underneath it. That's why you "check your progress . . . "

POST_SCRIPT: Waiting for delivery of an SATA-III Veloci-Raptor to use with an Intel Elm Crest SSD for "ISRT" on this Z68 mobo. Delivery -- scheduled for Friday. First pass at an OS (VISTA64) install on a Caviar-Black drive -- no problem. I didn't install the chipset drivers or anything additional to that. But ran the latest CPU-Z and CoreTemp with the room-ambient at 78F.

Coretemp shows fluctuation between 25C and 26C on all four cores at idle. Keep in mind -- idle doesn't mean much -- you have to do stress-testing for the load temperatures. But for that room-ambient, the idle temperatures can't get any lower. . . .