1. Vise. I bought this one at Home Depot yesterday: LINK
Just clamp it to a sturdy table. You could always get a better vise, but this one was just fine for the job. The important thing is to clamp it to a sturdy surface. I used my desk, but a dining table, work bench, kitchen counter, etc. would work.
2. Coollaboratory Liquid Ultra - LINK
You might be able to find it cheaper elsewhere, but I just went ahead with Amazon for convenience.
3. Regular old hammer.
4. Solid block of hard wood. This is to press against the PCB as a buffer between the hammer and the chip.
Optional: latex gloves to avoid finger prints/oils (i didnt bother with this, I've just seen people use them in videos).
The TIM is very important. Apparently there's no other TIM that really comes close to the performance of Coollaboratory Liquid Pro/Ultra except solder. It behaves like solder (beads up, wants to bind to other metal) except it remains a liquid without the need for heat. I saw a performance chart of delidded CPUs, and someone who decided to use AS5 on die and on the IHS only got a 7C reduction (which is still significant, but clearly the liquid ultra/pro is much better.
Don't use that stuff on anything but the die, though. Otherwise, one day when you remove your block you're going to need some sandpaper and metal cleaner to get it off the copper. And it straight-up eats into aluminum. So yeah, Liquid Ultra on the die, and your favorite thermal paste between the IHS and your block. In the grand scheme of things, that won't matter as much. AS5, Noctua, MX4, etc.
CPU clamped by the upper perimeter of the IHS. Just stuck some clothes beside it in case the CPU were to fly off while hammering (it didn't).
Since I of course approached this cautiously, it took a couple of tries to secure the CPU. I wasn't tightening it hard enough and it would come loose after a few hits.
The initial piece of wood I was using. A stake that had been outside. It was too soft, and the PCB was actually splitting the wood as I hammered. I then wrapped the strike point of the wood with duct tape to try to prevent the PCB from wedging between the wood. The PCB ripped through the duct tape.
That should alleviate any concerns you might have about how strong the PCB is against that kind of shock.
The much harder furniture wood I switched to. Once I started using this, it took two solid hits with a hammer. (hint: you probably won't hammer hard enough at first either). Notice the damage that the PCB did to this wood as well:
Freshly delidded, shitty TIM still intact. Not surprisingly, the TIM on the die had a similar look and texture/viscosity to the pre-applied TIM on a stock intel cooler.
Delidded, black glue scraped off with a credit card, die and PCB cleaned with alcohol. Scraping the glue was the longest part of the job, but still only took a few minutes. You can press pretty hard on the PCB with a credit card.
Pic 6: Scary moment with the razor.
Before I found that harder wood, I was starting to think that maybe Intel started using more or stronger glue in the later batches to prevent or discourage delidding. So I got my xacto knife kit and started working the corners of the IHS to try to loosen it up. Bad idea. I immediately stopped when I realized that as I was pivoting the blade to work it under the corner, I shaved an ever-so-thin sliver of PCB.
Once I'd finished delidding and cleaning, I put the PCB under a digital microscope to assess the damage. In my minute or two being an idiot with a razor I'd managed to expose a tiny tiny dot of copper tracer (could only be seen with a magnifying glass or microscope). I've seen delids with far worse damage that still worked fine, but that doesn't mean much. The important thing was that I hadn't cut into it vertically. I was hopeful but still anxious.
Pic 7: Coollaboratory liquid ultra applied. If that were ordinary thermal paste, you'd say that's the worst application of TIM ever. But this stuff is different. As I said, it behaves like solder that won't harden. Those little lumps go away when it forms a cohesive bond with the IHS. They don't give you much liquid ultra to work with, but it doesn't take much at all. Less than other TIM because it spreads thin. I also applied an even thinner layer to the underside of the IHS.
Also, this is where a Haswell delidding takes a little more care and precision than an Ivy Bridge delidding, but it's nothing to really stress about. Notice the exposed transistors to the right of the die. Obviously, these aren't present on Ivy Bridge CPUs. If you get that liquid ultra on those, you're fricked. But, the liquid ultra doesn't ooze, run, seep out, etc. So don't spread it like a clumsy gorilla and you'll be fine. I saw some posts on OCN where people were so paranoid that they were recommending you cover them with nail varnish or a non-conductive TIM. I didn't do anything with them. The liquid ultra comes with a small brush for application and spreading. I ended up using one of my artist brushes with a finer tip to get the edges of the die.
Then I reseated the CPU, gently placed the IHS on top of it as precisely as possible(didn't reglue), and clamped it down as normal.