I think a lot of things late at night, but lately I've been wondering: What if you removed the heat spreader on a modern CPU so that the die(s) made direct contact with the heatsink, just as we all used to do back in the good old days? There could be an issue of excess weight on the die, but with a proper mounting system the heatsink should just be touching the die, not exerting force on it.

Ah, the trembling hands whenever I changed the heatsink on my Athlon XP! I don't miss those times I think since modern CPUs (especially Intel) tend to have the heatspreader soldered to the die, the risk is just too high and the gain is too little.

Besides - the heatsink mounting mechanisms are designed for the heatspreader in place - the pressure would probably be too light due to the missing milimeter.

Anyway, if you want to go ahead with it, google "IHS removal" and see what others did to mutilate their CPUs

A little how-to on IHS removal.
The IHS on most CPUs is soldered on. To remove it you have to cut the rubber and heat it to 85 Celsius.
Then you have to take the mounting bracket apart since you removed the extra height of the CPU.
Gently reseat your heatsink and, viola!
Temps should improve by 5-10 degrees.

Assuming the chip is RoHS compliant, this is probably not accurate. I did a bit of solder rework station design in college (~3 years ago), and most lead free solders were 210C - 230C reflow temperatures. This actually makes it a lot harder to remove a device without damaging it (or the PCB). I doubt a soldering iron/toaster oven would be a safe approach without a good way to check thermals.

The article may have been based on non-RoHS chips?
Edit: or the solder may be different, given that its electrical properties are not important?

And why would you risk messing up a perfectly working CPU just to get at best 5-10c advantage ? Is it not just simpler to change cooler or get a better fan or two ?
And if there is no pressure on die thermals will suffer, you need pressure from CPU cooler in order to have effective heat transfer. There are many heatsinks out there that put low pressure on heatspreader and those coolers usually suck big time. So one can put like this, not enough pressure=low performance but if you remove heat spreader that pressure can mess up that core. IMHO not worth bothering.

Hahaha you're talking to the wrong guy here. I only asked because the last time I did something like this was with an athlon 64 X2, in which the heat spreader was not soldered to the CPU die at all. I would have to check if newer processors are indeed soldered, as the thermal compound they use between the die and the heat spreader can behave strikingly like solder given enough pressure and heat from regular usage. Thus, it might be easier to remove it from an unused chip. Furthermore, if you know exactly what you're doing, the risk is negligible. Also, the amount of pressure does not matter as long as there is some. The arctic silver 5 or whatever you're using will take care of the rest.

You'd be surprised how much pressure those dies can take, anyways. Hell, back in the day I used to have to push down on the old athlon XP's with all my strength just to get the damn heatsink on with those terrible clips they used. That's much much more force than even a 1kg heatsink will exert, especially since the heatsink would be sticking out parallel to the ground, meaning gravity is not much of a factor.

I guess it's more a thermal paste than what you would call solder.

This is what I was leaning towards, but I can't think of anything lead free that would melt at 100C below the lead reflow temperature. I was thinking silver, but I looked it up and the melting point is actually quite high. Any guesses? (Or perhaps it is a gummy substance, and heat merely makes it easier to deform?)

Edit: I know! Mercury. That's not on RoHS, right?