Are you sure? I would imagine aluminum heat spreaders would have a higher rate of transfer of heat to the surrounding air than the ram chip material would. In other words, I think aluminum would give up heat more readily than the ram chip ceramic.
You are correct, but the ram chip itself must give up the heat before the aluminum can, so you're just adding another piece to the chain. Heatsinks do not work because they conduct heat better than CPU cores, they work because they offer vastly increased surface area (much more surface contacted by a cooler substance: air, generally). Heatspreaders do not add enough surface area to work as proper heatsinks do; any advantage offered by the 50-100% improvement in surface area is cancelled out by the disadvantage of having another step between the device and sweet, sweet cool air.
If you don't get what I mean by "another piece in the chain", imagine that you have a length of steel wire. Steel isn't very conductive, so you solder another length of silver wire onto the end. While silver is a lot more conductive than steel, the electricity must still pass through the steel before it passes through the silver, so in the end you're not helping anything-- in fact, you're ADDING electrical resistance. The same applies here: while aluminum is indeed much better at conducting heat to air than the material the chips are made out of, the chip must still conduct the heat to the aluminum before the aluminum can conduct the heat to the air, a procedure that is inefficient unless aluminum is perfectly conductive (which it isn't, obviously). And that's in an ideal scenario, where the aluminum contacts the chips as well as air contacts the aluminum; in reality, they do not make good contact, and as such there's a thermal pad between the two (which, naturally, is not very conductive either).
Add on top of all of that that memory, well, doesn't get very hot. If there's not a lot of difference between case air temperature and memory temperature, the best heatsink in the world isn't going to do much, and heatspreaders acting as heatsinks will do three fifths of sod all.
There are tests ("experimental evidence") that back this up. All differences between heatspreaders and no heatspreaders, whichever way they go, are well within the margin of error for those tests.
Any heatspreader isn't spreading heat out among the chips exactly, it is simply spreading out the heat as in increasing the surface area available to transfer heat to the surrounding area. I'm sure you know that, i just think it's a matter of semantics. (The heat isn't going out to the chips, they chips are the source of the heat)
Actually, a heatspreader IS spreading heat out among the chips. See, a single chip on a stick of RDRAM can put out a lot more heat than its neighbors over a non-trivial length of time. While the stick did not put out enough heat total to justify bothering a proper heatsink, a single chip could definitely cook itself under the right circumstances, so heatspreaders were added to ensure that heat would be distributed as evenly as possible.
Yes, I know that DDR-SDRAM != RDRAM. People apparently liked the look of heatspreaders on RDRAM, and they're cheap, so memory sellers started putting them on their own chips to make them look cooler.