Thanks!

For a full node I've been running them all on anything that has a decent SSD + 32GB of RAM and still retaining the entire chain in-memory (using /dev/shm or some other ramdisk method). The CPU's themselves don't seem incredibly important unless a ton of remote connections are established to it and you use them for haproxy/nginx. You might be able to get away with 24GB RAM, but it's not as future proof and within months it may not be enough at all.

If it's an incredibly low traffic full node you could probably get away with 8-16GB RAM and just use disk i/o. I have a few development environments setup to use lower amounts of RAM like that, but nothing production ready.