I need to have maximum versatility and flexibility in my hardware setups for image processing, due to my pretty complex life and working situation. So no way I can use a single stationary machine, at least not exclusively.
Currently I do most of the processing stages on a powerful laptop, either on one of my Thinkpad W520 machines (both set up with Quad i7, 32 GB RAM, 2*512 SSD, 1 TB HDD), or lately, on a smaller Thinkpad X220 (i5, 16GB, 256 + 256 GB SSD) for lighter travels. On the road, they are used without any auxiliary monitors, whilst at the office they are docked to support 1-3 additional monitors to provide screen estate. In the office, there is a networked structure into which all laptops plug in to share the gathered data. All images are eventually stored on one or several networked nodes (Linux based, RAID-1, RAID-5, RAID-6, or RAID-10 systems) and backups done through a backup server cluster with duplicated LTO tape stations. In addition, there are networked boxes for accounting, image indexing and searching, data base servers etc. I'm working on making the utility servers virtualised and agglomerated on a massive Linux-based server. Plans call for 256 GB or more and at least quad CPUs.
Typically about 20-25 network nodes are active at any moment. Total storage capacity on the network now exceeds 200 TB.