If you want a 16 channels (or more) system with resolution up to 720x480, you would need to buy PCI-E based capture cards, or a special motherboard. The standard PCI bandwidth is not sufficient for that high resolution at that many channels.
Another problem is recording. If you just need "2fps" and round-the-clock for each channel, it would be easy. A low-price dual-core CPU can do this.
But if you want very smooth (at least 25fps) recording for all 16 channels, one recommendation would be to buy one 8 channel PCI-based D1 hardware compression card, one PCI-E based software compression card (general support D1 resolution), and a quad-core CPU (for software compression). And then buy a software that can mix-using hybrid cards.