The massive amount of data recorded by dense electrode arrays which are routinely connected to Nyquist-sampling signal conditioning blocks introduces new design challenges for implantable and wireless biological signal acquisition. Five different architectures of implantable multichannel neural recording systems are compared in terms of power and area constraints. Silicon results of a 16-channel spatial-domain compressive recording system implemented in a UMC 0.18 μm CMOS technology are presented. Applying intracranially recorded EEG signals, the proposed system achieves up to 16-times compression rate, consuming an extra compression power of 0.95 μW within a die area of 0.008 mm² per channel.