However, it looks like this was too small a step for most industry players, so PCIe appeared, which is a serial bus (unlike PCI and PCIX). Instead of having 32 or 64 parallel lines at relatively low clock rates, they choose to go with serial (1 bit wide) at much higher clock rate. This is a general trend in the industry, same thing with SATA, for example.
PCIe x1 means just one serial channel, while PICe x16 has 16 serial channels, and should therefore have 16 times the throughput. The reason this still works better than parallel is that in 16x, each serial channel has its own clock, so skew is not a problem like it was in the parallel configuration (when all the data lines shared a single clock).