The transfer speed will be limited by the device. A good USB thumb drive will give you around 80-150MB/s on the read transfer with USB3 right now. When you look at what that was with USB2 it was, much, much slower. Most 7200RPM SATA-3 drives never make the 6Gbps specification at all. You would need to RAID those drives until you maxed out the controller at that throughput level. I still feel eSATA was not given a fair shake. On old laptop with the side miniport for PCI-Express cards, its possible to run eSATA or USB3 off of that type of port. These are no longer being built into the laptops because they have native USB3 support with the chipset. Today you still see 6Gbps SAS controllers linked up to server boards as if this is some amazing innovation, with the PCI-E cards often being an outrageous charge.
Either way. it is a close approximation of what you would get, if the drive was installed internally. But the technical specifications for eSATA were in fact faster. The other issue with eSATA was that it often required an external power source, which USB3, on newer devices, do not require. In fact, some of them will now stay on and continue charging. All sorts of rapid charging technology is coming from the implementation of USB3.
I found eSATA to be a better option, early on, simply because you are connecting what would be internal storage using the same technology that is used when you assemble a system. Its the same basic SATA cabling (granted there is the eSATA cabling, but the end connector is basically so that it doesn't fall out all the time like a normal SATA cable would on the inside of a system. The cables also need to be more shielded than internal cables.
eSATA was better than USB3 but it wasn't as easily as accessible. When people go to the store they do not want to hear about putting together their own external hard drive and all of these things - even though it could save them a lot of money. Otherwise no one would buy a Western Digital Passport USB3 drive. They would just buy a laptop hard disk themselves, the external enclosure, actually use a screw driver, and create the partition when they plug it in. The requirement of having an external power source was also a major problem for eSATA, but I still use it to diagnose hard disks, with a StarTech 4-drive that connects via eSATA and a KingWin EZ-Dock for one additional drive.
Better support for ESATA could easily have been built into laptop chipsets. Remember, the early implementation of eSATA ports was using PCI-E cards and cables that connected to real SATA on the inside of a workstation. In a lot of ways this was very similar to the implementation to SCSI on server platforms (Today, the idea of SCSI is becoming even more and more silly as people realize they can just buy SSDs for a much cheaper price, even industrial ones, and create a legitimate off-site backup plan). Some motherboards came out with the eSATA functionality, built into the chipset, but this is more or less a curiosity since USB3 has become widely popular.
Server technology still relies on the five 9 idea. The 99.999% uptime guarantee, which with SMBs and outside of datacenters, is mostly a myth. A small company that needs one domain controller will often find themselves spending $50,000 dollars on server technology for their office that uses outdated, expensive technology like SCSI, that still won't prevent them from a melt-down if the operating system is infiltrated or the system does not have credible backup. I have seen the Active Directory implementation done poorly and incorrectly, even after so much money was spent on a bunch of soon-to-be obsolete server hardware.
Today, I would fully endorse setting up a server that uses eSATA, USB3, redundant backups, and a server motherboard for SSD- not SCSI or SATA storage. It would be best if SCSI would go away completely... some how this is still a sticking point for large consulting firms, presumably due to the ridiculous cost associated with it.