Oh I see what you did there, you compared the card I used for my 16 drive example and to the 4 port USB PCIe card and USB bridgeboard.... okay. Thanks for stacking the deck, you just compared an enterprise-level SAS controller capable of 64Gbps to a bunch of USB3.0 toys instead of the obvious eSATA example I offered. Let's examine this both ways:
In order to get a 16 drive RAID over USB you would have to... oh wait you can't, USB3.0 is only 5GBps per port. So unless you don't care about bandwidth, that would require you to have FOUR of those USB3.0 PCIe cards to get your 16 USB 3.0 ports, run them JBOD, and use Software RAID in OS X. Even then, you'd be throttling them to 5Gbps per drive from SATA III's possible 6GBps. What an insanely idiotic setup; why, WHY would you do that?! By the way, total cost? $400 for the 16 bridgeboards and $560 for the four USB cards for a total of $960. Oh, and you haven't even found a case or power supply to put them in, also good luck finding a motherboard with that many PCIe slots. So does that $470 SAS PCIe card look so expensive now?
You should've compared your $140 4 port USB card PLUS $25 per drive for the bridgeboard to a $100 4 port eSATA card with greater throughput (4 x eSATA III = 24Gbit/s Vs only 4x USB3 = 20Gbit/s). Also there no bridgeboard required with eSATA, so therefore it is cheaper with less latency. Then you can tell me why you're using latency-prone, bottleneck-creating bridgeboards in a setup and paying $140 more for it ($40 more for the PCIe card plus $100 for four USB Bridgeboards).
Your setup is bad, and you should feel bad!
USB3 is not a replacement for SATA any more than Thunderbolt is a replacement for PCIe. Slower, requires more equipment, more expensive, period.
Hehe, yes, I feel soooo bad.
Both ways have their advantages and if a user needs an enterprise-level SAS controller capable of 64Gbps then the $600 card is superior. As far as why I selected the $140 HighPoint USB3 card is because it has a dedicated controller for each of it's 4 ports - and it supports SuperSpeed on each. I haven't seen a cheaper Mac compatible "PC Card" that has that. The best I've seen is 2-ports per controller.
Sure if you're buying the very fastest SSD drives to populate your JBOD or RAID then you will see a little clipping in benchmark utilities over the USB3 setup. SATA3 tops out at 2.4GB/s (probably 2.2 GB/s in reality) and the USB setup tops out at 2GB/s (probably 1.8 GB/s in the real world). But both of those "real world" speeds are mostly only achievable in benchmark utilities - a fact most people never stop to consider.
Even the most aggressive video
editing software with the fastest CPUs with the most cores will almost never use those (2.2 ~ 1.8 GB/s) respective bandwidths - closer to half that is about right. If we could consider only those times during access I guess the breakdown would look something like:
80% of the time: Under 300 MB/s
18% of the time: Under 600 MB/s
1.5% of the time: Under 1 GB/s
0.499% of the time: Under 1.5 GB/s
0.001% of the time: Under 2 GB/s (The USB3 setup will be slower here)
Interface and connection latency would be important if we were timing boot speeds or massive database I/O but for video, photography, or really any other kind of content creation it's not very consequential.
When you also consider that the user might be using rotational media then there becomes absolutely no differences between the two configurations. And if you take into account diversity where the user might wish to use two or three individual 4-drive RAID boxes leaving the other USB3 ports for memory card readers, audio interfaces, 1080p monitors, 2G ethernet interfaces, USB2/3 hubs, TV tuners, Video recorders, and so on, then there's a little extra advantage in the USB3 setup.
Right, Even that youtube guys TB box was only doing just slightly over single connection USB3 speeds and when he went to RAID10 it was WELL under! Someone should tell him: If you're going to use RAID10 like that then you can skip the expensive TB enclosure and just use USB3 for the same identical performance.
