Sign in with
Sign up | Sign in
Your question

USB3 Bus Saturation, Throughput performance of multiple drives

Last response: in Storage
Share
January 26, 2011 5:41:22 PM

OK, this is a bit complicated but here is the scenario. Under USB2, I would see (predictably) slower write performance per disk as I added more drives to a USB2 hub (the drives being SATA converted to USB2 via bridge adpaters). For example, one drive could sustain 35MB/s, two drives about 19-20/MBs, three drives about 13/MBs. Assuming the drives had similar characteristics and were performing basically the same operation. It was pretty evident that the bus was the limiting factor here.

Now we have USB3, with its advertised 10x increase in bandwidth. Soooo, one would think that if I ran the same test I would get different results.

I installed a USB3.0 dual port card in my Dual Processor 6GB Xeon Supermicro motherboard. From the card I cabled out to a 4 port USB3.0 hub (using a USB 3.0 cable). I plugged all three drives in (still converted to USB2 via bridge adapters) and guess what?

I get exactly the same result.

Why is this happening? I expected the new bus to allow all three drives to operate at 35MBs and instead I see no improvement whatsoever.

Am I missing something in this equation?

Yes the drives themselves are still USB2 devices but I dont see where that has an impact when the heavy lifting is being done by the USB3 hardware and a well equipped motherboard.

Any help is greatly appreciated

a b G Storage
January 26, 2011 6:02:06 PM

Proably because the USB hub can't buffer the data, so it effectiely runs at the speed of the slowest active device.
m
0
l
January 26, 2011 6:13:51 PM

MrLinux said:
Proably because the USB hub can't buffer the data, so it effectiely runs at the speed of the slowest active device.


If USB 3 can run anywhere near its claimed speed that seems unlikely, the hub should be designed to buffer data back to the card at speeds vastly exceeding those of these old IDE drives. USB3 was designed to be backward compatible with USB2 but if that means the whole system drops to the lowest common denominator then its a pretty worthless standard.

Also the degradation is almost linear. The devices are identical, so there is no slowest active device. As more identical devices are added, every device slows down.
m
0
l
February 10, 2012 6:56:07 PM

Actually, I'll bet your problem is that your USB card has different controllers for USB 3.0 and USB 2.0 devices, with the latter subject to the same limitations as your previous USB 2.0 connection.
m
0
l
!