SSDs on USB 3

How is SSD USB 3 performance?

I'm not referring to hitting the drive's max seq. speeds; I'm much more interested in whether or not a SSD will maintain its snapiness.

 

Do you mean as in running the OS or simply loading programs on it?

Honestly, i really don't know, but i could test it, i have a SSD that is currently not in a computer and a USB3.0 enclosure and a desktop with USB3 as well.

 

with a complete OS setup. think of it as taking the SSD out of your one-bay lappy and plugging it back in through the USB 3 port*.

*i am not literally doing this, but if it helps someone visualize what i mean....

EDIT: i don't really care about benchmarks, but someone else might. subjective evaluations are enough for me.

 

Well, i can't really do OS on USB 3.0, but i can certainly bench the drive though, should take about 10 minutes at most.

Bench now attached, in short, speeds look on par with SATA II for the most part aside from QD32 randoms and sequential read, but even those are still "fast" so i'd say most of the snappiness should be there. 128GB Crucial M4 by the way, benched on a P8P67 Pro desktop motherboard.

 

Just as an idea, if you have a one bay Lappy with an ODD, you might want to try putting the ODD in an external enclosure and get a caddy for the HDD, put it in the ODD bay, and put the SSD in the HDD bay. You will get the most out of your SSD, storage from the HDD and unless you use the ODD very often, you won't miss it. And you won't see a drop in speed of the ODD unless you have a really old laptop.

 

That is what i would personally do, but the OP asked about SSD performance on USB3.0 so i got my SSD in an enclosure and benched away. The OP may have a pretty valid reason to want an OS on a portable device though, i have a usb with gparted and one with parted magic for troubleshooting stuff if i ever need to. I also successfully got ubuntu on a flash drive for the sake of trying it too.

 

It will. The drive's interface has no noticeable impact on the access times or "snapiness" of the drive, whether the drive is an SSD, 7200rpm, 5400rpm, etc.

I have been using external drives often for quite awhile, going back to 2007 when I had to run two virtual machines off a usb 2.0 drive. My current full time client requires me to use their desktop, which of course has an archaic single platter 500GB drive in it, so to bathe in SSD goodness I do my development off an external Intel 320 series 120GB SSD.

Since I have so often ran development virtual machines of external drives, I've studied and benchmarked numerous configurations and here are some of my lessons learned.

In my experience, running virtual machines is maybe the best real-life usage test of how a storage device performs. Booting, shutting down, pausing and resuming virtual machines thoroughly stress the interface bandwidth like nothing else, whereas once the virtual machine has started, most other actions (other then those mentioned previously) are heavily dependant on access times (snapiness). There are some exceptions of course - compiling in Visual Studio .net does a lot of read/write and both bandwidth and access times are important, but even in this scenario the bandwidth is not quite as import.

The interface, usb 2.0 or 3.0, Sata I, II or III, has no significant impact on access times or the "snapiness" of a drive. In other words, once a virtual machine is booted, while working in it, it is almost impossible to notice the difference between the various interfaces, but the access times are noticeable, especially when the VM and the host share a mechanical drive as it thrashes like no tomorrow.

When it becomes time to shut down or pause the virtual machine, then the interface bandwidth is a huge factor. On the Intel SSD mentioned above, when I run it off a USB 2.0 port, once the machine is booted, it performs very well even though it is supposedly "bottlenecked" by the usb 2.0 interface. However, pausing, resuming or attempting to shutdown takes forever, more than one minute while on a Sata II cable it is around 10 seconds.

I use snapshots often, and delete snapshots often (again, this is a bandwidth intensive operation and relates directly to boot/shutdown speed of the VM on the drive), and here is the pattern I typically saw on deleting snapshots off Sata connections: A single 7200 rpm drive would take well over 30 seconds to delete a snapshot, A pair of striped 7200 rpm drives would delete a snapshot in around 20 seconds. A pair of velociraptors would delete the snapshot even faster, 10-15 secs iirc. An SSD (Samsung 830, Intel 320 or Crucial C300) would all delete a snapshot in about five seconds off a Sata II port. On a Sata III port a Samsung 830 does it almost instantly.

My last comment, not all USB 3.0 ports are created equal. My toshiba r835-p56x portege has a slightly higher bandwidth ceiling on its eSata port than on the USB 3.0. However, the USB 3.0 hub I have plugged into the back panel of my desktop (Asrock pro4-M) does much better than eSata with an Intel 320 SSD.

I hope this info is useful to whomever.

 

i was very curious as to whether or not the lack of NCQ woud affect the drive.

 

most USB 3.0->SATA enclosures have a USB3.0->SATA II adapter... so sata 2 speeds

USB 3.0 adds more latency than "regular" SATA

I would expect it to be fine for moving files / VM, but not full speed

 

Without NCQ an SSD cant reach peak performance, but its not like you can get peak performance on USB3.0 anyway. Im more curious about TRIM.