Are USB 3.0 drives and peripherals CRAP?

[nemophoto]

I'm curious whether anyone has actually had decent success or performance with USB 3.0. I've had several USB 3.0 drives and find that, at least on my system, there is very little real world improvement over USB 2.0. I still find the "true" high performance external drives are eSATA. I did a copy test from an Iomega 500GB USB3 pocket drive to a WD My Book 1TB USB3. The folder was a simple drag-and-drop of 18GBs of 1Ds3 RAW+JPEG images in about six subfolders. The best I could manage was 6-7 MB/sec. Then, I did the same copy to a 500GB G-Tech eSATA drive. I averaged 30 MB/sec. At first I thought it might be a driver issue, but I have the most recent one.

Any thoughts? I have an AMD Phenom x6 1100T and 16GB of fast RAM, a Gigabyte motherboard.

Nemo

[kaelaria]

You can't compare performance like that, you were reading and writing through the USB bus at the same time on your first test.  Copy from your HD to both to test. 

[JeffKohn]

At the risk of asking the obvious, you do have a USB3 controller on your motherboard?

I haven't used any external USB3 hard drives, since I have a eSata ThermalTake Black-X duo for my external drive needs. But I did recently purchase a Delkin USB3 card-reader, and it can download from a UDMA CF card at sustained speeds of 75-80MB/s, which is considerably faster than any card reader I've previously used (including FW-800 and ExpressCard). So I'd say yeah, the performance benefits of USB3 are real.

[StuartOnline]

I gave USB 3 try with my current Macbook Pro 2.66 (June 2009) 15".
Purchased LaCie USB 3.0 ExpressCard34.
My main two issues where that it would not stay mounted (connected) and it only with work with LaCie products.
Had a number of communications with LaCie about this product, however never got anywhere.
What really got me was less then two months after I purchased this device they discontinued it. Go figure.

Have now just stuck with Firewire 800 drives using the Nitro AV 8 port hub:
http://www.nitroav.com/product/113/ that is connect to my Macbook Pro.
This allows me to connect 8 Firewire 800 drives simultaneously.
Have never had the first problem with this hub.
Also had no problems with the hub after upgrading to OS x Lion.

Stu

[nemophoto]

Thanks for the various replies.

1) I believe it was a valid test because I was using both USB 3.0 ports as well as the USB 2.0, in addition to the eSATA. Regardless of port (which are all built-in -- no add-in cards), copies are managed by a combination of OS, CPU and bus.

2) Not so obvious a question, but, yes, the USB 3.0 ports are part of the motherboard.

3) I had/have a Delkin USB 3.0 reader. I had to send it back for replacement because the first was faulty. I've never had high copy speeds, though, with it. (I have a few 600x and 675x CF cards.)

4) Sorry about your Mac. I've heard Apple has NO plans to support it since they are pushing Thunderbolt, Lightspeed or whatever the hell they're calling the latest connection. I have Firewire 800 ports on my computer (add-in card), but have never really had spectacular speed from it. I actually have a Gen2 Drobo attached to it, but never manage to get over about 14 MB/sec. Actually, I did my copy test to that as well. Pretty much same dismal copy speed as USB 2 & 3.

I checked, and my NEC USB 3.0 driver is the most recent. I final got a reply from WD. Sort of vague, but in looking at some links they sent, the problem may be related to the chipset/driver from NEC and CPU clock speeds and memory timing. I'll have to sort the info out, but if I read correctly, the USB 3 does not function properly when you have memory clocked at what's called XMP speed (which is where my memory is set -- because that's it's rating), and also my Phenom x6 1100T is lightly overclocked. I'm not a game player, but those things do help Photoshop processing. Thanks for the input.

[DeanChriss]

FWIW, I did a bunch of real world testing a while back. I forget the numbers, but USB 3.0 was very roughly twice as fast as USB 2.0 for large transfers between non-RAID mechanical hard drives. The transfer rates were nearly identical to the data rates between two SATA III drives on a SATA III interface, which incidentally is the same as on a SATA II interface, because the mechanical hard drive itself is the bottleneck. If you're using SSD or a fast version of RAID, you'll see more throughput in all of these cases. the only case where the interface is the bottleneck is USB 2.0.

[mediumcool]

AFAIK, USB is still host-dependent (as distinct from FireWire which has long had the ability to connect suitable non-computer devices together, such as a FW movie camera and recorder).

Assuming TB is fast in part because of the additional circuitry (even in the cables!).

So I think that USB will always be behind the 8-ball for sheer speed. But, cheap, yes.

[fike]

I agree that testing USB3 throughput with two USB3 devices on the same USB3 bus is NOT a valid test. Go from a SATA device to a USB3 and you will notice a substantial improvement.

I have used a USB3 hard drive with substantial performance improvements, though I did not benchmark.  Copying a day's worth of photos from my hard drive to my USB drive took hours, now it takes less than one hour. 

As for all the compatibility and technical problems mentioned, I have not had any issues.

[kaelaria]

The entire USB pipelines are shared.  It's not a valid test when reading and writing to two USB devices unless they are on separate controllers.

[PierreVandevenne]

As much as I dislike USB3 and love lightpeak (for architecture reasons), this is not a valid throughput test. But you have put your finger precisely on what's wrong with USB3 in practice for most users. In addition to having bi-directional transfers, you have bi-directional file system queries. Try the same test with smaller files and you are likely to be even more disappointed.

The Promise stuff may seem overpriced, but I have seen a demo unit and I was amazed. I have a couple of fairly recent Netgear Readynas, and my work isn't time critical so I won't upgrade. But with the right disks, Lighpeak is close to an order of magnitude faster than the Readynas. A very good price performance ratio.

[darylgo]

I have had problems with the connector from computer to drive enclosure to the point that I now don't touch the drive once it is mounted on the mac.  Part of the problem was solved by using a different mac usb port, but the poor connector on the drive side is the real issue. 

[mediumcool]

I have had problems with the connector from computer to drive enclosure to the point that I now don't touch the drive once it is mounted on the mac.  Part of the problem was solved by using a different mac usb port, but the poor connector on the drive side is the real issue. 

One of my biggest gripe about USB is the flimsy connector (talking about the full-size one here—the little ones seem to be even frailer); unlike FW400 and FW800, the pressed-tin nature of USB connectors loosen over time, and the hard-to-orient sameness of the connector is annoying. The FW400 connector is brilliantly designed and built, with the FW800 a bit behind to my taste. But both way more rugged than USB.

And my mini-Displayport plugs are going strong—such a solid connector for its small size.

[John.Murray]

I'm seeing much better results with USB 3.0 with some caveats:

 1) As others have pointed out, the USB Bus itself is a shared resource, transferring data between devices on the same bus will make evaluation impossible
 2) That said, USB 3.0 is supposed to be bi-directional, given that, transfers between USB 3 devices on the same bus should be improved - I have yet to see that in practice

I'm also not impressed with the USB  connector, basically printed copper traces, but "printing" makes the technology incredibly affordable.  I've found cleanliness to be key - dirty connectors are going to rapidly degrade and destroy those traces.