Test suggests USB3 is almost as good as TB for external SSD

[Vghty]

My MacBook Pro Retina has a pretty large SSD (768GB) but for my work I often deal with very large files. So I wanted to have some portable external storage too, both for copying as well as for direct access by software applications of various types (including database access).

Being exited about Thunderbolt, I was wondering whether bus-powered TB drives would be preferable over USB3 enclosures, even if they are much more expensive?

I took an OCZ Vertex-2 240GB SSD that came from a previous computer and tested in in several enclosures I had available. I also compared these to the internal SSD of the MBP itself. These are the results (for whatever its worth, I’m not a professional tester, but want to see myself what works best for me):

My tentative conclusions:

• The Lacie Rugged in TB mode is hardly faster than USB3. This is true for cheap USB3 enclosures as well as for the Lacie’s own USB3 mode. The enclosure is well crafted and sturdy, but also big and heavy – adding almost 20% of the laptop’s weight in to bag.
  
• The cheapest USB3 enclosure I tried, called “Diginote USB3” had a very good performance but also is very portable and solid made.
  
• One of the cheap USB3 enclosures I tried (König CMP-MOBSTORE96) had an abnormal slow performance. It also required an unusual cable (USB3 A-to-A).
  
• Not surprisingly, the only HDD I used compares bad to the SSD. However, in the real life copy of a 900MB file he did quite OK, confirming me that such drives are still pretty well for backup purposes.

So, I’m going to stick to the Diginote enclosure for a while!

Now, my test certainly has limitations. I did not consider Native Command Queuing, which I believe USB3 has not, and which the TB configuration I used might have (not checked). And perhaps all sort of other things.

Anyway, hope this information is of some use of others...

 

[throAU]

Try running 2 USB3 transfers at the same time, and take a look at your CPU utilisation whilst testing both and compare.


The 10GB vs 5GB speed thing is only a tiny part of the USB vs thunderbolt performance issue.


However if all you need to bulk storage to use occasionally, USB3 is likely by far the better bet due to cost. I run USB external drives myself for this reason.

 

[Weaselboy]

Being exited about Thunderbolt, I was wondering whether bus-powered TB drives would be preferable over USB3 enclosures, even if they are much more expensive?

Nice write up. Thanks

Bare Feats just did a similar test with the same results you had. Makes it hard to justify the added cost of a TB interface for these drives.

 

[duervo]

The Vertex 2 drive is a SATA2 unit. The USB3 and TB rates you're getting are in line with what you should expect from a SATA2 drive.

I would be curious to try the same tests with a SATA3 drive. The single drive will probably still be the bottleneck. You'd most likely have to start doing RAID to see the benefits of TB for speed. Daisy-chaining is also another benefit you don't get with USB3.

However, if a single drive setup is all you need, then it's practically a no-brainier, based on cost, to choose USB3 over TB.

 

[alphaod]

I use a SATA III SSD in a combo Thunderbolt/USB 3.0 enclosure.

Anything that leverages the sequential performance of my SSD, I find Thunderbolt to be significantly faster than USB 3.0; however for day-to-day use (random access), I find USB 3.0 is fast enough.

 

[Vghty]

Thanks throAU and duervo. You raised valid points, and indeed I have no doubt that for a large RAID set-up like the Pegagus stacks, TB will outperform USB3.

But the little test that I did was purely for the intended application I described: having an external, very portable drive using an SSD unit I already had lying around.

Having that said, whether even a modern SATA-3 drive would be faster in this particular LaCie enclosure, might also depend on the particular controller/drive protocol that LaCie adopted for this device.

Perhaps other people have posted something on this, or perhaps the LaCie factory configuration with 128 or 256 SSD have SATA-3 drives - don't know, I don't have any of these.

 

[leman]

Random benchmarks score TB: 936
Random benchmarks score USB3: 570

That is 60% difference! And random access is what makes SSDs feel fast in the first place. Sequential tests are only interesting if you are copying large amounts of data which also happen to be well-ordered on the disk. So basically never.

Although, I agree with the overall conclusion - USB3 is a much more cost-effective solution for usual things like backups or external data storage. However, if I were to use my OS from an external drive, I would certainly go TB.

 

[Drmash]

Thermaltake Raid 0

Hardware - Samsung 840 250gb x2
Thermaltake USB 3 cases
1. BlacX 5G
1. SilverRiver 5G

I ordered 2 SilverRiver 5G's, but just like other reviews they are poorly made and one was DOA. FYI Newegg doesn't pay return shipping for DOA and S/H will be 10.95. Amazon is the only consumer friendly place I have found, if others know places that pay return shipping for defective items let me know.

So I used what I had to post speed test.

Blackmagic speed Raid 0 - 5GB 500gb W and 594 R
Xbench Raid 0 - 539.5

Single Drive test
SilverRiver 5G - 250GB W and 427 R
BlacX 5G - 266.3 W and 387 R


I picked the Samsung 840's up at Amazon for 150 dollars each, but they would sold out of the Silver River's. I paid 60 dollars, + 10 S/H to return the DOA item. so $370 dollars for 500GB SSD. Regardless if I had 1 drive or 2 RAID 0 I would be using a secure backup and online. So I'm not too worried about increased risk of Raid 0.

I don't know of any 500gb SSD with these speeds at the price I paid.


Thanks

 

[KPOM]

My own experience with USB 3.0 enclosures and the Seagate Thunderbolt adapter is that the USB enclosure is slightly better with the HDDs I have tried it with, and the Seagate Thunderbolt adapter is slightly better with the SSD. The differences are not noticeable, though (less than 5% in terms of throughput).

 

[Freyqq]

The Vertex 2 drive is a SATA2 unit. The USB3 and TB rates you're getting are in line with what you should expect from a SATA2 drive.

I would be curious to try the same tests with a SATA3 drive. The single drive will probably still be the bottleneck. You'd most likely have to start doing RAID to see the benefits of TB for speed. Daisy-chaining is also another benefit you don't get with USB3.

However, if a single drive setup is all you need, then it's practically a no-brainier, based on cost, to choose USB3 over TB.

This. USB3 realistic speeds run at about the same as SATA II. Therefore, there isn't likely to be a huge difference. SATA III SSDs would should very different numbers.

 

[Erasmus]

As always, USB will always be better for consumer level stuff than the professional equivalent, and vice versa. Clearly your typical "consumer" will rather choose a cheaper, more universal connection, which is fast enough for simple backups and extra storage.

Clearly your typical "professional" will choose the faster connection for their double-redundant external RAID array, as not only would their setup actually benefit from the extra speed, but what they do demands it as well.

So, as it has always been; USB is for a general, simple, cheap connection. TB is for a super fast connection that comes with a price premium to match.