HomeHardware


USB 3 performance
Ken in Regina
I recently started to recycle some of my computer technology to make room for some new toys. I've given a couple of desktop systems to my daughter's family for the grandkids to use. The last system I sent over on Sunday had an SSD drive in it that I wanted to keep, so I swapped it for a spare hard drive I had laying around.

While I was at it - crawling around under the desk tracing and retrieving cables and such - I decided to freshen up the main tower system by adding a USB 3.0 interface. At the same time I bought a little USB 3.0 case for the SSD I saved out of the other system. This all makes backing up the system drive on the main tower a much faster process when using an external USB drive. Data transfer is so much faster with the USB 3.0 interface than the USB 2 interface!

I did a test transfer both directions between the SSD in the main tower (Samsung 840 EVO) and the spare SSD (Samsung 840) in the USB 3.0 case on the USB 3.0 port. The transfer speed was identical in both directions (125MB/sec).

That was using the Copy function, dragging and dropping a 450MB file from one drive to the other and clicking on the More Details dropdown in the Copy dialog window. Hardly scientific but it gives a good basis for comparison of relative performance. And it's real world performance, not an artificial benchmark.

Using that same test I found that there are significant differences in USB flash drive performance. I have a 16GB Kingston Data Traveler that I was using to compare the difference between my old USB 2 hub and the new USB 3 interface. My first tests were discouraging.

I'm using hubs because all of the USB interfaces on the main tower are at the back and I have no interest in fighting dust bunnies and cobwebs under the desk every time I want to use a USB device.

Initially I connected the new USB 3 hub to the new USB 3 interface with a 6 ft. USB extender cable I had laying around because the cable on the new hub is only 1.5 ft. long and won't reach the back of the tower. Using the old extender cable turned out to be a mistake.

My initial testing showed that the performance of the Kingston flash drive, which is supposed to be USB 3, was virtually identical using either hub.

In order to eliminate the hub and the extension cable I crawled under the desk and stuffed the flash drive directly into one of the ports on the new USB 3 card. The read speed improved immensely!! It nearly quadrupled.

I have a short extender cable that is specifically for USB 3.0 and it's just long enough to get the USB 3 hub out where I can use it. With the Kingston flash drive plugged into the USB 3 hub on the short USB 3 extender cable the performance was identical to when it we plugged directly into the USB 3 card.

I was really impressed with the read performance of the Kingston flash drive once I got the connections all sorted out. The write speed was another story -- and that's why I decided to write this. The write speed on this particular flash drive (which is sold as USB 3) is no better than a good USB 2 flash drive. The performance is the same whether I test it in the USB 2 hub or the USB 3 hub. I also tested it against a good quality USB 2 flash drive I have and the write performance is identical in either hub.

I checked the specs on Kingston's web site and it was illuminating. This particular model comes in 8GB, 16GB, 32GB, 64GB and 128GB sizes. The product documentation lists a write speed specification for the 32GB and larger models but it does not specify the write speed for the two smaller versions. Now I know why!

This flash drive is only half USB 3. It's a rocket when you read stuff from it. When you have to write anything it's no better than a good quality USB 2 flash drive.

So, two things to take away from this. If you are going to use USB 3 devices on a USB 3 interface and have to use an extender cable and/or a hub to make it practical to use, make sure all the stuff is guaranteed to be USB 3 capable. Otherwise you will only get USB 2 performance.

If you are buying a USB 3 flash drive and the write performance is important, you will need to do some research before purchasing. It looks like read speed is fast on almost any USB 3 product but write speeds are a completely different issue.

...ken...
Boyd
So what were the read write speeds for the drives you tested? Did I miss them? The only speed I see that you mentioned was 125MB/sec for the SSDs.

I have always been underwhelmed by flash drive and SD card performance, they're well below the speeds of external hard drives in my experience.

I just clocked a 3TB USB 3.0 external drive hard drive on my Mac Mini at 150MB/sec. I use the Mini as a media server and the external drive contains over 1TB of music and movies. Before I loaded that drive up with media I clocked it at closer to 170MB/sec.

I get a little over 100MB/sec with a cheap bus-powered USB 3.0 WD Passport hard drive on my macs. The USB 2.0 version of that same drive only clocks around 35MB/sec. I have never really measured them, but it seems that these same drives perform a bit faster on my Windows 7 based PC. My guess is that this is due to the Mac's journaled filesystem which trades some speed for reliability.

The internal 512gb SSD on my MacBook Air clocks about 700MB/sec write and over 700MB/sec read - it really flies. But the newest model MacBook Air evidently delivers over 1GB/sec performance which boggles my mind.

I use a free program from the video company BlackMagic Design on the Mac that is pretty standard for benchmarks. It just continually writes/reads a 1gb file.
Attached Images
speed.jpg  
Ken in Regina
I probably shouldn't have posted the one speed I did. As I described, it was not a benchmark. The exact numbers I saw were relevant to my specific setup and nothing else.

Based on my experience (lots) with benchmark programs I would guesstimate benchmark numbers would be as much as 3 to 5 times what Windows Copy function reports.

And each benchmark program does things enough differently that they aren't directly comparable. Unless I could get the exact program you used it wouldn't be worth comparing numbers. As it happens I looked on Blackmagic's website earlier today and could find no mention of the disk testing program that gets such good mention everywhere you look.

I would be willing to post some numbers if someone can persuade me they will serve some useful purpose. Or if you can point me to a Windows version of the program you used I'll be happy to run some tests with it.

...ken...
tcassidy
I got the feeling this post was more about how you have to use USB3 certified cables and hubs to actually see real USB3 speeds. You see so much information about various cables for speakers, HDMI etc. as being better when they aren't, it is nice to know that some things are actually as advertised. Thanks for confirming it!

Terry
Boyd
Thanks Ken. I guess I got a little "lost in the weeds" in your post since it was titled "USB 3 performance" and I couldn't find any actual performance stats.

The Blackmagic test is available for the Mac at the App Store. I found this with a Google search, downloaded the installer from the Google Drive link in the comments and it worked fine on my Windows 7 machine. It clocks my 1TB internal hard drive around 150MB/sec

https://youtu.be/69h3Pt_dzVM

https://drive.google.com/file/d/0B2-HjvKcdFPZMm9jR1kyekllMkE/view


I find it useful for comparing different drives to each other, not as an absolute performance benchmark.
Ken in Regina
Mostly my fault, Boyd. I was poorly focused when I wrote that. It was sort of stream-of-consciousness. I was just finished the troubleshooting process and wanted to share what I discovered while it was fresh. That's not always a smart thing for me to do because I don't get it very well organized. Terry is right that the main thing I wanted to share is how critical the individual pieces are to getting full USB 3 performance. The other discovery about flash drive performance was an accidental sidenote that I also wanted to share but did so in a way that made it a distraction.

I'll see if I can get that benchmark to work and run a few numbers to see what it says now that I have the USB 3 hub functioning at what should be full speed.

...ken...
Ken in Regina
Okay, here are some benchmarks using BlackMagic Disk Test 1.0.

Marvin please note that I used the word "map" at least twice in this post to ensure relevance.

The system is an old home built Intel system:
Windows 7 Ultimate SP1
Intel Core2 Q6600 Quad core @ 2.4GHz
4GB RAM
Gigabyte EP45-UD3L motherboard with SATA2 and USB 2.0 interfaces
USB 3.0 add-in interface card
USB 2.0 powered hub
USB 3.0 powered hub

It was a pretty hot system when I built it many years ago. It still satisfies me for all that I do with it, especially with the change of the system drive to an SSD.

Here are the numbers and some observations.

-------------------------

Samsung 840 EVO 250GB SSD internal system drive on SATA2 (3Gb/s):
Write: 125 - 260 MB/sec
Read: 140 - 263 MB/sec
This is the system drive and performance was all over the map depending on what other activity required hard drive access. When I saw this I shut down everything else to reduce the variability but there are still background processes doing stuff.

-------------------------

Samsung 840 250GB SSD mounted in USB3.0 case on USB 3.0 hub and interface:
Write: 134 MB/sec
Read: 147 MB/sec

Same drive on USB 2.0 hub and interface:
Write: 29 MB/sec
Read: 33 MB/sec

-------------------------

Kingston Data Traveler USB 3.0 16GB flash drive on USB 3.0 hub and interface:
Write: 7 - 19 MB/sec
Read: 115 MB/sec

Weird behaviour on this one. Write performance was all over the map. This might be because of the way the tests are run. The program does a write to the drive for 8 seconds, then reads back what was written, then erases the file and starts again.

The read speed was so fast that it only took 2 seconds or less to read back what it took 8 seconds to write.

It's possible that the read operation is so quick that the erase operation, which is effectively another write operation, sometimes doesn't have time to complete before the next write test begins.

This seems to be confirmed by the tests on the same flash drive in the USB 2.0 hub and interface. Write performance this time was very consistent and it appears that the slower speed of the reads on the slower interface allowed enough time for the erase operation to complete before the next write test began.

Write: 16 MB/sec
Read: 33 MB/sec

If you were to graph or compute the write test results of the test runs in the USB 3.0 connection (above) it would likely average out to about the same 16 MB/sec as you see when it's on the USB 2.0 connection. That's better than the USB 2.0 flash drive in the next test but a far cry from the write speeds in many USB 3.0 flash drive published specs.

-------------------------

Patriot Flex 16GB USB 2.0 flash drive on USB 2.0 hub and interface:
Write: 9 MB/sec
Read: 32 MB/sec

Same drive on USB 3.0 hub and interface:
Write: 9 MB/sec
Read: 32 MB/sec

-------------------------

And a couple of rotational type hard drives for comparison:

Western Digital Black 1TB on SATA2 (3 Gb/Sec):
Write: 116 MB/sec
Read: 116 MB/sec

-------------------------

Western Digital Blue 1TB on SATA2 (3 GB/sec):
Write: 171 MB/sec
Read: 171 MB/sec

-------------------------

...ken...
Ken in Regina
Here are a couple more drives. They weren't immediately at hand when I did the previous tests because they're my offsite backup drives. That's so if an airplane crashes into the house or something I can recover everything that matters.

Western Digital MyBook 3TB USB 3.0 drive on USB 3.0 hub and interface:
Write: 63 MB/sec
Read: 106 MB/sec

----------------

Patriot Slate 16GB USB 3.0 flash drive:
Unusable results

I have this drive encrypted with Bitlocker because I often carry it around in my pocket. It appears that the encryption/decryption absolutely kills the performance. The write performance is abysmal. After the first test pass it writes at about 1 MB/sec to 3 MB/sec. The read speed is not particularly relevant because there is too little written in the 8 second write test to make a useful read measurement.

I'm tempted to erase the drive and test it unencrypted to see just how badly the encryption/decryption affects the performance and how much is that it's a cheap flash drive.

...ken...
Ken in Regina
Okay, I formatted the Patriot Slate 16GB USB 3.0 drive and retested. The results are just as bad. Stay away from this one, it's worse than any of my USB 2.0 flash drives!!

...ken...
Ken in Regina
I returned the Patriot Slate and picked up a 16GB Kingston DataTraveler R3.0 G2. The Kingston spec sheet says that it will Write at 25 MB/sec and Read at 120 MB/sec.

The BlackMagic Disk Test shows 25 MB/sec Write and 120 MB/sec read.

The Windows Copy dialog when I dragged/dropped a 2GB video file to/from the flash drive showed above 45 MB/sec through the entire write operation with a final tally of 45.8 MB/sec for the full 2.1GB file. The read operation was consistently above 120 MB/sec through most of the read operation with a final tally of 118 MB/sec.

This is a nice quick flash drive no matter which way you test it. And it's not the highest specs on a USB 3.0 flash drive. The 32GB version of this drive, for instance, is spec'ed at 45 MB/sec Write speed.

From now on I will be paying close attention to the specs before I purchase a flash drive.

...ken...
Boyd
Cool stuff Ken.

That is odd about the internal SSD performance varying so wildly. Have never seen this with the Blackmagic test. It was very steady testing the 7200RPM internal disk on my Windows 7 desktop. I guess SSDs behave a bit differently? Or maybe it's an issue with the Blackmagic program.

On my Mac the SSD figures are always very steady.


Similar Topics:
Site:   Home,   Search
General forums:   Software,   Hardware
Focused forums:   ALK,   DeLorme,   Garmin,   iGuidance,   MapFactor,   Microsoft,   Navmii,   Odyssey Navigator
laptopgpsworld.com About