Testing methods

Our test bench for this flash drive roundup is a desktop computer running Windows 7 Enterprise 64-bit and consisting of an Intel Core i3-2100 CPU, ASUS P8H61-I (Rev. 3.0) mini-ITX motherboard with two USB 3.0 ports using an ASMedia ASM1042 USB 3.0 controller, 2 x 2GB Patriot DDR3-1333, one Intel 320 Series (G3) 80GB SSD, one Western Digital 2TB Caviar Green HDD, and one LITE-ON IHAS124-04 optical drive.

Iometer is a standard storage drive benchmarking software capable of testing mechanical, platter-based hard disc drives, flash-based solid state drives, and USB-interface flash drives. We provide a number of measurements for each drive. However, since the vast majority of users writes data to a flash drive, and then read off of it, the most important measurements for each drive are the sequential write and read speeds. While this might seem obvious, it should be noted this is different from a boot (OS) and application drive. OS/app drives typically do not see many frequent, sustained writes after initial installation. Furthermore, random writes/reads of smaller files are very common for an OS drive – this is where SSDs truly shine over mechanical HDDs, giving an SSD-based system spectacular snappiness compared to platter drive-based rigs. Random write/read performance is not particularly important for a flash drive unless it is being used as a ReadyBoost device. However, given how cheap DDR3 currently is, if you can afford a motherboard with USB 3.0 ports, you can afford enough system memory to not need a ReadyBoost drive! Furthermore, all of the USB 3.0 flash drives in this roundup are fast enough to support ReadyBoost in Windows 7.

That said, software virtualization applications like Ceedo, which act as a portable 'OS on a flash drive' certainly benefit from improved read/write performance, though thorough benchmarking and real-world performance tests of a program like Ceedo installed on a USB 3.0 flash drive are outside the scope of this article.

In addition to Iometer readings, we provide real-world usage scenario data for the drives. The first test writes 3,364 PDFs totaling 3.20GB from the SSD to the flash drive, then reads those same PDFs from the flash drive back to the SSD. The second scenario first copies 100 MP3 albums containing 1,133 files totaling 7.94GB from the SSD to the flash drive, and then those same MP3s from the flash drive to the SSD. The third and final test writes a single 4.16GB DVD image file (ISO of Fear & Loathing in Las Vegas) to the flash drive and then back to the SSD.

All drives were formatted to NTFS with a 4KB allocation unit prior to benchmarking, and all tests in Iometer were run also using a 4KB block size. While default file system and block size are variable between manufacturers, NTFS allows files larger than 4GB to be written to the flash drive (critical for DVD and BRD ISOs, among other file types), and it allows drives larger than 32GB to be used under Windows XP (unlike FAT32).

Samples

We were graciously provided the following samples by their respective manufacturers:

Kingston must have a die-hard Cincinnati Bengals fan on their design team!

Product Usable Capacity Price $/1GB
ADATA S007 32GB USB 2.0 29.8GB $44 $1.47
ADATA S102 32GB USB 3.0 29.2GB $50 $1.71
Kingston DT Ultimate G2 32GB USB 3.0 29.8GB $85 $2.85
Kingston DT R500 32GB USB 2.0 29.8GB $62 $2.08
Mushkin Ventura Pro 32GB USB 3.0 29.4GB $75 $2.55
Patriot Supersonic 64GB USB 3.0 60.5GB $150 $2.48
Patriot Supersonic Magnum 64GB USB 3.0 58.9GB $196 $3.33
Super Talent Express DUO 16GB USB 3.0 14.6GB $28 $1.92
Super Talent Express RC8 64GB USB 3.0 51.2GB $145 $2.83

These are the prices as of the time of this article's writing, and are absolutely not static. Be sure to keep this in mind when determining their comparative costs and considering their bang for the buck. As you can see, the USB 3.0 drives have a wide dollar per GB cost, from ADATA's S102 at $1.71/GB to Patriot's $3.33/GB Supersonic Magnum. Find out how these flash drives performed on the next few pages!

Introduction: A Brief History of USB Flash Drives USB 3.0 and 2.0 Flash Drive Native Interface Write Performance
Comments Locked

42 Comments

View All Comments

  • coyote2 - Saturday, August 6, 2011 - link

    Are you talking about 6GB/s eSATA III?

    USB3 is so much faster than the last eSATA, I'd be surprised to hear it's "protocol overhead" could make it perform slower.
  • theangryintern - Friday, July 29, 2011 - link

    I have the ADATA S102, got it a few months ago. So far I've been very happy with it. At home where I have a USB 3.0 controller card and an SSD boot drive, copies to and from the ADATA are screaming fast. At work, even on USB 2.0 it's still pretty fast
  • PotablePots - Friday, July 29, 2011 - link

    I was actually hoping that this article would look at portable application performance. I use a ton of portable application software and finding a flash drive that will give the best performance when running applications is something I could use AT's help on. Most of my portable software comes from PortableApps.com. I use mostly Portable Chrome and OpenOffice but also GIMP and Blender on occasion.
  • Aikouka - Friday, July 29, 2011 - link

    Zach, is it possible that you can list some value to help identify how much data you're transferring to these devices in the real world tests? For example, while I may copy PDFs to my thumb drives at times, I don't necessarily copy over hundreds of megabytes worth. I normally spend the most time waiting when I'm copying really large files to a thumb drive such as a movie I shot on my HD camcorder.
  • Gigantopithecus - Saturday, July 30, 2011 - link

    Hi Aikouka - The details of the real world scenario tests are on the second page. For the PDFs, the test copied 3,364 PDFs totaling 3.20GB. As I said, there are as many real-world usage scenarios as there are flash drive users! I personally, on a near daily basis, will copy hundreds, if not thousands, of PDFs onto a flash drive to perform a dump on a colleague's computer (hmm, that doesn't sound good - but you know what I mean, ha). That is if someone asks me about topics x, y, and z, I'll simply say read these, and give them a bunch of articles. I addressed your interest in large file transfers with the 100MB Iometer benchmark numbers and the real-world DVD ISO file read/write performance times. Those should give you a clear picture of which drives read and write bigger files, like those shot on your HD cam, the fastest.
  • justcommenting - Friday, July 29, 2011 - link

    Hey,

    I don't mean any offense, but I thought the writing style with this post was below normal Anandtech standards. The introduction skips details on what USB 1.1 changed that saw the widespread adoption and jumps around between discussing USB flash devices and USB hard disk devices without distinction between the two.

    I appreciate the various graphs, but I don't think the author did a very good job of explaining why certain drives pulled ahead in various scenarios, why that might matter, etc. Instead, graphs felt tossed into pages with little more than a description of the picture underneath.

    Like the pages before it, the conclusion is also a tough read due to general poor sentence / paragraph structure. I love Anandtech articles because of the highly technical content and the well-phrased, well-researched, and well-backed opinions put forth. I'd pay for it if I could, but preferably with stronger articles than this one. :)
  • Pozz - Saturday, July 30, 2011 - link

    indeed and less exclamation marks would be a start :)

    still, very interesting article
  • MaximillianSterling - Friday, July 29, 2011 - link

    Nice random U-M reference.

    Ah, the VERY long nights spent there. Although I preferred the Media Union.
  • Gigantopithecus - Saturday, July 30, 2011 - link

    I hoped someone would catch that. ;) Lotta really long nights there, brother!
  • Goi - Sunday, July 31, 2011 - link

    I would've liked to see what controllers and NAND flash chips were being used in the different flash drives. I know some are using USB 3.0<->NAND flash bridge controllers, while others are using SSD controllers with a separate SATA<->NAND flash bridge, or perhaps using a NAND device with a SATA interface. It would be interesting to find out how these design decisions affect performance.

Log in

Don't have an account? Sign up now