On April 11th, 2011, both Texas Instruments and Cypress Semiconductor made announcements concerning upcoming SuperSpeed USB (USB 3.0) Products.  On the surface the two announcements are unrelated and in all honesty they really are totally unrelated.

 

We at TI announced that our 4-port SuperSpeed USB Extensible Host Controller (xHCI), TUSB7340, is the first 4-port xHCI device that has been certified by the USB Implementers Forum.  It extends TI’s Industry broadest portfolio of SuperSpeed USB Products that includes the industry’s only SuperSpeed USB discrete transceiver, TUSB1310A; a 2nd generation of USB-to-SATA bridge, TUSB9261; both the above 4-port xHCI device as well as a 2-port version, TUSB7320; and a 4-port hub, TUSB8040.

 

Cypress announced that they will be releasing a new general purpose SuperSpeed USB peripheral controller branded EZ-USB FX3.  The importance of this message is that this device will enable peripheral products NOT in the storage space.

 

The importance of these two messages is that SuperSpeed USB is real and continues to ramp at incredible rates.  Just a little over a year ago at CES 2010, the first certified products were announced.  Now 16 months later, there are well over 200 certified products including the first integrated core logic chipset from AMD!

 

SuperSpeed USB is here – what do you want to do with it?

Anonymous
  • Darren,

    You should be able to obtain the source directly from the Kernel source or the GIT tree and build using your platform configuration.  

    – Prior to Linux 2.6.31 require patches or Sarah Sharp’s kernel branch

    – In mainline Linux kernel since 2.6.31

    For more information about the driver

    – Sarah Sharp’s Blog: http://sarah.thesharps.us

    – Mailing List www.linux-usb.org/mailing.html

    As far as the "hit" that SuperSpeed USB has on throughput, the primary culprit is actually the existing native mass storage class driver from Microsoft.  It is not geared to take advantage of the efficiency improvements of USB 3.0 - to fully appreciate those, you need to run what is called the USB Attached SCSI Protocol (UASP for short).  In out testing with the TUSB9261 and SSDs and using Crystal Disk mark as our benchmarking tool, we go from sequential read speeds of 200 MB/s with the native MSC driver to almost 275 MB/s using our UASP driver.  The sequential writes go from 169 to 199.  I do not have numbers of that drive talking directly to the native SATA host.  We have done other testing with different drives and the USB impact was around 2% on both sequential reads and writes.  There is still overhead with USB 3.0, just not as much as with USB 2.0 and once UASP is native to the OS, end-users will see the effect of USB 3.0 on throughout.

    Regardrs, Dan

  • I noticed the TUSB7340 mentions an available Open source Linux driver but I can't find it.  Can you point me to a link, a person, or clue what O.S. support there is?  Thanks!

    2nd question - You mention the TUSB9261.  One of the knocks from "end users on USB3 connected storage has been slightly slower speed  than natively connected SATA or eSATA.  In theory since the limiting factor should be the drives, I'd like to see benchmarks that show no essential throughput difference between eSATA or USB3 connections to the same drive. Does the TUSB9261 address some of the performance issues?