Announcement

Collapse
No announcement yet.

USB Updates In Linux 5.5 Help Intel Ice Lake, NVIDIA Xavier + More - But No USB 4.0 Yet

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • USB Updates In Linux 5.5 Help Intel Ice Lake, NVIDIA Xavier + More - But No USB 4.0 Yet

    Phoronix: USB Updates In Linux 5.5 Help Intel Ice Lake, NVIDIA Xavier + More - But No USB 4.0 Yet

    Earlier this week as part of his series of pull requests, Greg Kroah-Hartman has submitted the USB subsystem updates for the in-development Linux 5.5 kernel...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    The official name for this specification is "USB4" and not "USB 4.0" as stated in the official announcement from USB-IF. Because... reasons

    Comment


    • #3
      The mainline kernel is disabling UAS for JMicron SATA enclosures.
      Whaaaaat? Really? I thought that JMicron had some known chipsets that worked quite well with UAS support. So what chipset and enclosure do you use now? There's not a lot of others out there afaik?

      ---

      From my notes:

      Official Inateck page: https://www.inateck.com/inateck-fe20...p-support.html

      Amazon: https://www.amazon.com/Inateck-Enclo.../dp/B01FQ5R0PG

      This one is the model that's been praised on Manjaro forums as working well and it also uses the good chipset.

      > Engages automatic sleep mode after 30 minutes of inactivity to save energy! LED flashes every three seconds to indicate standby.



      > Of the five USB to SATA chips I tested, only Jmicron’s JMS 578 can support Trim.
      And then some later information that indicates there can be issues with JMS578, depends on firmware from vendor:

      JMS578 and JMS567 have some issues(at least in late 2017):



      Foced spin-down timer(might be configurable according to discussion midway, and not affect SSDs?)

      Does cause Unknown state issue(ASM1153E does not have this issue)

      No working TRIM support

      Update the firmware:



      All those issues may be resolved. Do note that even chipset aside, there is differing vendor firmware versions, where some vendors have their custom firmware supplied from the chipset vendor which may not be available elsewhere using the chipset.

      StarTech products for example had issue with TRIM, but were later fixed with new firmware.
      ASM1153 can have it's own issues, again often vendor specific, I think recent Seagate models aren't blacklisted and work correctly these days(but then again they're probably not using ASM1153?):

      Enclosures use varying chipsets in their USB-to-SATA bridges, some are unreliable/blacklisted as they don't correctly/completely respect the spec or have some quirks:



      > Newer Seagate USB3 enclosures relying on ASM1153 with branded firmware need various quirks

      Related: https://www.reddit.com/r/archlinux/c..._work/d50r6mz/

      > Yeah, I did a bit more research and found that certain ASMedia chipsets work while others don't. Apparently the kernel can't differentiate between a few different models and so it disables UASP on them.

      Also related: https://forum.odroid.com/viewtopic.php?t=31232

      > (All Seagate external drives have an ATA-to-USB enclosure that is incompatible with UAS. One must manually disable the use of the uas module by setting the quirks option of the usb-storage module.)
      This Phoronix thread provides some insights of troubles with Linux and chipsets among other issues. Oddly it states issues with JMicron which I'm seeing praise for these days rather than the ASMedia chipsets they're praising in the discussion, mind you this was back in 2016 and prior to the 4.9 kernel which I think provided some improvements(USB 3.1?):



      Related, fast forward to 2018 and JMicron, specific chipset is praised with the problematic ASM1153E being as having I/O issues:

      alright thanks @edupv Oh, and one another problem that bugs me once every one or two week in armbian 5.27, but in 5.40 it's almost everyday, The UASP stopped responding requests, it freeze entire system for several minutes, and every I/O to my hard drive makes the OS unresponsive. I was using ASM...


      That thread also states that the chipset compatibility can depend on the firmware it's given, with Seagate using one that is problematic.
      I have considerably more in notes on the subject(spent a rather long time researching the topic). Bit sad that JMicron is being blanket blacklisted though.. is there a way to verify/reproduce the issues cited with their chipsets?
      Last edited by polarathene; 29 November 2019, 08:05 PM.

      Comment


      • #4
        Originally posted by numacross View Post
        The official name for this specification is "USB4" and not "USB 4.0" as stated in the official announcement from USB-IF. Because... reasons
        searching for USB devices at a specific version is a bitch because you will pull up all items that have "USB" and "3" in the title for example.

        With USB4 you can search easily for devices supporting this revision of USB.

        I like to believe that this was an intentional choice and that whoever gave names to USB 3 revisions was thrown out of a window.

        Comment


        • #5
          Seeing that Gigabyte just released the first USB 3.2 2X2 PCIe card last week, and that the Asmedia 3242 is just now hitting the streets in TRX40 boards, it will probably be next summer before we see any real hardware.

          Comment


          • #6
            Originally posted by edwaleni View Post
            Seeing that Gigabyte just released the first USB 3.2 2X2 PCIe card last week, and that the Asmedia 3242 is just now hitting the streets in TRX40 boards, it will probably be next summer before we see any real hardware.
            image_removed
            That graphic is not right. ASMedia controllers pictured are NOT doing true PCIe but NVMe. So they will work with SSDs but not for anything else that isn't using NVMe protocol.

            Also wtf is happening at the bottom slot. Take a native PCIe (that will support NVMe already), convert it to USB, convert it to NVMe. I guess that it shows what you can do, but it's kind of weird.

            Comment


            • #7
              Originally posted by starshipeleven View Post
              That graphic is not right. ASMedia controllers pictured are NOT doing true PCIe but NVMe. So they will work with SSDs but not for anything else that isn't using NVMe protocol.

              Also wtf is happening at the bottom slot. Take a native PCIe (that will support NVMe already), convert it to USB, convert it to NVMe. I guess that it shows what you can do, but it's kind of weird.
              Asmedia ASM2364 is a dedicated 4 lane bridge chip, so I would assume they run NVMe over the USB 3.2 2X2, since by design to get the 20Gbps speed you need 4 lanes.

              I agree that the bounce back and forth seems silly, but I can only guess it was done so they could support other 2X2 devices that didn't use NVMe. To get "driverless" operation, they had to let the NVMe device look like a USB device.

              After reading this:



              I just feel like the whole USB 3.2 is a stopgap until USB over Thunderbolt (USB4) comes out of the depths of someones lab and sweeps away all of this naming and speed idiocy. When USB4 comes around, there is still going to be a huge dependency on the quality of the cable and the number of pins wired in the C connector.

              Comment

              Working...
              X