Announcement

Collapse
No announcement yet.

USB Improvements Land In Linux 5.8 With Intel Additions, Non-x86 Thunderbolt

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

  • USB Improvements Land In Linux 5.8 With Intel Additions, Non-x86 Thunderbolt

    Phoronix: USB Improvements Land In Linux 5.8 With Intel Additions, Non-x86 Thunderbolt

    On Sunday Greg Kroah-Hartman began sending in his pull requests to the different areas of the kernel he oversees for the ongoing Linux 5.8 merge window...

    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
    Anyone interested in reading how USB3 and PCIe Tunneling will work on the USB4 bus, the high level presentation is available here:

    https://www.usb.org/sites/default/fi...0Tunneling.pdf

    They also cover USB3 Native vs Tunneling and Device Routing.

    Honestly, its getting pin information on DP Alternate Mode and the associated power states that has to drive the engineers crazy. While alt mode is required in the USB4 spec, I am thinking of all the logic that has to occur in the device router that has to determine when a new cable is plugged into the host. And is the connector wired properly for the services it might be trying to acquire.

    I get it, one spec, one cable, one connector type. But as many have seen, not all cables are created equal. It will be an interesting transition.
    Last edited by edwaleni; 08 June 2020, 12:55 PM. Reason: spell correction

    Comment


    • #3
      Originally posted by edwaleni View Post
      Anyone interested in reading how USB3 and PCIe Tunneling will work on the USB4 bus, the high level presentation is available here:

      https://www.usb.org/sites/default/fi...0Tunneling.pdf

      They also cover USB3 Native vs Tunneling and Device Routing.

      Honestly, its getting pin information on DP Alternate Mode and the associated power states that has to drive the engineers crazy. While alt mode is required in the USB4 spec, I am thinking of all the logic that has to occur in the device router that has to determine when a new cable is plugged into the host. And is the connector wired properly for the services it might be trying to acquire.

      I get it, one spec, one cable, one connector type. But as many have seen, not all cables are created equal. It will be an interesting transition.
      Well,... I recently purchased 0.5m long USB-C extension cable for my USB-C dock with 0.2m long cable. And, the extension cable doesn't transfer video, only USB keyboard and mouse works, and also power delivery (laptop charging) worked.

      Comment


      • #4
        Does thunderbolt works with Power now with those changes?

        The news only confirmed it works for ARM, any raptor engineer here?

        Comment


        • #5
          Originally posted by andrei_me View Post
          Does thunderbolt works with Power now with those changes?

          The news only confirmed it works for ARM, any raptor engineer here?

          Maybe. Assuming there's someone that created the POWER dependent glue. Wouldn't do much good though, a quick glance through the POWER9 systems on offer from Raptor doesn't mention Thunderbolt at all. I doubt they bothered with a Thunderbolt package when the Linux kernel didn't even support it when these were designed. Also, one of the reasons people buy RCS POWER workstations is auditible security versus traditional PCs and ARM. Thunderbolt is inherently insecure and the current generation of Thunderbolt hardware is likely impossible to fix.

          Comment


          • #6
            Raptor doesn't list any Thunderbolt adapters in their HCL.

            I would just wait for a USB4 silicon provider that has no blobs associated with it and simply integrate into the next Talon/Blackbird motherboard. Or do a System76 like arrangement by designing an add on board themselves.

            Honestly if I owned a POWER from Raptor, high speed and flexible peripheral attachments wouldn't be a first priority. GPU's or FCPGA's (like with Xilinx) would be.

            Comment


            • #7
              Originally posted by kravemir View Post

              Well,... I recently purchased 0.5m long USB-C extension cable for my USB-C dock with 0.2m long cable. And, the extension cable doesn't transfer video, only USB keyboard and mouse works, and also power delivery (laptop charging) worked.
              different pins connected through I guess? Maybe this helps:

              Comment


              • #8
                Originally posted by kravemir View Post

                Well,... I recently purchased 0.5m long USB-C extension cable for my USB-C dock with 0.2m long cable. And, the extension cable doesn't transfer video, only USB keyboard and mouse works, and also power delivery (laptop charging) worked.
                That sounds like USB 2.0 only then, so it's mostly a power cable(either 3A or 5A, probably 3A), besides 3/5A power options, a cable can be USB 2.0 data only, USB 3 gen 1, or USB 3 gen 2(double bandwidth of gen 1, extra wires). USB 3.1 is USB 3.0 effectively, the important part there is if it's gen 1 or if it supports gen 2, after that gen1/gen2 are the same with USB 3.2, but bandwidth for both doubles again and doesn't need any special cable support, just that both devices support USB 3.2 on their ends.

                - Data: USB 2.0(480Mbps), 3.0 gen1(5Gbps, x2 with USB 3.2), 3.0 gen2(10Gbps, x2 with USB 3.2)
                - Power: 3A, 5A

                Beyond that, higher bandwidth afaik relies on lower lengths. The USB 2.0 cable can manage 4M passive in length, gen1 2M, and gen2 1M. Thunderbolt cables are similar to USB-C but have active/passive cable variants instead, where passive can only manage at 0.5M length.

                Length can also reduce effective voltage that makes it across the wire due to resistance, so shorter lengths generally help there, but their AWG(American Wire Gauge) rating is more important as if the vendor cheaps out there, a cable of equal length but smaller AWG rating(better) will manage to keep the voltage higher, a device requires enough watts(volts * amps) to operate, thus can cause problems if its unable to deliver sufficient amount to device. There's more to it still, with power source(eg power bank or some device like an RPi which has limited power to work with in the first place, and can fail to output a stable voltage under load elsewhere), USB hubs or ports(downstream ports), etc can also cause issues depending on setup and how they're trying to power a device(not just USB-C specific).

                Definitely gets complicated, but basically get a USB 3 gen2 5A capable cable of <1M length and you should be good, unless vendor is notably cheaper than competition for some reason by cutting costs somewhere they shouldn't be. Ideally USB4 would just have this type of cable sold as the only "USB4" type of cable. We want one type of cable, but since it'd be using USB-C, it's going to get confused with legacy ones, which you can't really avoid since people that don't need full featured cable don't want to pay for the extra stuff, which lets them save a little at the expense of confusing many consumers..

                Comment

                Working...
                X