Announcement

Collapse
No announcement yet.

AKiTiO Thunderbolt Devices Begin Receiving Firmware Upgrade Support Under Linux

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

  • AKiTiO Thunderbolt Devices Begin Receiving Firmware Upgrade Support Under Linux

    Phoronix: AKiTiO Thunderbolt Devices Begin Receiving Firmware Upgrade Support Under Linux

    AKiTiO is the latest hardware vendor beginning to allow for firmware upgrades in an easy and reliable manner under Linux...

    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 cookie message on that Akitio is really good! You actually have a "I decline" button, and you have checkboxes so you can choose which types of cookies to allow! (including explanations of what data is collected for each cookie type!)

    Comment


    • #3
      As a long-term eGPU user myself, every time I look at Akitio it looks horrible.
      I use EXP GDC v8 (works fine under Linux too btw).
      EXP GDC: 35$, pure PCIe.
      Akitio: 300$, Thunderbolt (much less speed than PCIe due to protocol wrapping/emulation/whatever).

      Comment


      • #4
        I wasn't even aware the AKiTiO external GPU enclosures were working under Linux
        Most of these enclosures use the very same Thunderbolt controller, so if either one works then it is reasonable to expect the others to work too.

        Akitio: 300$, Thunderbolt (much less speed than PCIe due to protocol wrapping/emulation/whatever).
        This is actually not ture. In fact, as surprising as it may be, the Thunderbolt 3 port is directly wired to a 4x PCI-express port on your CPU (or 2x depending on your manufacturer), that is how it can achieve the specified speeds.

        Comment


        • #5
          Originally posted by Venemo View Post
          ... the Thunderbolt 3 port is directly wired to a 4x PCI-express port on your CPU (or 2x depending on your manufacturer), that is how it can achieve the specified speeds.
          In theory yes. In practice, all the benchmarks I've seen at egpu.io and techinferno show much lower speeds for Thunderbolt, so some shenanigans are going on with it.

          Comment


          • #6
            Originally posted by Aleksei View Post
            In theory yes. In practice, all the benchmarks I've seen at egpu.io and techinferno show much lower speeds for Thunderbolt, so some shenanigans are going on with it.
            As far as I've seen according to reviewers, it gives you approx. 15-20% less performance than a "normal" PCI-express in a desktop. However I have not yet seen an apples-to-apples comparison to the EXP GDC as of yet; also most reviews make some pretty unfair comparisons. (For example, one guy connected the eGPU to a pretty weak computer and then compared the results to something with a much more powerful CPU. Another guy compares a weaker eGPU to a superior desktop GPU model, etc.) Would be interesting if Michael could get his hands on both a Thunderbolt enclosure and an EXP GDC and could do a fair comparison for us.

            Comment


            • #7
              Originally posted by sandy8925 View Post
              The cookie message on that Akitio is really good! You actually have a "I decline" button, and you have checkboxes so you can choose which types of cookies to allow! (including explanations of what data is collected for each cookie type!)
              Yeah it's probably the only site I've seen that has a "i decline" button, and it seems to work as advertised. That's a massive PR boost for people looking at this news (from here and other sites).

              Comment


              • #8
                Originally posted by Aleksei View Post
                I use EXP GDC v8 (works fine under Linux too btw).
                EXP GDC: 35$, pure PCIe.
                EXP GDC is a x1 PCIe lane, at most a x2, you can't fit more in the HDMI cable they are using.

                "Support for PCI-E X16, actual X1 mode (according to different notebook configuration can be upgraded to X2 mode)"
                I personally doubt it can actually go in x2 mode at all. Could you run GPU-Z and tell me what it says in the "bus interface" field? It should say first the card's interface, then write the current amount of lanes and pcie revision in use.
                Or with lspci -vv if you are on Linux https://unix.stackexchange.com/quest...-the-pcie-card

                The minipcie EXP GDC are limited to one lane of PCIe 1.0 (because the host port in the laptop is, even if the laptop is relatively modern), period. I won't even go into the expresscard version, it's even sadder.

                If the card docking is actually wired for just a x1 PCIe lane (which is what I strongly suspect), the m.2 connectors can use PCIe 2.0 or 3.0, so even if it's one lane you are getting 2 or 4 times the bandwith of the mini pcie connector.

                if you actually have a m.2 host port in the laptop you'd get the best possible performance out of it (i.e. full x4 PCie lanes of PCIe 2.0 or 3.0) with a couple adapters, this to pipe the m.2 connection out of the laptop


                and this to convert the m.2 slot into a normal PCIe slot with an additional power connection (note the connector that is not "closed" at one end so you can slot in longer cards) https://www.ebay.com/itm/NGFF-M2-to-...d=173400805110

                Assuming your laptop's slot is positioned conveniently enough, otherwise you would need a PCIe ribbon on top of that.

                Note that all these adapters are passive, they only swap electrical traces around. The rubber ribbon in the first one is pretty stiff and strong, it's wobbly like Sata cables.

                I did some testing with the m.2 nvme slot of my mini itx card as I want to add a 10Gbit ethernet without sacrificing the GPU slot (also because it's a Ryzen without integrated graphics). The test card I mounted in there (an old GPU) reports it's actually running at x4 in GPU-z (on windows) and also on linux reports the same.

                Another thing I looked into was just external PCIe enclosures. http://www.dynapowerusa.com/product/...lot-na211a-g3/
                Which is a prosumer product spin-off of "external PCIe expanders" used for servers and stuff (somewhat obvious in the marketing material, look, it's a low profile card that fits in a rack server, connect it to this cool-looking brushed aluminum enclosure that looks like a sore thumb in a server room). There is a host card that takes a x4 slot on a mobo and sends it outside from the back, then you route the pure PCIe over the usual server-y cables used for high performance networking inside the same rack. You can plug that on the adapters above, for example.

                There are also bigger cards, full x16 http://www.ioi.com.tw/products/prodd...ProdID=1060167 and also murderously expensive full x16 rack-mountable external enclosures.

                But a x4 PCie 3.0 won't really be saturated even by gaming laptop CPUs, so whatever. They are ok already.

                Comment


                • #9
                  Yeah of course, I posted links to ebay (as an example of products) and vBullettin stupid antispam decides a highly informative post for Aleksei above this requires admin approval.

                  Comment


                  • #10
                    This is good news, for now Thunderbolt is still under embargo, if Intel will finally open the spec and make it royalty-free (as they promised) we will finally see it more widespread and these things will finally come down a bit in price.

                    As it is, they are just losing vs pure pcie adapters, even the EXP GDC running with a single pcie 2.0 or 3.0 lane is a lot of bang for the buck.

                    Comment

                    Working...
                    X