Announcement

Collapse
No announcement yet.

After Years Of Waiting, Hands On With The AMD ARM Board

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

  • #21
    Originally posted by wizard69 View Post
    Surprisingly i don't think ARMs future will be driven by their cores, but rather what is integrated around those cores. ARMs future will be shaped by the ability of companies to engineer their own SoC to bring new technologies like hardware acceleration of AI / ML technologies.
    Something that could be very interesting is an ARM SoC that integrates very fast networking and a lot of PCIe lanes with a reasonable number of cores on a single die. Those would be great for storage. Intel currently doesn't seem to have a successor to the Xeon-D which was wildly successful in taking market share from ARM, so if I were building ARM cores I'd be looking at this. Cavium ThunderX2 looks promising in that regard.

    Comment


    • #22
      Originally posted by Qaridarium View Post
      I just vote for the RISC-V option for AMD instead of ARM...
      RISC-V would be the much better choice.
      It would be a great move, if AMD would come up with an equivalent of RISC-V in the field of GPUs.
      Last edited by sverris; 07-13-2017, 04:07 AM.

      Comment


      • #23
        I had one of these pre-ordered too, but when they notified me of the PCIe slot issue, and offered me a Hikey960 instead, I opted for that. At this point a quad-core Cortex-A57 just isn't interesting to me any more.

        Comment


        • #24
          Originally posted by nils_ View Post

          Something that could be very interesting is an ARM SoC that integrates very fast networking and a lot of PCIe lanes with a reasonable number of cores on a single die. Those would be great for storage. Intel currently doesn't seem to have a successor to the Xeon-D which was wildly successful in taking market share from ARM, so if I were building ARM cores I'd be looking at this. Cavium ThunderX2 looks promising in that regard.
          I like the Xeon-D (except the silly price), but it is made in the same fab as the recently discontinued Atom's. I don't think we will see anymore after this generation finishes their run.

          It won't happen, but this would be a great CPU for a hobby board due to its low power use and commonality with certain Atoms. Intel has to protect its pricing structure.



          Comment


          • #25
            Originally posted by [email protected] View Post
            I couldn't care less about the architecture, as long as it is supported in the open. AMD could roll their own instruction set, or maybe a RISC-V chip, and I would be quite happy if the perf/watt is good enough (and the perf itself, of course). Give us some completely open boards, with decent integrated GPUs, and completely change the raspberry-like segment.
            Bonus points if it is compatible with standards technology (maybe produce different board form factors): PCI, Sata, DIMM, USB, etc...

            I actually quite liked their initial plan to make x86 boards and sockets compatible with their ARM chips. The current instruction set segregation makes no sense in my opinion, especially if the software is open source and can be recompiled at will.
            A RISC-V + little Vega APU would be spot-on for kiosk-style embedded applications (that is, once the RISC-V privileged and vector specs are firm). Probably also killer for set-top boxes and digital signage in the lower bins.
            Last edited by microcode; 07-13-2017, 04:28 PM.

            Comment


            • #26
              Originally posted by sverris View Post
              It would be a great move, if AMD would come up with an equivalent of RISC-V in the field of GPUs.
              Shader cores are a little bit more specialized, but the ISAs are basically scalar RISC (or small vector RISC). There's nothing really wrong with RISC-V as a base ISA for a GPU, but the contract would be a bit different around flow control, FP exceptions, and the suchlike. I don't think there's really any decent reason to do this though. GPU ISAs are highly specialized for very good reasons. Basically every innovation in GPU efficiency comes with a reorganization of the ISA.

              Comment


              • #27
                Intel has the Xeon D and Denverton cards they can play to block new entrants for small/micro server platforms. Because of that, no one is targetting this niche which is really in need for options.
                I am hoping the chip AMD is selling to HP for their gen10 microserver could become an option for us DIY system builders to build our storage boxes. I am sick of using "workstation/server" class mobos from Intel just to have ECC or having to go Ryzen and then buy a power hungry GPU plus a "gamer" mobo for no reason (really, why can't we have small-server Ryzen mobo with no LED, audio and wifi but instead 10-12 SATA ports, IPMI, and 10GbE port?).

                Comment


                • #28
                  I'm sadly trying to get my Cello board up and running. I think the memory I have is not compatible, but I have no way to be sure. It's frustrating waiting so long for these boards and not being able to use them. The lack of PCI Express port doesn't bother me since it has Gigabit Ethernet, 2 SATA ports, and can take more than 2GB RAM. (The HiKey doesn't have that feature set, sadly.) That's been my reason for using this, and right now they are paperweights. I really hope LoveRPi will make some RAM choices available on their store, and have a community site for others with these boards because I have been really disappointed with LeMaker's support thus far.

                  Comment


                  • #29
                    I only just now noticed this article; did the follow-up review ever get written?

                    I'd love to see how well various GPUs work with this.

                    Comment

                    Working...
                    X