Once you get away from the “rack” as the unit and go back to a tower/NUC, I don’t think Oxide has anything special to add, but I may be wrong since I haven’t used their product.
A Dell (using them as an example since I’m familiar with PowerEdge) server/tower/NUC is pretty well integrated; the Oxide value add, as I understand it, is that your entire rack is integrated as opposed to being your own assembly of PDUs/servers/switches/etc.
Perhaps their management software (for spinning up VMs, managing storage, etc.) is high value, but I’m not sure how that works when you lose the rack-level integration and go back to a tower/NUC.
What do you see Oxide as adding at the tower/NUC level that takes you beyond the integration of a traditional tower/NUC from a vendor like Dell? I have seen similar excitement for “hobbyist Oxide” on other threads, but I’m still not completely sure what this actually means and how it would distinguish itself from existing sub-rack commodity hardware.
A sibling comment mentions smaller racks, which is certainly something I could get behind, but that’s not the same as towers/NUCs (or maybe that’s exactly what you mean: something that looks like a tower/NUC but is really a mini-rack with multiple miniaturised computers/switches/etc. inside?).
> The CPUs in Oxide racks are AMD, so, presumably AMD-based compute rather than ARM.
These don’t run Hubris though; the x86 cores are running Helios [0], Oxide’s flavour of Illumos (i.e. Solaris) that includes a bunch of binary blobs that aren’t even available, let alone open-source: they plan to make these “public” at some point, but it’s not clear whether this means source-available.
Based on the chips directory in the repo [1], they’re targeting a mix of NXP and ST parts, which are Arm, and the user isn’t likely to see them or care what firmware they’re running: I don’t think replacing the firmware with something running on top of Hubris is a value add, or even particular interesting for the hacker/hobbyist at all, in the sense that they’re a tiny cog in a much more complex system, and if you’re interested in hacking on something like that then it’s much easier/cheaper to do it in isolation rather than trying to do in-circuit debug on the part that’s keeping your server from catching fire (but that’s just my opinion!).
A Dell (using them as an example since I’m familiar with PowerEdge) server/tower/NUC is pretty well integrated; the Oxide value add, as I understand it, is that your entire rack is integrated as opposed to being your own assembly of PDUs/servers/switches/etc.
Perhaps their management software (for spinning up VMs, managing storage, etc.) is high value, but I’m not sure how that works when you lose the rack-level integration and go back to a tower/NUC.
What do you see Oxide as adding at the tower/NUC level that takes you beyond the integration of a traditional tower/NUC from a vendor like Dell? I have seen similar excitement for “hobbyist Oxide” on other threads, but I’m still not completely sure what this actually means and how it would distinguish itself from existing sub-rack commodity hardware.
A sibling comment mentions smaller racks, which is certainly something I could get behind, but that’s not the same as towers/NUCs (or maybe that’s exactly what you mean: something that looks like a tower/NUC but is really a mini-rack with multiple miniaturised computers/switches/etc. inside?).
> The CPUs in Oxide racks are AMD, so, presumably AMD-based compute rather than ARM.
These don’t run Hubris though; the x86 cores are running Helios [0], Oxide’s flavour of Illumos (i.e. Solaris) that includes a bunch of binary blobs that aren’t even available, let alone open-source: they plan to make these “public” at some point, but it’s not clear whether this means source-available.
Based on the chips directory in the repo [1], they’re targeting a mix of NXP and ST parts, which are Arm, and the user isn’t likely to see them or care what firmware they’re running: I don’t think replacing the firmware with something running on top of Hubris is a value add, or even particular interesting for the hacker/hobbyist at all, in the sense that they’re a tiny cog in a much more complex system, and if you’re interested in hacking on something like that then it’s much easier/cheaper to do it in isolation rather than trying to do in-circuit debug on the part that’s keeping your server from catching fire (but that’s just my opinion!).
[0] : https://github.com/oxidecomputer/helios
[1] : https://github.com/oxidecomputer/hubris/tree/020d014880382d8...