[Beowulf] recommendations on ARM distro?

[C Bergström]

On Sun, May 15, 2016 at 11:59 PM, Gerald Henriksen <ghenriks at gmail.com> wrote:
> On Sun, 15 May 2016 09:39:50 +0200, you wrote:
>
>>Was actually thinking of a clustered server setup using 64bit arm board
>>if possible
>
> First problem is that while the RPi3 has a 64bit cpu, there is only
> support for 32bit operating systems (you need binary blobs to get
> Linux running on these ARM boards, and for RPi3 they only support
> 32bit).
>
> [the Pi people chose the chip for its speed while maintaining
> backwards compatibility, thus they don't support or care about 64bit]
>
> Your next problem is that there is only 1GB RAM, and most of the I/O
> goes through a single USB2 port that has reports of being buggy.
>
> The RPi3 (or any of the Pi models) can certainly be used for
> educational purposes regarding a cluster, but unlikely to be actually
> useful.
>
> If you really want a cheap 64bit ARM board there are better options
> available that actually run 64bit Linux though you may need to pay a
> bit more.  Best advice is to find the ARM community for your preferred
> version of Linux and see what boards they support given the
> problematic nature of the cheap ARM boards.
>
>
> Is there hope for the future of 64bit ARM?
>
> Maybe.
>
> Red Hat is pushing hard for the the ARM vendors to implement the
> standard BIOS/UEFI boot process for AAarch64 boards which should
> remove most of the issues booting Linux on AAarch64, but the ARM
> hardware side has been very slow in coming to market and the cheapest
> board so far supporting a BIOS/UEFI is in the $300 range I believe.
>
>
> But for now, until the AAarch64 board makers can get their act
> together, for getting work done (like the 3D rendering you mention in
> another message) you are better off in the Intel/AMD world.

Very nice reply Gerald, not meaning to nit, but for certain workloads
I'd emphasize that accelerators make more sense than Intel.

Power and ARM have some uphill battles ahead of them, but I'm
optimistic that in the next 2-4 years we're going to see Intel go
against increasingly interesting products.