[conspire] Quiet, Freedom-compatible NAT/firewall/misc box?

[Dana Goyette]

Incidentally, there are some server-grade "Avoton" Atom processors
that have 2, 4, and 8-core models, and support up to 64GB of ECC
memory.  For comparison, the Core i7 / Xeon E3 series only does 32GB
memory.
The Avoton SOC contains a four-port Intel i354 network controller, but
oddly, some OEMs ignore that and waste PCIe lanes and watts for
discrete i210 network controllers instead.

Another thing that seems hard to find: low-power AMD boards with IPMI.
IPMI firmware is another can of (usually GPL-violating) worms, but
remote KVM access can be useful.

I'm looking forward to seeing what the AMD Opteron A1100 (aarch64) is
capable of.  The SOC has two integrated 10-gigabit controllers, and
can take up to 128GB of ECC memory.  I believe AMD is working directly
with the upstream kernel to get the thing supported.

Considering how often the AMD processors are artificially restricted
to bottom-of-the-barrel devices (such as laptops with terrible
screens), it really does make me wonder if Intel is still doing
anticompetitive arm-twisting.

On Tue, Mar 17, 2015 at 6:48 AM, Rick Moen <rick at linuxmafia.com> wrote:
> Just wanting to elaborate a bit on:
>
>> Suspicion #2 (less conspiratorial):  AMD suffers because it's been
>> crowded into price competition with really anaemic, low-spec ARM-based
>> boards best suited to smartphones and low-end tablets that are so cheap
>> that the OEMs just aren't willing to pay for better performance and
>> fewer limits.
>>
>> #2 would explain both why it's so difficult to _find_
>> enthusiast-friendly (i.e., not an HP Pavillion sealed-tight mini-something)
>> units based on the newer and truly amazing low-power, high-performance
>> AMD SoCs, _and_ why low-power computing has been almost completely
>> abandoned to stripped-down ARM offerings with all their
>> proprietary-software dependencies and utter lack of standard kernel
>> support, even after _many_ years of ARM-based Linux computing.
>
> Somehow through dominating the upper end of the market, Intel seems to
> have perpetuated a perception that its CPUs and chipsets are preferable
> to AMD's at _every_ price point and in every specialty within the x86_64
> market space, which is very demonstrably very untrue.  In the low-power,
> low-cost segment, AMD has lately had _massively_ better offerings,
> especially per dollar.
>
> You look for an Intel-based low-power system, and inevitably find an
> Atom-series CPU, which frankly is a bit lackluster in performance, in a
> chipset that, at best, maxes out at 8GB RAM.  The AMD alternatives are
> things like the ASRock mini-ITX / AMD Kabini bundle I cited on Newegg,
> which has a quite fast CPU that draws less power than the Atom, runs
> cooler, and plugs into a motherboard that maxes out at 32GB.  And the
> ADM package ends up being cheaper, as well.  But people will recommend
> Atom-based things anyway, because Intel.  Which is kind of crazy.
>
> The lower end, which is what low-power is considered to be, rightly or
> wrongly, is dominated within the hardware industry by the
> embedded-computing mindset (that surrounds ARM).  As Don Marti will tell
> you (he having been Editor of _Embedded Linux Journal_ for some years,
> the embedded-computing offshoot of _Linux Journal_, secrecy and
> proprietary components are totally routines in the embedded space.  GPL
> enforcement against embedded-computing hardware companies typically
> fails for a number of reasons including chipset churn being so rapid
> that the violator can just stall for a few months until the model in
> question gets EOLed and then says 'Hey, we've ceased violating.'  Rapid
> chipset churn also means that reverse-engineering is less fruitful
> because so many things are moving targets.
>
> But the effect of the _standard_ low-power offerings being relegated to
> ARM is that everyone gets so used to the user-facing hassles and
> compromises of the ARM platforms that the hassles and compromises fade
> to background: Rarely do they get talked about, even by Linux people.
> They become 'normal', so people don't say 'Wait, this is stupid.  Is
> there an alternative that's _almost_ as power-thrifty and _almost_ as
> cheap that doesn't have these problems?  With no strange boot
> configuration and bootloader setups, no out-of-tree patchsets against
> old and moldy kernels?  No proprietary blobs just to run X11?'
>
> And that is exactly what you _can_ get with the sort of AMD SoC /
> miniature motherboard setup I've been talking about in this thread.
> In general, the 'HTPC' market is producing some really kick-ass small
> form-factor, quiet, low-power hardware at very low prices, and AMD by
> all rights ought to own that market, given the exceptional things the
> Kabini/Temash, let alone the hard-to-find Beema/Mullins SoCs seems
> capable of, provided the OEMs bother to ship motherboards that can use
> them instead of assuming that everyone wants ARM-based alternatives just
> because they hit absolute bottom dollar.
>
> I'm really not quite understanding why the OEMs have in general done
> almost nothing with those chips in SFF PCs, and instead apparently put
> them only into tablet computers.  It's a pity they haven't.
>
> And all I can add to that is:  Linux people who want quality low-power
> gear need to be better at voting with their dollars, or this bad
> situation will get worse.
>
>
> _______________________________________________
> conspire mailing list
> conspire at linuxmafia.com
> http://linuxmafia.com/mailman/listinfo/conspire