Plug computers

[Lennart Sorensen]

On Wed, Aug 25, 2010 at 11:50:09AM -0400, D. Hugh Redelmeier wrote:
> No device should overheat just because you use it at its capacity.
> Especially if it does not have any automatic method of detecting the
> problem and throttling the use.
> 
> I'd say that message smelled slightly of "blame the victim".

No, it sounds like they admitted they screwed up, and are giving
suggestions for ways to avoid the problem until they can provide a
proper fix.

> The guidance they give does not give you technical details of just
> what generates the heat and a formula for avoiding problems.  Is it
> the CPU or the ethernet that is a problem?

The ethernet controller is part of the CPU.

> Even though they talk about cutting down bandwidth as a solution, it
> seems unlikely to me that that would reduce the power used by the
> ethernet (running at 100M instead of 1G doesn't change the duty cycle
> of most of the chip for the better).

It takes a lot more power to operate at 1Gbps than at 100Mbps.
So actually it does make sense.  A fairly common estimate seems to be
about 100mW for 100Mbit, 1W for gigabit and 10W for 10Gbit.  Or at least
in that order of magnitude.

> The advice might be trying to reduce the duty cycle of the CPU.  If
> so, they should say so.

I don't think it is load on the CPU.  I think it really is that gigabit
operation generates a significant amount of extra heat (relative to the
total heat generated by the sheevaplug devices).

> The manufacturer / distributor ought to have the technical depth to
> determine an exact description of the problem.  It should be possible
> for them to measure power use of each chip under varying loads.  If
> they had, I would expect a crisper technical description.
> 
> I'd stay away until the problem is actually fixed.  And maybe longer
> than that.

Well it sounds like they are going to make a new enclosure for it with
better cooling.

> If it were manufactured in the US, that would be true, kind of.  HST
> is not "duty".  We have NAFTA so actual duty on things manufactured in
> the US is 0%.  But if this is manufactured in China (guess), and not
> enough value is added in the US, then there would be a duty (I don't
> know the rate).

Well some of the chips are by US companies, although probably
manufacturered elsewhere.  Now if they are assembled in the US, that
might just be enough.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists