[coreboot] Does the Core2Quad Mobile Q9100 require microcode updates?

[ron minnich]

On Mon, Jan 9, 2017 at 2:09 PM Taiidan at gmx.com <Taiidan at gmx.com> wrote:

> Reasons to hate microcode updates:
> * They enable companies to ship broken CPU's and fix them later thus a
> CPU undergoes less testing (remember when software/games didn't have and
> worked fine without a day one patch?)
>

that's actually not a very good reason. Companies ship broken hardware all
the time. A company that has a microcode machine would be irresponsible
were they not to allow the option of an update. Further, hardware is ....
hard. It's simply not possible to catch every possible bug before it ships.


True story: when the Y came out, vendor abc used a Y-1 to check its
floating point calculations. The machines were to stop when a difference in
computations was found. The machines stopped. The Y-1 had a bug that the
newer machine found.

Why did I not use names? Because I've heard this story from architects at
just about every computer company. Hardware has bugs. Microcode fixes can
work around the bugs. So it makes sense to take advantage of that.

* Theoretically a nation state actor could screw around with a CPU and
> have an internal microcode update to secure their own systems, or
> something else like that.
>

All kinds of things are possible in theory, but it seems to me you're
making an argument against microcode, not microcode updates.


> * It is a black box (at least with intel) that is just another step of
> the war on general purpose computing-  the tivoization of hardware.
>
>
This is pretty much the same argument. If you don't like the CPUs you are
using, get different ones.

Given the use of a machine that has microcode, I don't think the opposition
to microcode updates makes any sense at all. You're far more likely to be
harmed by bugs in microcode than problems in a microcode update.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20170109/f8b4007c/attachment.html>