Re: [PLUG] Question about AMD Ryzen 3000x rdrand bug

[Rich Freeman via plug <plug@lists.phillylinux.org>]

On Sun, Dec 6, 2020 at 1:21 PM Rich Mingin (PLUG) via plug
<plug@lists.phillylinux.org> wrote:
>
> The rdrand issue has not been resolved in any 3000 series CPUs manufactured to date. I am still waiting for confirmation if it was corrected in the 5000 series or not, the long delay makes think it has not been corrected.
>

Seems a bit moot.  New AMD CPUs don't boot without new firmware, and
it sounds like newer firmware issues fix the issue, so it would be
impossible to boot a zen3 CPU with the issue.

I don't really have a problem with microcode fixes to problems as long
as the resulting solution still meets the original specs.  Solutions
that tank performance by 10% or whatever aren't great, but modern CPU
instruction sets are already a software abstraction of the actual
hardware, so doing the fixes in this layer seems reasonable.

Biggest frustration I have with my new zen3 CPU is that I can't get
the FCLK to go over 1700MHz, but that already is over the spec so it
is hardly something to complain about.  There is speculation that
they'll try to improve this in future firmware.

One thing I don't care about with AMD lately is that they don't seem
to officially release microcode updates.  When Spectre/etc was new
figuring out what issues are fixed in what microcode files was hard to
confirm.  The files themselves weren't on any kind of official
distribution site with any kind of versioning.  You'd look at the
linux-firmware git tracker and it sounds like they'd basically get
them emailed from some random person at some random company as a
favor.  I'm guessing that AMD distributes them to motherboard vendors
to incorporate into firmware updates, which they would distribute as
they saw fit, or not.  I'm guessing Microsoft probably would get a
copy eventually as well.

In theory if you have updated firmware you don't need the latest
microcode in linux.  However, if your distro is current any microcode
fixes should be getting installed at boot time via the
double-initramfs mechanism built into the kernel.

-- 
Rich
___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug