[mythtv-users] Question on 8GB of RAM. Was: 32 bit or 64bit???

Stackpole, Chris CStackpole at barbnet.com
Mon Jan 26 21:46:36 UTC 2009


> From: mythtv-users-bounces at mythtv.org [mailto:mythtv-users-
> bounces at mythtv.org] On Behalf Of Francis Hartojo
> Sent: Monday, January 26, 2009 2:18 PM
> Subject: Re: [mythtv-users] Question on 8GB of RAM. Was: 32 bit or
> 64bit???
> 
> On Mon, Jan 19, 2009 at 3:02 AM, diane mittnik
<dianemittnik at gmail.com>
> wrote:
> >
> > Don't remember exactly where, maybe wikipedia, or other tech
articles
> > on the AMD cpus, but the TLB bug was supposedly eliminated in the
cpus
> > ending in 50, ie: 9550, 9750, 9850, 9950, 8450, 8650, 8750 (cpus
> > currently listed on pricewatch).  Also, I recall reading at least
one
> > source other than Intel/AMD about the TLB bug with the source (don't
> > recall who) stating that most or all TLB bugs can be eliminated
> > through a BIOS update, along with a tech article also stating that
the
> > AMD TLB bug was eliminated in affected cpus with a either a BIOS
> > update or through the workaround switch in the BIOS as the parent
> > states.  Once either the BIOS update (if there is one that is
> > different from the workaround) is installed/adjusted, there is no
> > longer a TLB bug.
> >
> > Like to know further details on this, please update us here as well
if
> possible
> >
> > Agreed, if it's a 8x50 or 8x50 cpu it's not TLB, or if the TLB
> > workaround or BIOS update is used, not likely? or not? TLB?
> >
> > Please let us know if this issue is resolved with a definitive cause
> 
> Sorry for the late response.  Here are some details of the problem I
> run into.  First the tech. specs.:
> 
> - ASUS M3A with the latest BIOS and the "TLB Fix" option enabled.
> - Phenom 9600 BE Quad Core.
> - 8GB RAM DDR2-6400 Crucial Ballistix.
> - NVidia FX5200 256MB PCI graphics.
> - Hauppauge PVR-500 tuner.
> - Serial IR blaster, StreamZap remote.
> - 500GB Seagate SATA/300 system disk (i.e., /boot and /) under ext3.
> - 1TB Seagate SATA/300 MythTV recording partition (i.e.,
> /var/lib/mythtv) under LVM2 and JFS.
> - Mythbuntu 8.10 x64 with the latest updates (i.e., kernel 2.6.27-9
> SMP) as a combined FE/BE box.
> - Anything else you'd like to know?
> 
> This system ran solid as a 32-bit system with Mythbuntu 8.04 and 4GB
> RAM---kernel saw only about 3.1GB---before I "upgraded" it.  After the
> upgrade Live TV started to freeze after anywhere from a couple of
> hours to a couple days.  More accurately, it'd freeze after about a
> couple of hours of watching Live TV following a fresh boot or a few
> days of up time.  When it freezes the system wouldn't respond to any
> keystroke (e.g., Ctrl-Alt-Del, Alt-SysRq-<whatever>, etc.).  Some
> times existing SSH connections would still be alive, but I wouldn't be
> able to initiate a proper shutdown/reset because processes would hang.
>  And the kernel trace would always be the same:
> 
[snip]
> It's always the mythfrontend process that gets stuck and it's always
> when it's the middle of handling a page fault while doing some TLB
> operation.  That's really the only reason why I ever think of I may be
> hitting the AMD TLB bug (i.e., the mention of TLB in the trace).  Of
> course, I could be completely wrong and this has nothing to do with
> that bug.
> 
> In any case, a few days ago the system which had been running quite
> stable with the BIOS TLB Fix option enabled again started to freeze
> rather consistently with the same trace as above.  Currently, I
> throttled down the number of CPUs once more down to 2 via "maxcpus=2"
> and it's been running stable.  For now.  I left Live TV on for hours
> over the weekend and it didn't freeze.
> 
> FWIW.  If you need more details, just ask and I'll provide what I can.
> 
> Thanks.
> --
> Francis
> echo "sunegbwb at tznvy.pbz" | tr '[a-z]' '[n-za-m]'

To quote a popular net phrase: ZOMG!!1!!

I just looked at some of the notes I took when I was having these
problems and they match up almost exactly with your description. I think
I am going to have to look into this bug a bit more.

However, since I did those updates / patches (see messages from last
week) I have not had a single mythtv related problem (neighborhood lost
power and the battery backup only gave me 2 of the 10 minutes *shakes
fist in anger*). So I think that the updates fixed it, but it can't hurt
to look into this bug and see if I can find more information on it.

Also, as I side note: I used to have a script that would run every
coupld of hours to update the time. If left unattended in 64bit mode the
time would drift by ~15 sec/day. NTP had issues keeping it in sync. It
kept perfect time in 32bit mode. Since the update it drifts maybe 2
sec/day. NTP seems to be working well.

Thanks for the info
Stack


More information about the mythtv-users mailing list