Pentium 4 "Northwood" not testing, and etc.

It has been a while. I have been testing FreeBSD-12.0 on a Pentium 4 "Northwood" and it's D865GBF motherboard for a while. It seems P4 "Prescott" and "Cedar Mill" will work with the FreeBSD-12.0 (Alpha, BETA, and RC's) but not "Northwood."
The EM64T feature seems to be essential and my box does not have it.

Otherwise, I have been resolving some conflicts at home and at work, so it has taken some time and money to resolve them. (Laundry room, washing machine, plumbing, etc. etc.) Anyway, thanks for listening to my two cents worth.
 
Thanks for the info. FYI I have a Pentium 4 which runs every FreeBSD version I've tried, up to and including FreeBSD-12.0-RC3-i386, with fewer problems than I've experienced with amd64 installs on off-brand laptops. I think it might be a later cousin of Northwoods, but I'm not sure, and don't really know the details of the lineage of these Pentium 4 processors. I know nothing about the EM64T feature. BIOS says...

Code:
Processor Info:
Processor Type    Intel (R) Pentium (R) 4 CPU 2.80 GHz
" Clock Speed     2.80 GHz
" Bus Speed       800 MHz
" L2 Cache        1 MB
" ID              0F34
Hyperthreading?   Yes
64-bit tech?      No

System Info:
System            Dell System 4700
BIOS Version      A10 (01/04/06)
This system refuses to run the amd64 version of the install, saying something like "Long mode addressing not found" or "Long mode addressing required." I can't recall the exact text of it.
 
Phishfry: I was able to avoid kernel panic and progress further
into the bsdinstall sequence using your advice to turn off "ACPI"
on the boot options; (when the .iso disk would boot at all).
https://forums.freebsd.org/threads/disabling-acpi-when-booting-from-iso.68584/#post-408735
Thanks.
Further, adding kernel modules from a shell prompt
(manually) also made the box more functional.

Others have described the 32-bit processors as sailing off into
the sunset; much as, deprecating drivers which few if any are
using on their boxes anymore. So from a long range
perspective I can understand this. The security risks are no
doubt a large factor - many have described risks in other posts.
 
Turning off ACPI totally is the nuclear option. You should at least enable ACPI-APM because this controls power states of the CPU.
Especially valid on older CPU's with no internal thermal trips. ACPI also controls the fans. So best case scenario they run %100.
Worse case.....
I now wonder if I should have even commented about the option. I would use it very carefully.

There are probably less security concerns with these older CPU's because they don't have all the Spectre vulnerabilities.
You only have to worry about i386 going away.
You know newer is better (supposedly)?
 
The "em0" driver and interface do not function on the many tests and workarounds I have used recently for FreeBSD-12.0. (Including the previous Alpha, Beta, and RC's).

This is an old and refurbished machine - Pentium 4 "Northwood".
I am ready to chalk this up to system hardware failure.

Finally, a Bugzilla report with ?id=229432 details a new patch
for the /etc/dhclient.conf that affects "em" drivers. But using bsdinstaller I only have a Read-Only privileges and any attempts using the shell prompt are not saved. Thanks, only 2 cents.
 
If it helps, I have a Pentium 4 that is 64-bit but is unable to run Windows Server 2012 because it doesn't provide SLAT (If that helps identify it somehow ;)).

I use this machine as one of my main test machines running FreeBSD 8. The main reason why it runs 8 is because there was a regression in FreeBSD 9+ that causes issues with ACPI. The main symptom is that powerd is unable to reduce the CPU clock speed causing it to always run at 3.2GHz.

I never quite found out a solution and because there were no other reports of similar, I assumed it was just my specific setup so never reported it myself. Plus I am actually very happy running older versions of software because it is offline and because I could keep with a pretty decently working Gnome 2 (that was broken around ~9 plus the whole Gnome 3 bullsh*t).
 
There does seem to be some problems with the em0 recently. I just tried to help another user and it ended in failure.
https://forums.freebsd.org/threads/freebsd-gui.68307/page-2#post-407701
Luckily he had wireless to fall back on.
Odd thing is not all Intel em(4) ports are affected. I have some that work fine. Had none that didn't work right yet.
To edit your FreeBSD memstick installer you must remount in rw mode.
From LiveCD shell you can use: mount -o rw /
Then you add your settings to /etc/dhclient.conf. You could also mount the memstick installer from another machine and modify the file.
That allows you to customize your memstick installer.
/mnt/etc/dhclient.conf
Code:
interface "em0" {
  supersede interface-mtu 0;
}
 
This has proven to be a good lessons learned thread. Many detailed concepts: boot sequence, ACPI, 32-bit processors,
have enriched my understanding of FreeBSD. Thanks to all.

EDIT: A fourth rebuild of this refurbished PC was given it's third motherboard in 16 years;
but sadly only going to be used as a beta-testing and games only box!
 
Last edited:
Back
Top