DoN. Nichols wrote:
On 2009-01-02, Christopher Tidy wrote:

Hi Don,

A while back you gave me some advice about fixing my ailing Sun Ultra 2
workstation. Well, shortly after that it died completely. I posted
another reply in the "Shopmade grinder with winch" thread but I had to
use Google Groups. I think you may have missed the message if you're
blocking messages from Google Groups.


I don't always read all threads to the end. That was one which
I skipped over a lot of, when I was overloaded with articles to read.

That's fine. It happens to all of us!

And I'm going to be heavily occupied Saturday and Sunday,
probably not even get to the newsreader on either day -- or perhaps near
midnight on Sunday.


The same thing has now happened to me twice. With the first workstation,
it worked for about 10 days after the clock stopped keeping time. This
time, it worked for about a month. It leads me to believe that Ultra 2
clock problems are perhaps most often caused by motherboard faults as
opposed to a dead NVRAM battery. Anyway, I have now managed to build a
working Ultra 2 from parts with 2048 MB RAM. 2048 MB seems to make the
Ultra 2 slightly faster than it was with 1280 MB. I notice it when the
machine is booting up, which surprises me as I would have expected boot
speed to be limited by hard disk speed.


When it is booting it is loading a lot of programs. With plenty
of RAM, it doesn't have to swap just loaded programs to disk while it is
loading the next one.

I didn't expect it to make a big difference, because I thought that most
of the programs being loaded into RAM during the boot process would be
staying in the RAM. Also, once my old machine was up and running, it
often didn't use all the of the 1280 MB RAM. But I don't know of any way
of looking at RAM usage during the boot process.

I've noticed a significant improvement in boot speed (and in
workspace startup speed) with my boost from 3 GB of RAM to 6 GB, and
want to go to the max of 8 GB for my Sun Blade 2000 to have plenty of
RAM for when I'm doing lots of image processing (which is what is coming
over the weekend, with someone else critically involved in the project
who can't get up here very often.


This means that perhaps my old NVRAM chips are worth preserving. You
mentioned a piece of software which could be used to stop the NVRAM
clock. I searched but couldn't find it online. Looking at the NVRAM
datasheet, it seems like it is not something which can be done with a
crocodile clip. The only reference I could find was for modifying the
contents of an ancient sun4c machine's NVRAM at the "ok" prompt, and I
couldn't immediately see how to use this information with a sun4u Ultra
2. Do you know of a source for software which can be used to stop the
NVRAM clock?


I don't remember for sure where I read about it, but it was
probably in the first hit on a Google search for "Sun NVRAM FAQ"

http://www.squirrel.com/sun-nvram-hostid.faq.html

That's the information that I had seen, too. I was just thinking that
there might be a piece of software which makes the process easy, but I
could not find any mention of it on that page.

This one was posted in February of 2004, FWIW.

And I see that the procedure only mentions the Sun 3/80 (68030
CPU) and the sun 4C (SS1 SS2 IIRC).

I'm not sure whether it would work on the Ultra-2 but I don't
see any reason why it should not, if you can get to the "ok" prompt.

I tried the method intended for sun4c machines on my Ultra 2. I get the
error "Fast Data Access MMU Miss" after the second command.

My best guess is that the addresses for the Write and Stop bits given in
that document for the M48T02 chip are not the correct ones for the later
M48T59Y chip, which contains more memory.

From the M48T59Y datasheet I can understand the structure of the
register in which the Write and Stop bits are held, but I cannot
understand the syntax used at the "ok" prompt (or perhaps not: the
prompt is indicated by ?) to modify the contents of the register. I
cannot find any information online about this syntax other than the
NVRAM FAQ, which does not explain the syntax used to program the NVRAM
at bit level, as opposed to byte level.

I'm sure that someone with more knowledge could figure this out, but I'm
reluctant to experiment with my working machine in case it ceases to
work. It also seems that the two failures which I thought were caused by
dead NVRAM batteries were not, so perhaps the M48T59Y has a longer
lifetime than the M48T02. There do not seem to be any documents
referring to NVRAM problems with the sun4u machines using the M48T59Y chip.

Check the contents of your NVRAM chips (HOSTID and Ethernet MAC
address) and keep a hardcopy version in case you need to perform the
battery surgery which is documented in the website above.

For both the working motherboards I have, I've written down the details
displayed on the screen when the machine boots, such as the serial
number, ethernet address and host ID. For the machine I'm using at the
moment I've also saved a copy of the output of the command "eeprom -v".
Whether or not I need the extra details such as the system board serial
number, I don't know.

I think that's all there was of importance in the post. I speculated a
little about whether common applications such as Firefox and Nautilus
are able to make good use of multiple processors on Sun workstations.
I'm still not certain. But I think that's all.


I depends on whether they are multi-threaded. Though they will
probably use other CPUs for plugins.

I did a bit of reading online and came to the conclusion that Firefox is
multi-threaded, but that one thread handles the user interface,
JavaScript, layout and image decoding. Probably this thread sometimes
gets bogged down!

Thanks for the advice.

Best wishes,

Chris