SGI: Hardware

A followup to this thread . I have since finished reworking the IO4.

The DS1397 with exposed battery contacts:

Here's the final result:

Full story and more photos HERE .

And the best part: the Onyx LIVES

Very nicely done! Thank you for taking the time to write that up and post it for the benefit of others.

Just wanted to point out that anyone else doing this needs to cut at least one battery contact or you'll just drain the new battery across the old one - it's obvious you did this and you mention it in one photo caption - just not in the write-up

Excellent photos - I'm about to have to do a similar mod to my Indy once (if?) I get it back out of storage again. Mine probably won't look as professional

bri3d wrote: Excellent photos

I love my Nikon D90

bri3d wrote: I'm about to have to do a similar mod to my Indy once (if?) I get it back out of storage again. Mine probably won't look as professional

The 'professional' look is because:
1) I did some investigations so I knew beforehand where to open the Dallas.
2) I bought a 3mm grinding stone for the Dremel rather than use whatever I had handy and butcher the thing.
3) The SMD CR2032 socket is simply made for mods like this because it sits comfortably on top of the Dallas for most systems, or on the PCB next to it in case of the IO4. (It would get in the way of IO4 mezzanine cards if installed op top of the IO4).

The CR2032 sockets were only 1 EUR I believe and I ordered 10 immediately because I expect to rework all of my SGIs over the coming years. Thinking about it, I should have ordered 20

My first experiment was a SUN SparcClassic. I expect Indys and Indigo2s to be similar.

There's a second DALLAS chip in every Onyx/Challenge, on the System Controller. I've seen both DS1287 and DS12887. In my Onyx IR it has run out, so I modded it:



The Onyx stores the system serial number in both the IO4 and the system controller. If you replace either of those one with a blank specimen, it is written automatically. If IO4 and system controller have different serial numbers, the system controller 'wins', unless you start up in maintenance mode with the debug bit set -- then it doesn't do anything. Still, a system controller serial number can be set only once. Interestingly, if you mod both system controller and IO4, the system has a blank serial number, and you can set it (as described in the Challenge/Onyx Diagnostics Roadmap). Pulling the CR2032, you could set it any number of times

Unfortunately, the system controller doesn't seem to initialize the RTC of the DALLAS properly, so the time on the LCD is garbage (IRIX is fine, probably uses the IO4). It looks like it doesn't start the built-in oscillator, or there is confusion about the time format (BCD vs. something else). The NVRAM clock registers are not touched by the infamous 'zap' command in POD mode, which will clear other NVRAM data such as the serial number, the event log and PROM password. I'm thinking of getting some sort of 8051 embedded development board to talk to the DALLAS and find out what's going on here.

Oh, reworking a DS1287 is described in detail HERE . The black hotglue I used to close the DALLAS is more heat resistant.