[CB_HK]

Special thanks to Raion, Dodoid, Jan-Jaap, BackPlaner, and everyone else who has provided assistance, insight, or humor along the way!      
     
Introduction

I recently acquired an Onyx2 from Homeboys Electronic Recycling in Los Angeles. You may be asking yourself, doesn't this seem familiar? Didn't I just see a post about an Onyx being restored that was acquired from the same place? You would be correct! Though the difference is that this is going to be the restoration log of the machine I picked up, which just so happened to be sitting next to the Onyx that BackPlaner purchased. After conversation, he and I are in agreement that these machines were likely kept in the same filthy closet or storage area for who knows how long.

This is my attempt at documenting my restoration--something I neglected to do as I rebuilt and restored the Crimson I acquired last year. I realized too late that I had done the community a bit of a disservice by not showing all the issues I ran into and how I had to go about fixing them (though I have a good thread on PowerOne PSUs and how to fix those). This time I'm going to walk through my restoration of this Onyx2 and hopefully capture some of the things I've learned along the way. I'm also sure that I'll have plenty of questions as I go and hopefully the community can chime in and assist if I'm stuck.

Enjoy!

Part 1: Getting the Machine Home



Loaded up in the back of the rental SUV. I pulled the front panel off to make sure it wasn't damaged during transit.

I had originally planned to remove all the skins from the machine before transporting it. I'm glad that I didn't. Having worked on the Crimson prior, I knew that nothing was ever as simple as it first appeared. Even though the users guide shows the panels clearly and in a way that makes it seem that they pull off easily, they do not. I opted to be as careful as I could be given the fact that all the plastics were intact. Something I was not expecting. You can see in the photo above the front panel has a number of scratches on it and the entire machine was coated in a thick blanket of dust and debris.



Excerpt from Onyx2 Deskside Workstation Owner's Guide (Document Number 007-3454-004)

Once home I was able to take a look around the system and see a bit better what I was dealing with. This had been a gamble for me. All I knew at this point was that the machine had been tested for power and that it seemed to function. Looking at the photos on eBay I realized that this machine could be decked out pretty well, especially since it had two node boards installed.



The front panel is held in place with one screw at the very bottom, in the middle. Once removed, the panel slides down approximately 4mm and can be pulled out. Check out the Onyx2 Deskside Workstation Owner's Guide, page 34, for specifics on this.



Empty drive bays and no sleds unfortunately. Easy enough to rectify, however. But I will need some blanking plates too. The switch on the MSC no longer clicks in each position; irritating, but not the end of the world. It will be replaced if it doesn't power the system up correctly, though.



Removing the top covers was relatively straightforward... AFTER I successfully got one out. The manual isn't very clear, however there are tabs that a screwdriver can reach through the top plenum that allow the caps to be pried up. I used a soft spudger tool to pry the top while a screwdriver gently pressed on the tab to release it from the underside.



Excerpt from Onyx2 Deskside Workstation Owner's Guide (Document Number 007-3454-004) page 37



The top plenum is actually two pieces that can easily be separated by tabs located next to each of the four captive screws. This was going to make cleaning much, much easier. Additionally, I was able to get a good look at just how bad the dust inside was. Everything that looks brown is supposed to be a dark plum color.



Removing the power supply from the front yielded an incredibly dusty fan tray assembly.



After a bit of encouragement, the fan tray slid out.



The RMs were next after the fan tray to come out. The GE and Node Boards would prove to be tricky until I received some assistance from Dodoid.



If you've never worked on one of these machines, this is an extremely important part to know. The right connector actuator has been undone, while the left has not. These are used to keep the boards' compression connectors snug against the pads of the midplane. If these aren't released, then the boards aren't going to come loose.



And here is the first Node Board out! Looks like low-density 64MB RAM which means this board has 1GB of RAM on it. The second board would prove to be identical.



Eventually I removed the rest of the boards and side plastics. Thanks to Jan-Jaap for his guidance. The panels just need a good tug upward to get them to unseat. I doubt the ones on mine had ever been removed before. Now it was time to tackle the bottom tray of the chassis.

I'll stop and make a note that may help some folks. The bottom of the chassis comes off with four philips head screws. You may also notice that the casters, which are a known issue with these machines, can be accessed. It appears to be as easy as removing the bolt holding the screw in, popping out the caster and replacing it. Luckily mine are in one piece. If you happen to have one of these machines and your casters are broken, fear not. They appear to be easily replaceable.



Closeup of the nut holding in the caster bolt.



Slightly out of order as the side plastics aren't off in this photo and the bottom tray hasn't been removed yet. On the table you can see the two Node Boards (both R10K @ 180 MHz with 1GB of RAM) as well as the graphics setup. After cross checking part numbers, it appears that this system has an InfiniteReality 2 boardset installed.

---

Part 2: Cleaning the Plastics

Part of restoring, at least for me, is bringing the system back to as close its original appearance as possible. Obviously some things just can't be fixed: deep scratches on plastics, components that have worn out, etc. However, a clean and polished set of plastics is pure motivation for me. It's what helps me see what the machine will look like when it's done and it gives me a sense of joy.

Everyone has their own way of cleaning these machines. I take no credit for my process insofar as I have borrowed techniques from other folks--some of whom had posted to Nekochan about their experiences with removing deep scratches, scuffs, and dirt.

I use a combination of products: Simple Green to clean all the plastics, Goo Gone for any left over stickers, melamine sponges to work away any scuff marks, Meguiar's Ultimate Compound for polishing out scratches and making them much more difficult to see, and finally, Aerospace 303 as a top coat to bring back the luster of the plastic.

For this project I had to include Novus plastic polish (both Product 1 and Product 2) in order to clean up the front panel. I didn't want to haze or scratch the slightly shiny surface with Meguiar's.



The products that I prefer to use. The Dry Teflon Lube is excellent on squeaky casters (which mine were) as well as moving plastic parts. It kills any obnoxious squeaking noises and sticks to surfaces as a dry lube.



I'm incredibly pleased with how well I was able to restore the plastics. There will always be scratch marks if you look closely enough, but overall the system looks practically new.



Every panel was removed from the machine and every piece was cleaned in the same way, leaving it looking almost new. All my machines get a coating of Aerospace 303 once every few months. It's a great way to also protect them from UV damage if the machine is near a window.



So here's where things sit. The plastics have been cleaned and polished. The components have been removed and looked over. Next up will be a good cleaning of the chassis with compressed air. Right after I get the midplane out.

Coming up - Part 3: Cleaning the Chassis

I'll be working on cleaning up the chassis next. After using compressed air to remove all the dust and debris I can, I'll be looking for any signs of corrosion (spoiler alert: there are a few spots) and will be applying a rust killer along with a protective top coat to those spots.

Once the chassis is cleaned up it'll be time to look at what's going to go back into the machine. As it stands, I'm planning on swapping out the R10K boards for some R12K boards, courtesy of Doug Mashek. I'll also be installing 6GB of RAM instead of the 2 that was in there to begin with. I need to disassemble the power supply and clean it out. I can barely see any color other than brown dust inside. Once that's done I'll look at putting everything back together. It all sounds so easy, but I'm prepared for plenty of hiccups along the way. Hopefully luck will be on my side!

If you've read this far, thanks! I hope you've enjoyed taking a look at my little project. I'll continue to update as I go so be on the lookout for more photos and details!

[jan-jaap]

Wow, that cleaned up really good!

A few random things you might want to know that you won't find this in any manual:

1. The screws that keep the backplane in the chassis are numbered, 1/22 ... 22/22 and they are longer than most others used in the system. Keep them separate.
2. The nodeboards and IR boards have an EMC gasket on the left side (the 'fingers'). Because of this, it's best to install boards left-to-right, otherwise the edge of a board will mash the EMC gasket of the board to it's right side. There's a divider between CPU and gfx side so there's e.g. no need to undo the CPU boards to replace a DG5.
3. Tightening the large screws is not the preferred (or only) method to install the boards because it will "bend the ears" as can be seen on your RMs. Gently "wiggle" them mostly in place by pushing the top/bottom black handles in turn, and use the screws only to secure the boards.
4. If you install multiple boards at once, don't fasten anything until they're all in the card cage.
5. For CPU and GE boards: first put them in place, then fasten the 7/64" hex bolts and finally the large screws.
6. That "SCA" connector used on the fan tray is it's weak point. Clean it well with some alcohol and an old toothbrush. The MSC can tell you which fan(s) failed (if any). They may be actually dead, but often it simply that connector. Clean the backplane side of it while it's accessible.
7. IIRC that PSU doesn't really come apart all that well. The wires between the PCBs inside are soldered.
8. If your existing nodeboards were 180MHz you need to change some jumpers on the mainboard before installing the R12K nodes.
9. With the MSC out, inspect the brown flex cable connecting the CDROM. Early models connected only one of either 5 or 12V, and that doesn't work with newer drives. Since you have the early "silver heat sink" R10K CPU boards this may be the case. A mod is possible if necessary.

If the MSC key can turn by more than 180deg it's broken. This is fairly common. I've seen plenty of MSC keys that turn really light and give only the tiniest amount of clicky feedback when in the 90deg position. That's nothing to worry about as long as they work otherwise.

My first Onyx2 was a similar mess when I got it. I did a complete tear down and rebuild. IIRC I actually hosed down the metal chassis after I had stripped it.

Thanks for the tip about the dry lube. All my desksides squeak. The Onyx2 also leaves black marks on the floor if you move it such that the wheels don't roll (e.g. turn it around in place).

[CB_HK]

Excellent points, jan-jaap! I'm curious as to which jumpers I'll need to modify. I see a few on the back of the midplane. Any reference I can look at or do you happen to know off hand how I need to set them?

As luck would have it, as soon as I got the midplane out I found two problems I'll need to address. One of these may very well be the reason this machine was relegated to a closet.



Yay for exploding tantalum capacitors! I have a suspicion that this may have been the reason this machine sat for so long. Does anyone know the value of this capacitor? I can't extrapolate based on the others near it since they vary with no pattern I can see easily.



There should be a 330 Ohm SMD resistor sitting on the top pad. I know the value because it popped off when my hand brushed over it and I later found it on the floor. I've never had an SMD component come off that easily. I'll need to clean up the pads and replace that item as well.

Thankfully I don't see any other issues with the board. I'll have a closer look tomorrow and see if anything else is hiding.

[jan-jaap]

Excellent points, jan-jaap! I'm curious as to which jumpers I'll need to modify. I see a few on the back of the midplane. Any reference I can look at or do you happen to know off hand how I need to set them?

The jumpers are D4M0 and C2L2, behind the nodeboards.

The following table shows the setup of the jumper sets as seen from the back of the machine:

Code:

              180 MHz           Other

D4M0           [.|.].            . . .

C2L2            . . .           [.|.].

C2L2 installed, D4M0 open: valid for everything except 180MHz nodes (photo: sgistuff.net)

This info should be in that 300+ page scanned "quick reference to the Onyx2 / O2K series" PDF, the one with the purple cover.

The part# on the nodeboard should be able to tell you what you have right now.

I have a spare Onyx2 deskside backplane, I'll have a look tonight and make some photos of the locations with the missing/broken components of yours.

Oh: to remove the midplane: only undo the screws 1/22 .. 22/22 and leave the rest alone, they secure components to the midplane and aren't necessary to remove the midplane from the chassis.

[jan-jaap]

Midplane photos. "102" = 1kΩ, the capacitor is 33 µF / 25V

[BackPlaner]

Excellent work so far C! I'll have to get going to catch up with you, I'm almost done taking it all apart and cleaning everything. My skins also cleaned up really nice, I was quite amazed as were you.

Keep up the good work!

[CB_HK]

Excellent work so far C! I'll have to get going to catch up with you, I'm almost done taking it all apart and cleaning everything. My skins also cleaned up really nice, I was quite amazed as were you.

Keep up the good work!

Thanks much! I've found I have a bit of an addiction to the restoration process. It's very satisfying to watch something neglected come back to life and look close to new. I get why car guys enjoy it so much!


Part 3: Cleaning the Chassis

As with any project, it's always wise to do as many repairs as you can, including proactive things, so that you don't just have to pull it all apart again soon after when something else breaks. Part of the process I enjoy is making sure everything is as clean as it can be. Even though the plastics are what will be seen, the inside should be nice and clean as well!



I took Jan-Jaap's advice and separated the chassis for a good washing. The nice part about being in the desert is how quickly things will dry. Even so, I used an air compressor to blow off as much water as possible after scrubbing both parts.



You can see how much gunk is on the frame still. This is AFTER I used the compressor to blow off as much dirt and dust as I could. Whatever this gunk is, it didn't want to come off easily.



After hitting the chassis with Simple Green and letting it sit for a few minutes, the dust and dirt started to lift. Thankfully this made it fairly easy to spray down and only light scrubbing was needed.



The black rings you see are where the fans sit and pull air through the chassis. A lot of debris collected here. Also, a lot of moisture. I'm guessing that this system was used someplace where there was a bit of humidity. If it came from somewhere in Los Angeles, where I picked the machine up, that would make sense. I'll likely never know where the machine spent its previous life, though.



After a good wash and dry it was time to get rid of the rust! I used this product when restoring my Crimson and found it was very effective.



I covered both the top and bottom air intakes with the paint. It won't react with the metal that isn't rusted, so there's no issue there.



A few days of the rust converter drying and then a top coat of Rustoleum silver paint. This likely doesn't add much that the converter hasn't already done, but I feel better about making sure things are well protected, especially after rust was an issue before.



The IO6G had seen better days, at least as far as rust goes.



Thankfully it was just surface rust and the metal hadn't been badly pitted.



5 minutes with a dremel fitted with a wire brush and voila! Thankfully the rest of the board cleaned up well with an alcohol spray down and some compressed air.


Miscellaneous



The fan tray was scrubbed down with alcohol and a soft brush. All the fans were blown "clean" and then the remaining dust and dirt was gently scrubbed off with an old toothbrush.

Once all that was done, I made sure every fan had a few drops of oil added to their bearings.



The R10K boards that came with the system were maxed out with 1GB of RAM on each. These were sent in as a partial trade with Doug Mashek.



The replacement R12K 300MHz boards with 4GB of RAM installed! I sourced the RAM separately from eBay. The second node board will have 2.5GB of RAM with the same 256MB sticks.

Coming up - Part 4: Reassembly

The next part will likely be split into two updates. I'm currently waiting on a few parts to arrive from Raion and Jan-Jaap (Thank you both!). After reassessing the condition of the power supply I decided it would be a good idea to snag a known working one. I'll post some comparison photos showing how banged up the one in this system is versus the replacement. I also opted to snag a replacement midplane. While I will be fixing the two issues with the one that came with the system, this is another case of using a part that is known to work instead of gambling with an unknown. Even so, I'll keep it around as a backup just in case.

I'm looking into replacing the caster wheels. While mine are 99% intact, there are signs of chipping and cracking. I think it would be well worth it to replace them now, since getting to them later will require removal of all plastics and separating the top chassis from the bottom caster tray. If all goes well I'll be sure to post an update with that process!

A question regarding power

The below excerpt is from the Onyx2 Workstation Owner's Guide and covers how the system should be provided power given the boards installed. The manual is fairly clear about what conditions require 110v @ 20 amps versus 220v @ 16 amps.



I've read posts and spoken with folks who have both Onyx2 and Origin 2K systems who state they run their systems off 110v, regardless of the number of node boards or RMs installed.

I'm curious as to what the best answer would be. I have a dedicated 110v 20 amp outlet I can use. There is only one outlet on that circuit so I'm not concerned about overloading it. However, I'm wondering if the issue is more specific to the Power Supply. I'm assuming that 220v is used to provide more juice for the more demanding loads placed on it with a fully kitted out system. Is that the case? Or was SGI trying to prevent folks from getting upset about blowing circuit breakers when they kept pulling a full 20 amps with a 2x node and 2x RM system?

---

Added: One thing I almost forgot! I found this guy tucked away on the IO6G. It looks fairly similar to a Dallas RTC. Any thoughts on whether this needs to be replaced?

[CB_HK]

Fixing the Door Mechanism

One of the things that I decided would have to be fixed early on was the sliding door mechanism. The Onyx2 has a clever way of covering the MSC, power key, CD-ROM drive, and hard disks by way of a sliding door. The only problem with mine was that it made a terrible noise when it moved AND it wouldn't lock in the open position.

While I had the front panel off for cleaning, I made a point of lubricating the guide rails and toothed mechanisms that guide the door up and down. I also discovered the latch assembly which had broken.



This piece was missing the grabber arms which would hold onto the door rail as it bottomed out in the track. Without the arms, the mechanism was completely useless.

To say that finding a replacement was not going to be easy might be a small understatement. Even so, I wasn't about to just give up on something that I knew was going to irritate me.

After a week of Google searching and attempting to image compare shots of the latch, I stumbled across a company which looked promising. So promising, in fact, that I broke out my calipers and started comparing measurements. I have no idea how or why a company would still be producing this rather bespoke latch more than 20 years after the Onyx2 was designed, but I'm glad that there is still demand for it.

Enter ISC Plastics out of Spain.



Broken assembly on right, and a brand-new manufactured version by ISC on the left!

Once the new assembly arrived I was extremely pleased to find that it was indeed a direct replacement piece!



The guide rod will bottom out into the latch assembly, while the latch arms lock into the notches of the guide rod, holding it in place.



The guide rod locked into place.



Now the door will stay in the down position!

Look out for more updates as I continue to have parts roll in!

[jpstewart]

Wow, that's attention to detail in the restoration!  I'm amazed at the the lengths some of the members will go to, to get things just right.  I don't think I'd have the patience to get through such a detailed process.  Great job tracking down that part!

[CB_HK]

Thank you! I definitely have a thing for how SGIs look and restoring them to their former glory is a very rewarding experience for me.

Check back later when I get the rest of the hardware I need and get it all squared away!