The Quad Chronicles - Restoring a HOT G5 Quad

[eyoungren]

Well... I may have hit the end of the road. The CPU pad intake barb, which sheared off while I was removing those ^%$#@& metal clips Apple so considerately put on each hose end, is proving to be impossible to reattach with any form of epoxy, Super Glue, etc. It breaks off again every time I try to attach a hose to it, or if I pre-attach the hose, when I attempt to attach the other end of that hose to the radiator. I am trying one last bit of serious MacGyver'ing to overcome this, but if that fails, I have no place else to turn.

You might try hot knifing it. Might be hard to do and impractical but at this point maybe worth a shot.

What you do is is heat up a knife over the stove until it's real hot. Probably want to use a pocket knife that you don't necessarily care about messing up. You certainly do NOT want to use a table knife or any other knives intended for food.

Anyway, the idea is that like a 'hot knife through butter', a hot knife through plastic cuts easy. In this instance I'd think you'd want to get one end of the two broken sections nice and soft with the hot knife. Then, you'd press them together and hold. With any luck the two ends would rebond as the plastic cools and solidifies.

I once put a new radio in the dash on my car, but it was a 1980 Nissan, so I had to cut out the plastic section that held the old radio. A friend of mine hot knifed it using his pocket knife. Made it real easy to just slice through the plastic without using a saw and kept the edges from being sharp.

So, that's my thought here. Don't really know if it will work or not.

 

[Certificate of Excellence]

Is it something that can be 3D printed?

 

[mac57mac57]

@eyoungren, interesting idea. I thought about heating both ends and pressing them together as well... sort of like welding metal, but with plastic.

I am still worried about the lateral force on the reattached part, as a result of connecting a hose piece to it and then connecting the other end of that hose piece to some other part of the cooling loop. Making that connection exerts a modest amount of lateral force on the reattached barb, and with every approach I have tried so far, it has come apart at that time - epoxy, super glue etc. are just not strong enough.

I have ordered a new product called Plastic Weld, a plastic-specific sort of glue, that says once cured, it will withstand up to 600 PSI of lateral force. I will try that next, but if it too fails, I will try the "hot knifing" approach.

Last but not least, I could extract the equivalent part from the other CPU assembly and try to get a replica 3D Printed. That option is likely quite expensive, so it might have to wait.

Thanks for the suggestion.

 

[Certificate of Excellence]

Check THIS out. It’s a cool DIY of aftermarket pumps circa 2021 that this guy installed in his G5. Looks fantastic. This or a conversion to air cooled could be a fun journey. Hopefully though your plastic weld reacts properly with the plastics of your pump. I am a fan of and use plastic weld for repairing things (lately my kids toys and my oldests remote model helicopter which got in a fight with a pine tree) It is hit very dependent on the type of plastic. I’m hoping it reacts & bonds with you lcs bits.

 

[mac57mac57]

Thanks @Certificate of Excellence! Very useful link.

I have to admit that I am starting to think seriously about air cooling... SO much simpler mechanically, and much more reliable nr long term. The "up front" work is daunting, but it may be the way to go if nothing else works.

 

[mac57mac57]

Well... My final jury rig of the sheared off part failed miserably (leaked like a siege), so onwards to the Plastic Weld glue product.

Sigh... this is SO tough. I have never run into any task in my retro computing work that comes anywhere near the complexity and failure rate of an LCS rebuild.Sheesh!

 

[MisterKeeks]

I also think JB Weld's Plastic Weld product is probably your best bet for something like this. How'd the piece snap off? I didn't have the special hose clamp pliers, I just used pliers to lift the clamps above the plastic barb area. Then I believe I sliced the hoses open (i.e. I cut parallel to the length of the hose) and kind of peeled it off the barb. If you do this, though, you should slice the hose down to the barb, but don't try and slice past the barb because you'll slice into it and make things more likely to leak when the new hose is fitted.

You might consider fitting a hose to the plastic barb and clamping it down *before* you attach it with the plastic weld so that you can avoid applying force to that area. I would also recommend *after* the Plastic Weld is completely cured that you use a silicone sealant around the area (like this). The Plastic Weld will provide structural integrity, while the silicone should help more with leaks. I don't know how familiar you are with automotive repair, but this silicone sealant is the main ingredient in "gasket maker" products that are designed to seal imperfections in mating surfaces.

3d printing might be an option but that's going to be a complex one. Modeling the part aside, you have two issues. One is printing it to the tolerances required to prevent leaks around the O-Rings. Two, 3d printed parts are not necessarily watertight.

 

[Certificate of Excellence]

3d printing might be an option but that's going to be a complex one. Modeling the part aside, you have two issues. One is printing it to the tolerances required to prevent leaks around the O-Rings. Two, 3d printed parts are not necessarily watertight.

Completely agree on the tolerances needed to prevent leaking. That is a tough one to get right with 3d printing. I do think replacing the aging LCS with a new one would be a cool project + you'd have convenient access ports and a fantastic story to tell. The aircool mod discussed HERE also looks like alot of fun. I have some metal cutting tools so I think I could machine a very good result. Alas, a quad is something I have never found locally and I would never pay to have one shipped, so Ive never had the opportunity LOL.

... Then I believe I sliced the hoses open (i.e. I cut parallel to the length of the hose) and kind of peeled it off the barb. If you do this, though, you should slice the hose down to the barb, but don't try and slice past the barb because you'll slice into it and make things more likely to leak when the new hose is fitted...

This is very good advice. I learned this the hard way LOL. About 15 years ago I was working on the vaccuum system on a w123 and the original line connectors were bloated and kinda gooey and stuck/glued to the lines so I used a box knife to slice up the connectors to get them off. After the job, I had a stupid leak that would not go away which in this car made the car really rough at idle when it started each morning until pressure was regained. I ultimately found the issue a month later - excactly this. I had been a bit to aggressive on the passenger front door and I had sliced into the line a bit too much which created a leak that would leak out each night. It was fixable but man what a PITA lol. From then on, I always used wooden toothpicks, skewers or golf tees to pull stubborn connectors off lines in that car. Slicing into a nipple like that is so easy to do especially on plastic or soft alloys like aluminum.

 

[mac57mac57]

Thanks all, this IS good advice. I have been very careful not to slice into the barbs... they have ended up scored a bit, but it is very surface level. I am keenly aware that I can't just hop onto eBay and buy replacement parts if I really mess this up (like shearing off a barb entirely - doh!), so I absolutely don't want to break anything else!

... and on that front, I am presently attempting to bypass the sheared-off barb entirely by directly connecting a hose segment to the CPU pad the barb is a part of. Given the shear forces that would normally act on such a connection as the other end was attached to the radiator, I applied @MisterKeeks suggestion in a slightly different context and created and attached a custom-shaped set of hose pieces so that there will be almost no shear forces - the resultant piece sits almost naturally on the CPU pad.

To create a hose piece that fit this need, I used one segment of the old Apple hosing, which incorporated a near right angle and then barbed that together with a regular piece of transparent hosing, which I connected to the relevant outlet of the radiator. Both ends were pre-clamped so that there is no need to fuss with this section of hosing once it is complete. I then "glued" the hose piece down with J-B PlasticWeld. This putty-like product sets in 20-25 minutes and is fully cured in 2-3 hours. The result looks like this and is curing right now:

In the photo above, you will see that I have also threaded a large twist tie though the CPU pad and used it to tie down the hose segment, so that I don't have to stand there with my hand on it for 20-25 minutes, applying the pressure needed to ensure that it cures in the original position.

The J-B Plastic Weld packaging SAYS that once cured, Plastic Weld can handle up to 700 PSI, so it SHOULD do the job, as long as it doesn't leak. I have J-B's Plastic Bond for that eventuality - if it leaks, I will create an encircling "collar" with the Plastic Bond and that should seal any leaks (... I hope).

I will let the Plastic Weld cure overnight, so no further progress will occur today.

In the meantime, as part of my last (failed) attempt at a new cooling loop, I found that bleeding even most of the air out of the system was incredibly difficult. This attempt WAS failing due to leaks, and I suspect that the leaks were introducing new air into the system almost as fast as I was bleeding it out, so that might be the issue.

Nonetheless, assuming the above barb replacement effort succeeds, the next battle will be trying to bleed most of the air out. My current loop design has only one fill/bleed port; the next one I do will have two, one in each of two different hose segments. I am guessing that one of these ports should be directly in the intake (return) line to the pump, and the other in the output (source) line of the pump.

If there is an obvious flaw in this thinking, I would be delighted if someone would point it out so that I could learn from that, vs. "the school of hard knocks"! Thanks!

More tomorrow!

 

[mac57mac57]

Well... it seems to be working! I hooked up the intake barbs of each CPU pad to the radiator, and then used my hand operated siphon pump to flush distilled water through the system, just to see if (a) water came out of the output barbs of each CPU pad, and (b) the same amount of water (roughly) came out of both output barbs. In both cases, the result was positive. Water came out of both output barbs, and it was approximately the same volume of water from each barb.

SO... this looks encouraging. I noticed a little leakage from around the Plastic Weld area of the newly attached hose (replacing the sheared off barb). I expected that this might happen. As I mentioned in a previous post, I applied J-B Plastic Bond epoxy around the whole thing, and am now waiting for that to cure.

Once cured, I will try filling the new cooling loop (this design sports two feed/bleed lines, one in each of the output line from the pump and the input line to the pump) and see if it will actually pump coolant around the whole thing. If it does, I may be done. If not, it will be time to carefully examine what is happening and devise a solution.

 

[mac57mac57]

Sigh... one more thing that didn't work! Under pressure, the newly attached hose which replaced the sheared off barb leaked copiously at the point where it was glued to the CPU pad. No "collar" of additional glue could stem the flow. So, that approach goes onto the trash heap too. At this point, I know dozens of things that DON'T work; I have yet to discover even a single one that DOES!

With two newfound "tools" however, J-B Plastic Weld and J-B Plastic Bonder, I am taking one last crack at glueing the sheared off barb back onto the CPU pad it sheared off of. The packaging says that the Plastic Bonder can handle up to 3770 PSI. Maybe, just maybe...

If this fails, I am out of options, and this particular G5 Quad becomes a "donor machine", to be used for parts. I will strip it down, save everything of any value, and put the carcass into a deep, dark closet, not to be removed until some magic new technology seems to me to have a hope of resurrecting it.

I do have one other Quad, and it does not have any sheared off barbs! Applying what I have learned from the current machine, I will very gently break down its LCS and rebuild it. I now have all the right tools and some painfully acquired experience to guide me. At this point, building out a cooling loop has become easy to me. Only the leak from the sheared off barb on the current machine has foiled all my efforts. Hopefully I can rebuild the cooling loop on the other Quad quickly, efficiently and without leaks. Time will tell.

One final note: don't decide to recondition your Quad's LCS unless you are REALLY dedicated to your craft. It will be an epic battle. I can tell you plainly that after nearly two months of work, with nothing to show for it yet, I am well and thoroughly sick of this task. However, I am persistent: I will not stop until I either succeed with at least one of the Quads, or run out of options on both of them!

 

[MisterKeeks]

It looks like you have way too much epoxy in that picture.

I didn’t realize JB Weld had multiple plastic specific products, I suspect the plastic bond product is more suitable for this. I actually used it on my hose barbs, but only to fill in a small gouge which occurred during hose removal. I would only use a small amount to start with. You could let that cure then give a bit more. I still recommend silicone after for leaks.

This would probably be easier if you removed the black plastic parts from the assembly.

What’s the red stuff inside the pump tubes?

I would not use a brass fitting in this system.

 

[mac57mac57]

Agreed, @MisterKeeks, I used only a small amount to start with, but then I smeared a whole lot more around the outside afterwards as a guard against leaks... it didn't work!

The red stuff inside the pump tubes is residue of my initial coolant fill, done with a blood red colored coolant. It was subsequently drained, leaving those telltale signs.

For this last attempt, I have in fact removed the black plastic part with the two barbs and have used only a modest amount of J-B Plastic Bonder to glue the sheared off barb back on. It is cured now and seems quite solid. I am going to do some preliminary leak tests this time before putting it back into the LCS.

I will also attempt to attach a hose segment to it BEFORE reassembly into the LCS. That way, I have easy 360 degree access while I am fitting the hose piece on, and if the barb does snap off again during this process, it is a lot easier to start over than it would be if it was already assembled into the LCS when it broke.

 

[MisterKeeks]

I’d be suspicious of any coolant leaving a residue like that, I don’t like the idea of that kind of particulate matter in the radiator

 

[mac57mac57]

Once again on this lengthy, tortuous LCS journey, I have managed to snatch defeat from the jaws of victory...

My efforts with the J-B Plastic Bonder were a 100% success. The reattached barb supported the installation of a new hose section and proved to be completely water tight.

I then filled the cooling loop, which also was a 100% success.

I then spent about two hours bleeding almost all of the air out of it. Again, a 100% success - there was almost no visible air in the system when I was done.

So what happened? I was just starting to seal off the two fill/bleed lines that are part of the loop when the pump itself started leaking copiously and then simply stopped working! Not a good time for a pump failure! 😡 It has not worked since.

This is frustrating beyond words... I finally solve the problem I have been working on for weeks only to run into an entirely new one. Air cooling is starting to look REALLY good, but I want my 2.5 GHz Quad to still run at 2.5 GHz when done. Most, if not all, of the air cooled mods I have read about involve using the 2.3 GHz CPU, not the 2.5 GHz CPU. This is a non-starter for me... I want my Quad to be a full Quad when done, not a speed reduced version of it.

Whats next? Back to "square one". I have a second Quad. I will attempt to replace its LCS cooling loop in as few steps as possible, and with as little physical jiggling of the system as possible. If for any reason THAT effort fails, air cooling will be the next step. A whole new journey, but one that is a little more in keeping with my skillset. As I said at the outset, I am NOT a plumber - this whole liquid cooling stuff is a nightmare, for me at least.

 

[Certificate of Excellence]

Sorry to hear about the bad outcome but I wish you the best of luck on your next LCS journey

 

[mac57mac57]

I'll say one thing - I have gotten VERY efficient at tearing down a G5 Quad, its LCS and rebuilding the LCS' cooling loop. I tore down my "other" Quad, extracted the LCS, tore down just the hosing sections that I had to (the input to the radiator and the two outputs from it) and rebuilt those using transparent hosing (with the addition of a feed/bleed line to the input to the radiator), all in about an hour and a half.

The result is a hybrid of Apple hosing (the two output lines from the CPU pads and their common return line to the pump) and new transparent hosing for the rest, all in the spirit of disturbing as little as possible while managing to flush the radiator (a bit dirtier than the first one I did) and remove the filters from the inputs to the CPU pads (they have to be restricting coolant flow at least a little).

With the new clamp removal tool, plus a trick I figured out with needle-nose pliers, removing the clamps from the hose lines I was replacing was much, much simpler and faster than before. As a result, no damage done to anything this time around.

Tomorrow I will fill the system with coolant, bleed air, fill, bleed, etc. until there is little to no visible air in the system, and then reassemble and try it out.

This SOUNDS simple enough, but Murphy has been dogging me the entire way on this LCS rebuild effort, so there may be more "surprises" in store. I am however keeping my fingers crossed - maybe Murphy will take a break tomorrow.

Stay tuned!

 

[MisterKeeks]

I am super curious what happened with the pump. It might be repairable. I would do an autopsy of it. They can be disassembled, I opened mine so that I could replace the o-ring in it.

Where did it leak from?

 

[Eriamjh1138@DAN]

Just because the pump leaks doesn’t mean give up. …Unless you’re trying not to spend any money at all.

It’s not about the value of the machine. It’s a project. It’s a restoration. It’s sweat equity. It’s your time.

There’s some good rebuild efforts out there and people who have done what you’ve done. Don’t give up.

If you haven’t already, read all the G5 threads on the 68kMLA forum for the G5:

G3/G4/G5 Power Mac, iMac & eMac

From the Beige G3 to the quad-core G5, via various iMacs!

68kmla.org

Keep searching for LCS rebuilding on the web. Look for parts machines. People are throwing these things out because they are “old”. And some are asking hundreds of dollars.

Hell, I have two of them. I should sell mine, too. PM me if you’re interested in a bargain. Local only. I’m not shipping it, but I might be willing to meet you halfway if you’re not too far. I’d like my G5 to find a good home.

 

[mac57mac57]

Yesterday, I reported that the pump of my LCS had failed. I can and will open it up and see if I can repair it, but I have read, somewhere in these forums, an account of replacing the stock pump with a new one.

Would anyone reading this be able to point me to a compatible pump? Since the pump seems to be the final failure in this instance, replacing it might bring the Quad back to life.

Thanks!

 

[Eriamjh1138@DAN]

I remember reading something about how these pumps were “special” meaning not very common. It was on the 68kMLA board. I’ll see if I can find it, but you’re welcome to keep trying as I’m juggling a lot of projects right now at home (all not computer related. Stupid house…).

 

[mac57mac57]

A little positive progress to report today.

I finally managed to build out an LCS cooling loop, got it filled and bled most of the air out of the system, all without any leaks. It proved impossible to get all of the air bubbles out, but I got most of them. I then reassembled the LCS into the Quad and, fingers crossed, fired it up. It started at least, gave a robust startup chime and booted. The fans stayed at a low speed initially, creating the expectation of success, but Murphy couldn't resist the opportunity to throw another curve ball at me - my monitor would not display anything, so I could not see what the CPU temperatures and fan speeds were. I restarted several times and finally swapped in another monitor. With the new monitor, I could finally get a display.

I didn't like what I saw. CPU A was running in the high 50s C, but was stable at that temperature. OK, but not great. CPU B was running in the high 60s C and was slowly climbing. By the time I stopped it and put in the ASD 2.6.3 DVD, it was up to the low 80s. It was clearly a slow runaway - there is obviously not enough cooling to keep CPU B operational.

What I would like to try is disabling CPU B and see if the result is thermally stable. I have read various discussions of how to accomplish this, but none of them seem to work on my system. Any pointers?

I am now attempting thermal calibration - we shall see what happens there. I did reset the PRAM for whatever value that may deliver.

A few photos are in order:

First, my hybrid "FrankenCooler" completed but not yet filled - the long overarching hose is the fill/bleed line, clamped off when done:

Next, the same cooler, now filled. This is really just a "technology trial", so I used distilled water as the coolant, with a a small amount of the blood red commercial coolant I plan to use when all of this is working, which I added to give the coolant a color, so that I could see air bubbles.

Finally, a photo during the long and arduous process of bleeding the air bubbles out of the system:

All in all, I loaded about 220 ml of coolant fluid into the loop, but even so, there are still some visible air bubbles.

Since the result is not cooling CPU A to a "nice" temperature (45 C or below is my target), nor cooling CPU B in a stable way, I am guessing that I did not do a good enough job bleeding the residual air from the loop, even though I was at it for several hours. I would welcome tips on how to bleed out all the air! Thanks in advance.

Next up: thermal calibration, which is running as I type this.

 

[MisterKeeks]

First of all, nice job.

Thermal calibration is not going to be of use here.

Your hose is nearly kinked exiting into what I believe is CPU B. That’s going to cause problems. Give it a little more length (easier said than done, I know).

I gotta say, I’m still concerned about the residue the coolant is leaving in your machine; especially so after your pump failure. I didn’t notice anything like that with the auto stuff I used.

 

[mac57mac57]

Thanks @MisterKeeks, I noticed that too. I tried lengthening the run... the kink got more pronounced. Shorter was better, but certainly not perfect. Good catch though... I'll see if I can find any way to reduce the restriction there. I kept all the "useful" Apple hose segments (right angles, U-bends, etc.). I'll see what I can cook up!

I agree that thermal calibration is not likely to fix anything, but it is worth doing just to see if it completes. To date, except for once, it has not. No error messages... the machine just dies.

 

[MisterKeeks]

The risk you run with a thermal calibration is cooking the CPU. It really can only hurt the system. Yes, it should shut itself off before it gets too hot. Why test that functionality.

I’m also not sure where the calibration data is stored, but if something bad happens, like a thermal shutoff or power loss while it’s writing data, you could have much bigger problems on your hands.

If you have a CPU going to 80C, you’ll need to fix the cooling system anyway, at which point you’d need to rerun thermal calibration.