Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
First results from Air Quad Rev 1:

1/ Plug the machine in, and Diagnostic LED #1 (CPU A Installed Incorrectly) comes on immediately.

2/ Press the front panel Power On button and the machine starts, all the fans spin (including the FROZN ones), but there is no startup chime.

3/ Shortly thereafter, Diagnostic LED #2 (CPU Overheat) and Diagnostic LED #7 (CPU Checkstop) come on.

4/ Left like that for a few minutes, the Apple fans ramp up to full speed, as is usual when the machine isn't working properly.

It goes without saying that this is not the result I was hoping for!

It means that I need to disassemble the FROZN coolers, check the CPU cards and their motherboard pin grid very carefully, and try reassembling. For now, I have no information at all about whether this new cooling system cools the machine or not - the machine simply won't boot.

I don't know if I can draw any conclusions from the fact that Diagnostic LED #6 (CPU B Installed Incorrectly) does not come on. Is this because CPU B is properly installed, or is it simply because the diagnostic process doesn't get that far, stopping at CPU A? I just don't know.
 
After repasting CPU A, the machine did come up and run, but CPU temperatures were not controlled and it check stopped a while thereafter. While it was running however, it was lovely - quiet as the proverbial church mouse!

I have tried numerous short term fixes to keep it running (adding an extra fan, for example @joevt!) but it was all to no avail.

My guess is that the twist-tied front FROZN mounting bracket is just not tight enough, and there is not a good thermal seal between the CPU and the FROZN. You called it @Doq, and you may well have been right. The twist tie "jank" may be just a little too "janky" to work. We shall see...

I have torn the Quad back apart and am trying one alternate twist tie arrangement. If that does not work, I am going to have to "bite the bullet" and remove the motherboard. This will allow me to get at its underside and remove the four mounting posts that are in the way of the FROZN mounting hardware. Then I will be able to get a clean, properly screwed down heat sink attachment.

The Apple service manual shows that extracting the motherboard is rather like extracting the LCS - a long and arduous process with almost millions of screws and connectors to remove. I am not looking forward to that, hence the "one last twist tie attempt".

I have to wonder sometimes if a cool, quiet Quad is worth the effort and aggravation this is causing me. I am now into my third solid month of continuous labor on this matter, all in an attempt to get a Quad whose fans do not deafen me while I use it! It shouldn't be this hard!
 
I have torn the Quad back apart and am trying one alternate twist tie arrangement. If that does not work, I am going to have to "bite the bullet" and remove the motherboard. This will allow me to get at its underside and remove the four mounting posts that are in the way of the FROZN mounting hardware. Then I will be able to get a clean, properly screwed down heat sink attachment.

The Apple service manual shows that extracting the motherboard is rather like extracting the LCS - a long and arduous process with almost millions of screws and connectors to remove. I am not looking forward to that, hence the "one last twist tie attempt".
Before committing to logic board removal, see if you can unfasten the posts from the above side. I wasn't able to do it on mine but you might have better luck. You will of course now have exposed screws but that should still be enough clearance to get the coolers and cards seated properly.
 
I tried my final "twist tie" solution; it did not do the trick. The Checkstop light goes on immediately and the machine will not boot. I did get a few startup chimes out of it initially, but now it will not chime either, which is consistent with the Checkstop light going on right away. No CPU = no startup chime.

SO... I am either doing a full motherboard removal, OR I am having some custom mounting hardware made by a neighbor who has a metal shop in his garage. I have drawn out a detailed diagram of what I need - I will check with the neighbor and see if he can build such a thing; else it is onward to motherboard removal. Sigh...

After several cycles through the air cooling setup, I am now out of Kryonaut Extreme. It was a small tube even though it was over $US 20 - it handled about two full applications of two CPUs and two shims. I am back to my trusty Noctua NT-H1 paste, of which I have a much larger tube.

So, not a lot of luck with my "Air Quad", but even though the fans ARE running at 3200 RPM, my liquid cooled Quad is fully operational - I am typing and submiting this post from it, using Aquafox 2.2 and Sorbet Leopard. It continues to run CPU A in the high 30s and CPU B in the mid to high 40s. Excellent values all around... but at 3200 RPM!
 
"Magic Smoke" ... never let the "magic smoke" out of chips, or else they stop working.

Regrettably, today while testing my latest air cooling mod, I smelled the unmistakeable scent of burning electronics... magic smoke! The machine has not given a startup chime since. I fear something got "smoked", for reasons I do not understand. The machine MAY be toasted...
 
I talked to a neighbor about his metal working capabilities. With his garage-based shop, he can build the custom front A400 mounting bracket I need, allowing me to bypass the existing motherboard-based alignment pegs for the CPU cards. Better still, those alignment pegs can now be used to screw down the CPU cards in the front. Up until now in my air cooling mods, they have just been screwed down in the rear.

I am going to have him make a dozen or so of these. If you decide you want to try air cooling your Quad, just PM me and I will ship you the parts at cost... it takes two (one per CPU) per Quad.
 
Decision made. I have purchased two plain sheets of 1/8" aluminum from which we will machine two new front mounting brackets for the A400 FROZN units. This is what we will be building as an initial prototype:

2.Custom FROZN Front Mounting Bracket, v0.1.jpg

The metal sheets arrive Saturday and so all further work on air cooling goes on hold until that time.

If these brackets work out as intended, I will be happy to provide them at cost to any other MacRumors member who wants to also try air cooling their Quad.
 
i look forward to seeing this get finished and working properly. i just got my hands on a quad but the liquid cooler refused to work properly which made it lock up before it even got a chance to chime.
 
Status update on the air cooling effort. Regrettably, the "magic smoke" I reported on a few posts ago clearly smoked something important. The machine will not boot at all now - when the power button is pressed, the red Checkstop light comes on immediately, and a few seconds later the power cuts off.

I have examined everything I can see, but I cannot detect any damaged components.

I will proceed with building my custom front mounting bracket and try it out, but at this point, this is an exercise in mechanical fit and finish - even if it is perfect, the machine itself is quite dead at this point. After a suitable break, I will tear the machine all the way down, including removing the motherboard, to see if I can detect and correct whatever got smoked, but I don't hold out great odds of success. I think I am down one soldier.

This was my original Quad. I have one other Quad. See my thread on "cooling a hot quad" (URL immediately below) for ongoing updates on that machine.

The Quad Chronicles - Restoring a Hot G5 Quad

I will post here in this thread on the results of the custom mounting bracket, just for everyone's future reference.
 
Last edited:
Good news on this thread at last. I exchanged the CPU cards between my two Quads, created a new FROZN mounting schema and that did it. The Air Cooled Quad came right up, and with excellent results:

2025-05-24, CPU Temperatures.jpg


2025-05-24, Fan Speeds.jpg


As you can see, both CPUs are now idling in the high 30s, and fan speeds are completely controlled... the fans on the FROZN units are doing all the heavy lifting of course, so the Apple ones are just loafing.

I am out of time for today, so I will post the "magic" that got me here tomorrow. I neither had to create custom FROZN mounting hardware nor completely disassemble the machine to remove the motherboard so that I could get rid of the alignment pegs that keep the CPU cards in place. Instead, I found a way of reusing those pegs, along with the mounting hardware that came with the FROZN units, to get the job done.

The result is not perfect however... I have some degree of graphical "color corruption" showing up on screen from time to time. This is clearly caused by the CPU cards, since the same color oddities existed on my liquid cooled Quad when these CPU cards were in it. I have checked them over thoroughly, but I cannot find anything that looks wrong. However, the problem traveled from the liquid cooled Quad to the air cooled Quad when the CPU cards were exchanged, so I am pretty sure the cause of the problem lies with the CPU cards.

Finding new CPU cards will be ... challenging ... to say the least!

More details, and photos, tomorrow.
 
Last edited:
Regrettably, today while testing my latest air cooling mod, I smelled the unmistakeable scent of burning electronics... magic smoke! The machine has not given a startup chime since. I fear something got "smoked", for reasons I do not understand. The machine MAY be toasted...
Good news on this thread at last as well. I exchanged the CPU cards between my two Quads, created a new FROZN mounting schema and that did it. The Air Cooled Quad came right up, and with excellent results:
So the machine that got smoked was not toasted - only its CPU cards were damaged? Can you mix the cards up to see if it was one card or both that was damaged? Can a Quad work with only one CPU card installed? I wonder if it can be repaired.
 
Exactly, @joevt. Whatever got smoked was on one of the two CPU cards, since with my "other" Quad's CPU cards installed instead, it worked just fine (setting aside the occasional color distortion).

I have two Quads, which equals four CPU cards. I am guessing that only one of the two from the newly air cooled Quad got "smoked". Further, for my "other" Quad, I am guessing that only one of its two CPU cards is dead.

So, I *may* have two good CPU cards spread across two Quads, but to find out who is who would take a lot of [extract, swap and repaste] iterations. All that handling of the CPU cards significantly increases the risk of damaging them further, and so I don't think I will try this approach.

Instead, I will put up with the occasionally "off" colors until I can lay hands on two known good CPU cards, and then I will do a one time swap. In the meantime, the machine runs, the CPUs are nicely cool, and it is REALLY quiet.

I will deal with it for now... I am just so darn chuffed at successfully air cooling a Quad that a bit of color madness seems a small price to pay.
 
I have been reading that Apple used to support a software package that allowed you to enable/disable each of the Quad's processors and run the machine as a G5 Dual 2.5 GHz. Does anyone reading this know what that software is, and where I could get it? Thanks!
 
I have been reading that Apple used to support a software package that allowed you to enable/disable each of the Quad's processors and run the machine as a G5 Dual 2.5 GHz. Does anyone reading this know what that software is, and where I could get it? Thanks!
I don't know about that software.

I use the following in Open Firmware to make my Quad G5 bootable:
setenv boot-args cpus=3

It's either a coincidence that the CPU that it disables is the one that causes problems on my Quad G5 or that disabling any one core is sufficient.

Probably I should take it apart and reapply thermal paste or something so it can be a Quad again instead...
 
Good news on this thread at last as well. I exchanged the CPU cards between my two Quads, created a new FROZN mounting schema and that did it. The Air Cooled Quad came right up, and with excellent results:
Congrats. You might have just done it better that I did.

I mean.... at least your coolers aren't upside-down 💀
 
Congrats. You might have just done it better that I did.

Thanks @Doq, you were the inspiration - you blazed the trail. You should "take a bow"!

I could never have accomplished this without both your written post information and the comments you provided back to me over time as I progressed through this. Thanks again!
 
As promised, full details today on how I went from "its dead, Jim" to an operational air-cooled Power Mac G5 Quad.

To start with, the "magic smoke" incident I reported earlier in this thread clearly smoked one or both of the CPU cards on the Quad in question, a fact I confirmed when I swapped the CPU cards between the LCS-cooled Quad and the Quad I planned to air cool. After that swap, the LCS-cooled Quad no longer booted, and when I finally succeeded with the air cooling, the onscreen color artifacts that had showed up pretty consistently on the LCS-cooled Quad followed the CPU cards and moved with them from the LCS Quad to the "Air Quad".

SO... one or both of those CPU cards has been damaged. To work around this "smoke issue", I set both of the cards aside and proceeded with installing the CPU cards from the LCS Quad into the Air Quad.

For the Air Quad, the key issue all along has been the front mounting hardware for the FROZN units. The back mounting hardware presents no issues - just put a small bolt through each of the two holes in the CPU card nearest the heat pipes and use them to screw down the front bracket. Easy.

Well, nearly easy anyway. When you do this, you will quickly realize that there are some rather tall (for a chip) surface mount components near the holes you need to use, and thus a tall plastic washer is needed to ensure that when the nut is added, the front bracket rests above the nearby components. I used hard white plastic (non-conductive!) washers that I had in stock, probably a quarter of an inch tall.

The photo below shows the back mounting hardware now in place. The white plastic washers are underneath the bracket, and so are not visible.

2025-04-29.1322, The Two CPU Cards in Place, With Rear A400 Mounting Brackets in Place.jpg


So, that is the back mounting hardware in place and secured.

The front mounting hardware has been the key issue to overcome all along. Unless this is as tightly connected as the back mounting hardware, the FROZN heat sink doesn’t mate tightly enough to the CPU face and effective cooling simply doesn’t happen.

At first glance, this doesn’t look particularly challenging. There are four more holes near the front of each CPU card (two sets of two holes, one on each side of the card), and the inner set of holes fits the positioning requirements for the front mounting brackets almost perfectly. Looks like a slam dunk, but … Apple put those holes there for a reason! Four threaded pegs exist on the motherboard the slide up into these holes when the CPU cards are put into place and (a) ensure the correct physical alignment of the cards relative to the motherboard and (b) tighten down the front of the card, securing it in place - that is why each peg is threaded. These pegs are clearly visible in the photo above.

SO... Apple is already using the set of CPU card holes that the front mounting hardware needs. What to do????

Faced with the same problem, @Doq opted to tear down the Quad completely, remove the motherboard and then, with access to the motherboard’s underside, remove the four offending pegs entirely. He then added a nut/bolt arrangement similar to what was done for the back mounting hardware and thus, after reassembling the Quad, could now secure the front mounting hardware to the CPU card.

The only issue with all of this, other than the enormous amount of work required to completely tear down a Quad, is that once done, there was no way to secure the front of each CPU card – the now removed pegs had provided that function prior to their removal. Since the CPU cards are screwed down in the back, @Doq decided that this was good enough and did not attempt any further efforts to secure the front of the card. As it happens, his “air quad” worked out quite well, and so this turned out to have been a very effective decision.

For my part, after reading the Apple service manual for the Quad, it became clear (to me at least) that extracting the motherboard from the case was something akin to brain surgery, and I quickly decide to try almost ANYTHING to avoid having to do that.

I have a huge collection of nuts, bolts, screws, clamps etc., but I could not find ANYTHING that would screw onto the threads present in each of the pegs… nothing! They must be a very unusual size – classic Apple engineering: why use an industry standard size when you can throw everyone off by creating something proprietary! :rolleyes:

As readers of this thread will know, I attempted a variety of twist-tie based solutions to secure the front mounting hardware, but none of them were able to hold the hardware in place well enough to create a good thermal seal for the heat sinks that screw down onto it. So I decided to create a custom front mounting bracket with the help of a friend who has a metal working shop in his garage.

Then, an “aha!” moment. The mounting hardware included with the FROZN units is very complete and includes four knurled top nuts. I will refer to these as “knurls” for the rest of this post. These knurls turned out to be the lifeboats I needed. They will screw tightly onto the top of the motherboard pegs, but they will only screw down a tiny amount from the top before stopping and they cannot be tightened further. I tried them as a solution early on and dismissed them at that time – they fit the top of the peg just fine but provided no way to securely screw down the mounting hardware that had to sit below.

FROZN Mounting Hardware.jpg

The ”aha”? What if I built up the bottom of each peg, after it had come through the hole on the CPU card, with enough rubber washers such that the front mounting bracket was raised up into near contact with the top of the peg, close enough that the knurl would provide adequate tightening for it? That was “aha” number one.

I tried it, and let me tell you, it is a delicate dance – too many washers and you raise the mounting bracket up so high that the heat sink will not contact the CPU face, but too few washers and the mounting bracket ends up too low, and the mounting hardware on the FROZN unit cannot reach down far enough to grab onto the threaded screw post in the center of the bracket.

There was another complication too. The holes in the edges of the mounting brackets are too small to fit over and through the CPU/motherboard pegs. Hence the bracket would not go all the way down the peg to reach the washers. A complication? Yes. A big one? No. I got out my trusty electric drill, selected a drill bit that was intended for metal and was of the right diameter (I just eyeballed that one) and enlarged the hole at each end of the bracket. When done, the hole was large enough that the bracket would go all the way down the peg to the washers. The result looked like this:

2025-05-09.1012, Front Mounting Bracket Drilled Out to Fit Quad CPU Card Alignment Pegs.jpg


With trial and error, I found that three 1/16” rubber washers almost did it. With three such washers at the base of peg (on top of the CPU card hole), The mounting bracket almost reached up to the knurl, but there was a just a little bit of “wiggle” in it. With the knurls fully tightened down, here is what the result looked like:

2025-05-24.1528, Front and Rear Mounting Hardware In Place.jpg


“Aha” number two? This wiggle room didn’t matter – in fact it was exactly what was needed.

With three washers used, the front mounting bracket was just marginally too low to allow the front FROZN mounting hardware to reach down and grab the threaded bracket screw it tightens into. But leveraging that little bit of wiggle, I could tilt the FROZN unit forward slightly, lowering the front mounting hardware just enough to enable it to grab the front bracket screw and tighten down onto it, and then tilt the FROZN the other way, causing the now attached front mounting hardware to pull the front mounting bracket up to the small extent of its wiggle room. This provided just enough “wiggle room” in the other direction to enable the FROZN’s back mounting hardware to reach its bracket screw and tighten down.

Mission accomplished! The FROZN unit was being held firmly in place against the CPU/shim setup on the CPU card AND the card itself was being more or less held in place by the pegs and the knurl. In retrospect, if I had tried four washers each, it might have been even a little bit more effective, but it worked well with three, and since it works, I am not messing with it!

I repeated the above with the second FROZN and then attached the FROZN fans. As documented earlier in this thread, I rely on a combination of the mounting clips provided with the FROZN units, twist ties and duct tape to secure the fans. The fan closest to the top is properly secured on the side nearest the PCI Divider plate with a mounting clip. The far side of the fan case is secured to its paired FROZN with duct tape. The fan closest to the bottom of the Quad is secured on top entirely with duct tape, while its bottom is twist tied to the nearest edge of the top fan. The result looks like this:

2025-05-24.1723, Completed Air Quad.jpg


If you follow the wires in the above photo really carefully, you may just see that the FROZN’s fans are powered from a MOLEX that I split out from the DVD/CD area of the case and then pulled through a gap near the bottom of the PCI Divider plate. This is the only MOLEX I am aware of in the machine. Of course, the FROZN units do not directly consume MOLEX; they have traditional fan plugs. However, I had a MOLEX to fan plug conversion cable and that bridged the gap.

If I had chosen not to go with MOLEX, I could have (per @Doq,) cut into the Quad’s pump control cable (it is now extraneous after all - there is no pump anymore!) and extracted the +5v, +12v and Ground needed from there.

So, that is it! That is the story of how I got from “its dead, Jim” to a fully air cooled Quad! Lots of niggly little details, but the key issue to solve was attaching the front mounting hardware, and the knurls provided by FROZN turned out to be the answer.

I am still pinching myself; all four CPU cores in the 30s … I have never seen such a thing, nor did I even hope for such an excellent result. I think I am breathing pretty rarified air as well – as far as I know, not very many people have successfully pulled this off. I am not sure how many other people in the world have attempted a true air-cooled Quad conversion (the original four 2.5 GHz CPUs being air cooled, not the conversions that end up turning a Quad into a dual 2.3 GHz air cooled machine) but counting @Doq myself, I am only aware of about three. There must be more however, somewhere.

In the end, air cooling a Quad is SO much simpler than overhauling a failing LCS and less expensive to boot. It takes a little persistence and ingenuity, but I strongly recommend it. Never again will you have to contend with coolant leaks, dried coolant crystals, roaring fans and so on. The LCS may have been a good cooling decision in 2005, but we can do better two decades later in 2025. The results speak for themselves:

2025-05-24, CPU Temperatures and Fan Speeds.jpg


As I did with the LCS Overhaul work, I am busy writing a complete document laying out exactly how to accomplish an air-cooled Quad. When done, I will post it here, and load it onto my retro-computing.com web site as well.
 

Attachments

  • 2025-04-29.1316, A400 Rear Mounting Bracket Component Stack, Focused on the Stack Components.jpg
    2025-04-29.1316, A400 Rear Mounting Bracket Component Stack, Focused on the Stack Components.jpg
    283 KB · Views: 23
Last edited:
  • Like
Reactions: Doq and wkw
How do temperatures behave under a sustained multicore load?

I don't know yet. I'll let you know once tested. I just got to having the machine come up and run yesterday, and then ran out of time for the day.

Over the next few days I will load it up with apps, data, music and video, and then put it through it's paces. I am not all that concerned with artificial benchmark results for CPU temps, but rather real world day to day stuff like Photoshop, iTunes and VLC.

I'll post results as I get them.
 
  • Like
Reactions: wkw
Some more informal test results, obtained using GeekBench, GraphicConverter and others. I have been unable to push CPU temperatures above 54 C no matter what I throw at it. I am very pleased with this!

p.s.> During a several hundred GB Carbon Copy Cloner cloning operation, I saw CPU A hit 61.1 C. That is the highest I have seen since moving to air cooling.
 
Last edited:
  • Like
Reactions: wkw
An update on the onscreen video color corruption I have been experiencing.

I use the following in Open Firmware to make my Quad G5 bootable:
setenv boot-args cpus=3

I tried this, and "2" was the magic number. At 2 CPUs, the onscreen color corruption disappeared entirely. As far as I can tell, what this has done is disable CPU B, so now I know who my culprit is, and now I only need ONE new CPU card to get this Quad to full operational status.

In the meantime, I have a G5 "Half Quad" 2.5 GHz solution. I can return it to full Quad status anytime, by simply restoring the boot-args "cpus" argument from 2 to 4, but the net result doesn't look nearly as good.
 
  • Like
Reactions: wkw
Some additional oddity... now that the Air Quad is fully operational, I decided to swap the hard drives from my LCS Quad into my Air Quad and thus achieve a fully loaded HDD without having to spend all the time doing the loading. The main drive from the LCS Quad is a 4 TB spinner, partitioned into two 2 TB partitions (to allow for a limitation of the G5 Quad that no partition may be larger than 2.2 TB).

I did the physical install, but the Air Quad simply doesn't "see" the drive at all. Drive Utility is blissfully unaware of it, and hence there is nothing I can do with it. I added in another drive, an old 750 GB drive I had lying around, and it boots just fine, even though it is not supposed to... it uses a GUID Partition Map table vs. an Apple Partition Map table. According to everything I have read, you cannot boot a PowerPC Mac from a GUID partition... and yet here it is booting!

To straighten this out, as I type this post, I am cloning the booting partition to an external eSATA HDD, and then will confirm that this drive is bootable. Once that is the case, I will boot from it and repartition the 750 GB HDD to Apple Partition Map and then clone back. I am hoping that this may allow Mac OS X to "see" the 4 TB spinner. If nothing else, it will make me feel better about no longer booting from an unsupported partition type!
 
  • Like
Reactions: wkw
I added in another drive, an old 750 GB drive I had lying around, and it boots just fine, even though it is not supposed to... it uses a GUID Partition Map table vs. an Apple Partition Map table. According to everything I have read, you cannot boot a PowerPC Mac from a GUID partition... and yet here it is booting!
In the "List of Open Firmware versions" at
https://forums.macrumors.com/thread...l-work-in-a-beige-power-macintosh-g3.2303689/
it shows GPT is supported after Open Firmware 4.8.9f4 March 23, 2005 (exclusive) and before Open Firmware 4.9.4f1 July 12, 2005 (inclusive).
For 64bit PPC Macs, that is after Open Firmware 5.1.8f7 October 26, 2004 (exclusive) and before Open Firmware 5.2.7f1 September 30, 2005 (inclusive).
I am looking for ROM dumps for combinations of Mac model/Open Firmware version that do not exist in that list or are marked as - = don't got it

You want APM formatted if you want to boot the disk with different Power Macs.
 
Well @joevt, this opens up a can of worms!

I think you are suggesting that if I upgrade to the latest OF version my 4 TB spinner will then be recognized? Since my other Quad had no issues with seeing the spinner, this makes sense... it must be running a later version of OF.

Since I cannot find an OF reference guide however, I need to know how to:

- Query my current OF version

- Find the last OF version for the Late 2005 G5 Quad

- Find a way to flash that new firmware onto my machine

Would you happen to know the answers to any of the above?

Finally, I am worried that if anything goes wrong during flashing, I will "brick" the machine I have just so recently coaxed back to life...

All that having been said, I remember that someone here at MacRumors posted a lengthy set of OF commands (not a full set, just what they could find). I will hunt that down and see if it answers any of my questions above.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.